WO2024053728A1

WO2024053728A1 - Subject state prediction and application of same

Info

Publication number: WO2024053728A1
Application number: PCT/JP2023/032811
Authority: WO
Inventors: 伸手嶋; 光將栗田; 耕志山本; 典子西川; 大樹神山; 信孝服部
Original assignee: 住友ファーマ株式会社; 学校法人順天堂
Priority date: 2022-09-09
Filing date: 2023-09-08
Publication date: 2024-03-14

Abstract

The present disclosure provides a method by which it is possible to estimate the state of a subject with ease. A method for estimating the state of a subject of the present disclosure comprises: (A) a step for acquiring the value of a parameter pertaining to the blinking of the subject; and (B) a step for estimating the state of the subject on the basis of at least the value of the parameter pertaining to the blinking. In one embodiment, parameters pertaining to blinking include at least one or a combination of eye closure speed, peak eye closure speed, eye opening speed, peak eye opening speed, eye closure time, eye opening time amplitude, eye closure time amplitude, blinking continuation time, and blink count. In a preferable embodiment, the parameters pertaining to the blinking include at least one of blinking confidence, blinking interval, blinking depth, and blinking energy.

Description

Prediction of target state and its application

The present disclosure relates to a method of estimating the state of a target based on the value of a parameter related to blinking. Furthermore, the present disclosure also relates to methods of evaluating therapeutic or preventive drugs or other medical techniques for a subject.

Parkinson's disease is one of the progressive neurodegenerative diseases whose main symptoms are abnormalities in extrapyramidal function. Pathologically, dopamine neuron loss and alpha-synuclein deposition in the substantia nigra pars compacta are observed. Clinically, it exhibits various motor symptoms such as akinesia, resting tremor, muscle rigidity, and loss of postural reflexes.

The basis of Parkinson's disease treatment is drug therapy aimed at replenishing dopamine in the brain, and drugs containing the dopamine precursor levodopa (L-DOPA, levodopa) are used as the first-line drug for the initial treatment of Parkinson's disease. be done. However, as the disease progresses, in almost all patients treated with levodopa, circadian fluctuations in Parkinson's symptoms, levodopa-induced dyskinesia (PD-LID) in Parkinson's disease (hereinafter referred to as "PD-LID") Motor complications such as dyskinesia (sometimes referred to as dyskinesia) occur.

Typical symptoms of diurnal fluctuations include wearing-off, on-off, no-on, and delayed-on phenomena. There is. Among them, wearing-off is caused by, as mentioned above, when the ability to retain dopamine in the synaptic cleft decreases as the disease progresses, the dopamine concentration in the brain fluctuates depending on the blood concentration of levodopa, and as a result, the blood concentration of levodopa is below the safe therapeutic range, and the duration of the effect of levodopa is shortened.

The incidence of PD-LID within 5 years after starting levodopa treatment is 30-50%, and increases as the disease progresses, reaching 50-100% within 10 years after starting treatment. Peak-dose dyskinesia is known as a typical symptom of PD-LID, and it is an involuntary disorder that appears on the face, tongue, neck, limbs, trunk, etc. when the blood concentration of levodopa is high. It's exercise.

The condition of Parkinson's disease patients being treated with L-DOPA is often not known until dyskinesia actually occurs. On the other hand, it is known that the amount of L-DOPA in the brain correlates with the onset of dyskinesia, but it is not practical to measure the amount of L-DOPA in the brain in real time, so it is not realistic to measure it in real time. However, no method has been provided to estimate the state of Parkinson's disease in real time.

The present disclosure was completed by discovering that the value of a parameter related to blinking serves as an index for estimating the state of a subject. The subject is preferably a Parkinson's disease patient or a patient suspected of having Parkinson's disease, and more preferably a Parkinson's disease patient undergoing treatment with L-DOPA, an L-DOPA-related compound, or a dopamine agonist.

Accordingly, the present disclosure provides:

(Item 1)
A method of estimating a state of a target, the method comprising:
A) obtaining the value of a parameter related to the subject's blink;
B) estimating the state of the object based on at least the value of the parameter related to the blink.
(Item 2)
The parameter related to blinking is at least one of the following: eye-closing speed, eye-closing peak speed, eye-opening speed, eye-opening peak speed, eye-closing time, eye-opening amplitude, eye-closing amplitude, degree of eye opening, blink duration, and number of blinks. or a combination according to item 1.
(Item 3)
The method according to item 1 or item 2, wherein the blink-related parameters include at least one of blink confidence, blink interval, blink depth, and blink energy.
(Item 4)
The method according to item 1, wherein the blink-related parameters include blink confidence and blink interval.
(Item 5)
The step B) includes estimating the state of the subject based on the elapsed time after administering L-DOPA, an L-DOPA-related compound, or a dopamine agonist to the subject and the blink-related parameter. , the method described in any one of items 1 to 4.
(Item 6)
The method according to any one of items 1 to 5, wherein the condition includes a condition indicated by an index that can be used for clinical evaluation.
(Item 7)
The condition is according to any one of items 1 to 6, including a condition indicated by at least one of the presence or absence of dyskinesia, ON-OFF, MDS-UPDRS Part III score, UDysRS score, and plasma levodopa concentration. Method described.
(Item 8)
8. The method according to any one of items 1 to 7, wherein the subject is a Parkinson's disease patient being treated with L-DOPA, an L-DOPA-related compound, or a dopamine agonist.
(Item 8A)
9. The method of any one of items 1-8, wherein estimating the state of the object includes predicting a future state of the object.
(Item 9)
A method of estimating a state of a target, the method comprising:
A) acquiring eyeball information of the subject;
B) A method comprising the step of estimating the condition of the subject based on the eyeball information and the elapsed time after administering L-DOPA, an L-DOPA-related compound, or a dopamine agonist to the subject.
(Item 10)
A method for estimating the condition of a Parkinson's disease patient being treated with L-DOPA, an L-DOPA-related compound, or a dopamine agonist, the method comprising:
A) obtaining a blink confidence value of the patient;
B) estimating the condition of the patient based on the blink confidence value and the elapsed time after administering L-DOPA, an L-DOPA-related compound, or a dopamine agonist to the patient; The method, wherein the condition includes at least one of the presence or absence of dyskinesia, ON-OFF, MDS-UPDRS Part III score, UDysRS score, and plasma levodopa concentration.
(Item 11)
The step A) further includes obtaining at least one value of the patient's blink interval, blink energy, blink duration, and number of blinks,
The step B) is based on the value of the blink confidence, the elapsed time, and the value of at least one of the blink interval, blink energy, blink duration, and number of blinks. The method according to item 10, comprising estimating a state of.
(Item 12)
A system for estimating the state of a target,
means for acquiring the value of a parameter related to the blink of the subject;
A system comprising: means for estimating a state of the object based on at least a value of a parameter related to the blink.
(Item 12A)
13. The system of item 12, comprising the features described in one or more of the above items.
(Item 13)
A program for estimating a state of a target, the program being executed in a computer including a processor, the program comprising:
A) obtaining the value of a parameter related to the subject's blink;
B) A program that causes the processor to perform a process including: estimating the state of the object based on at least the value of the parameter related to the blink.
(Item 13A)
14. The program according to item 13, comprising the features described in one or more of the above items.
(Item 13B)
A computer-readable storage medium storing the program according to item 13 or 13A.
(Item 14)
A method of evaluating a therapeutic or preventive drug or other medical technology for a subject, the method comprising:
A) obtaining the value of a parameter related to the subject's blink;
B) calculating an estimated effective amount or effective level or usage or administration of the therapeutic or prophylactic drug or other medical technique based on at least the value of the blink-related parameter.
(Item 15)
The method according to item 14, further comprising determining a therapeutic or prophylactic drug or other medical technique recommended for the subject based on the calculated estimated effective amount or effective level or dosage or administration. .
(Item 15A)
15. A method according to item 14, comprising the features described in one or more of the above items.
(Item 16)
A system for evaluating therapeutic or preventive drugs or other medical techniques for a subject,
means for acquiring the value of a parameter related to the blink of the subject;
and means for calculating an estimated effective amount or effective level or usage or administration of the therapeutic or prophylactic drug or other medical technique based on at least the value of the blink-related parameter.
(Item 16A)
17. The system of item 16, comprising the features described in one or more of the above items.
(Item 17)
A program for evaluating therapeutic or preventive drugs or other medical techniques for a subject, the program being executed on a computer including a processor, the program comprising:
A) obtaining the value of a parameter related to the subject's blink;
B) Calculating an estimated effective amount or effective level or usage or dosage of the therapeutic or prophylactic drug or other medical technology based on at least the value of the blink-related parameter. ,program.
(Item 17A)
18. The program according to item 17, comprising the features described in one or more of the above items.
(Item 17B)
A computer-readable storage medium storing the program according to item 17 or 17A.
(Item 18)
A target health management method,
Performing the method described in any one of items 1 to 11,
determining whether or not to treat the subject based on the results of the method;
If it is determined that treatment should be performed on the subject, taking action for health management of the subject.
(Item 19)
The action includes performing the treatment on the target, issuing an alert that the treatment should be performed on the target, and administering a predetermined drug or therapy to the target. The method according to item 18, comprising at least one of the following.
(Item 20)
A target health management system,
An estimation system configured to perform the method described in any one of items 1 to 11;
means for determining whether or not to treat the subject based on the output from the estimation system;
A system comprising means for taking action for health management of the subject when it is determined that the subject should be treated.
(Item 20A)
21. The system of item 20, comprising the features described in one or more of the above items.
(Item 21)
A target health management program, the program being executed on a computer comprising a processor, the program comprising:
Performing the method described in any one of items 1 to 11,
determining whether or not to treat the subject based on the results of the method;
A program that causes the processor to perform processing including: issuing an instruction to take action for health management of the subject when it is determined that treatment should be taken for the subject.
(Item 21A)
22. The program according to item 21, comprising the features described in one or more of the above items.
(Item 21B)
A computer-readable storage medium storing the program according to item 21 or 21A.

In the present disclosure, it is intended that one or more of the features described above may be provided in further combinations in addition to the specified combinations. Still further embodiments and advantages of the present disclosure will be recognized by those skilled in the art upon reading and understanding the following detailed description, as appropriate.

The present disclosure allows the state of a target to be easily estimated. In one example, the present disclosure can provide a method of estimating the condition of a Parkinson's disease patient or a patient suspected of having Parkinson's disease, etc., which is practically possible.

Flowchart of how to estimate or predict the state of a target Flowchart of a method for evaluating therapeutic or prophylactic drugs or other medical techniques for a subject Diagram showing an example of the configuration of the system 10 A diagram showing an example of the structure of a neural network 20 for constructing a learned model that can be used by the estimation/prediction means 12. Diagram showing an example of the configuration of system 10A A diagram showing an example of the configuration of a system 1000 of the present disclosure A diagram showing an example of the configuration of the user device 100 A diagram showing an example of the configuration of a server device 200 A data flow diagram showing data exchange between the user device 100 and the server device 200 in one embodiment A data flow diagram showing data exchange between the user device 100 and the server device 200 in one embodiment Diagram schematically showing the test procedure of Example 1 Diagram showing an example of data measured from one patient Diagram showing the results of estimation performance evaluation for unknown data Diagram showing the results of estimation performance evaluation for unknown data Diagram showing the results of estimation performance evaluation for unknown data Diagram showing the results of determining the presence or absence of dyskinesia for unknown data and learning data of patient #20 Diagram showing the results of ON/OFF determination for unknown data and learning data of patient #3 and patient #20

The present disclosure will be described below. Throughout this specification, references to the singular should be understood to include the plural unless specifically stated otherwise. Accordingly, singular articles (e.g., "a," "an," "the," etc. in English) should be understood to also include the plural concept, unless specifically stated otherwise. Further, it should be understood that the terms used herein have the meanings commonly used in the art unless otherwise specified. Accordingly, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. In case of conflict, the present specification (including definitions) will control.

(definition)
Terminology and general techniques used herein are explained.

As used herein, "about" means ±10% of the following numerical value.

In this specification, "subject" is used in the same meaning as "subject" and "subject", and refers to a human being for whom the treatment, diagnosis, or test of the present disclosure is performed, and the subject is suffering from a disease. is used synonymously with "patient". The subject is preferably a Parkinson's disease patient undergoing treatment with L-DOPA, L-DOPA related compounds or dopamine agonists.

As used herein, "dopamine or a substance biologically equivalent to dopamine" refers to dopamine or any substance that has a similar function to dopamine in vivo. One such function is as a neurotransmitter, and there are three main neural pathways through which dopamine acts: the nigrostriatal tract, mesolimbic tract, mesencephalic tract, and the nigrostriatal tract. Since the striatal tract is associated with Parkinson's disease, this specification specifically refers to the biological equivalent of dopamine as referred to herein, if it can be said that it is biologically equivalent in relation to the nigrostriatal tract. It corresponds to an equivalent substance.

In this specification, substances biologically equivalent to dopamine include, for example, levodopa (L-3,4-dihydroxyphenylalanine (IUPAC name is (S)-2-amino-3-(3,4-dihydroxyphenyl) ) propanoic acid), also called L-dopa), L-3,4-dihydroxyphenylalanine, bromocriptine, pergolide, talipexole, cabergoline, pramipexole, ropinirole, rotigotine, apomorphine, etc. but is not limited to these.

As used herein, "L-DOPA, L-DOPA-related compounds, or dopamine agonists" refers to the pharmacological effects of L-DOPA through metabolism in the body when administered to a living body (for example, a Parkinson's disease patient). For example, it refers to a substance that exhibits the biological function of "dopamine or a substance biologically equivalent to dopamine," and includes L-DOPA, L-DOPA-related compounds, and dopamine agonists. Ru.

As used herein, the term "dopamine agonist" includes, but is not limited to, L-3,4-dihydroxyphenylalanine, bromocriptine, pergolide, talipexole, cabergoline, pramipexole, ropinirole, rotigotine, apomorphine, and the like.

In this specification, "L-DOPA" and "levodopa" are synonymous.

As used herein, "a compound related to L-DOPA" is different from L-DOPA but has a biological function equivalent to that of L-DOPA (particularly, for example, an activity related to the treatment of Parkinson's disease). say something For example, L-DOPA related compounds include, but are not limited to, esters of L-3,4-dihydroxyphenylalanine and salts thereof. Examples of esters of L-3,4-dihydroxyphenylalanine are levodopa ethyl ester (LDEE; ethyl (2S)-2-amino-3-(3,4-dihydroxyphenyl)propanoate), levodopapropyl ester; Propyl (2S)-2-amino-3-(3,4-dihydroxyphenyl)propanoate), levodopa methyl ester (methyl (2S)-2-amino-3-(3,4-dihydroxyphenyl)propanoate), etc. The ester of L-3,4-dihydroxyphenylalanine can be a salt, including, for example, a hydrated salt. The salts of levodopa ester are octanoate, myristate, succinate, succinate dihydrate, fumarate, fumarate dihydrate, mesylate, tartrate and hydrochloride. may include, but are not limited to. For example, the succinate or succinate dihydrate of the ester of L-3,4-dihydroxyphenylalanine is levodopaethyl ester succinate (LDEE-S) or levodopaethyl ester succinate dihydrate (LDEE-S). -S-dihydrate or LDEE-S(d)). Compounds related to L-DOPA include prodrugs and depot agents of L-DOPA. L-DOPA related compounds may overlap with dopamine agonists, but may be classified herein as such in either case.

As used herein, "adjunct" refers to a drug other than the drug that has the main effect, and in this specification, if levodopa is the main drug, tandospirone etc. are adjuvants. Can be done. As used herein, the term "L-DOPA adjunct" refers to a drug that suppresses the metabolism or decomposition of L-DOPA or regulates the release of dopamine derived from L-DOPA into the synaptic cleft. In the present specification, the daily dosage of levodopa, which is the main drug, is the usual dose for levodopa treatment as described in the 2018 version of the Parkinson's Disease Clinical Practice Guidelines or the corresponding guidelines in the United States and Europe. Generally, the usual daily dose of levodopa is 50 to 1200 mg/day, preferably 100 mg to 600 mg/day, in combination or as a combination drug with peripheral dopa decarboxylase inhibitor (DCI). It is day. For example, SINEMET® (Carbidopa-Levodopa combination tablets) (New Drug Application (NDA) #017555) approved by the FDA has a 1:4 ratio combination tablet (Carbidopa 25mg-Levodopa 100mg), and 1:10 It is provided as a ratio combination tablet (Carbidopa 10mg - Levodopa 100mg, Carbidopa 25mg - Levodopa 250mg). SINEMET (registered trademark) is administered as a daily maintenance dose from 70 mg to 100 mg of Carbidopa, and SINEMET (registered trademark) is administered as a maximum daily dose of 200 mg as Carbidopa.

Examples of levodopa auxiliaries include levodopa metabolic enzyme inhibitors and dopamine release controlling agents.

As used herein, "levodopa metabolic enzyme inhibitor" refers to any drug that has the effect of inhibiting the metabolism of levodopa so as to enhance its action in a broad sense. Dopa decarboxylase (decarboxylase) inhibitors (DCI) (carbidopa, α-methyldopa, benzerazide (Ro4-4602), α-difluoromethyl-DOPA (DFMD), or salts thereof, etc.) are exemplified. ), as well as catecholamine-O-methyltransferase inhibitors (COMT-I) (entacapone is an example), which prevent levodopa from being broken down before it enters the brain, and drugs that prevent dopamine from being broken down in the brain. Examples include monoamine oxidase inhibitors (MAO-I) (selegiline is an example). A "dopamine release controlling agent" refers to a drug that has the effect of controlling dopamine release from dopamine nerves, and examples thereof include zonisamide, amantadine, tandospirone, buspirone, istradefylline, and the like.

As used herein, the term "dopamine nerve function restoring agent" refers to any drug or treatment method aimed at restoring the function of dopamine nerves and/or preventing or suppressing their degeneration and loss, and refers to substances that have dopamine nerve degeneration effects. Examples include vaccines or antibody drugs against alpha-synuclein, and nucleic acid drugs that affect its expression.

As used herein, the term "dopamine-producing cell drug" refers to a drug that treats Parkinson's disease through cell transplantation of dopamine neurons. Examples include cell transplantation of fetal midbrain tissue and cell transplantation of dopaminergic neural progenitor cells derived from human induced pluripotent stem cells (iPS cells).

As used herein, "dopamine-producing gene therapy" refers to a treatment method in which the enzyme gene necessary for dopamine synthesis is introduced into the putamen using a viral vector. An example is a gene therapy method in which an aromatic amino acid decarboxylase (AADC) gene is introduced into an adeno-associated virus (AAV).

In this specification, "surgical therapy" refers to deep brain stimulation (DBS) to the subthalamic nucleus, internal globus pallidus, and thalamus, stereotactic destruction of the thalamus and globus pallidus, and focused ultrasound. Surgical treatment methods such as Focused Ultrasound (FUS) can be mentioned.

In this specification, when a certain "amount" is referred to, it may include not only an "absolute amount" in its narrow sense, but also a "relative amount" in its broad sense. The "relative amount" includes, for example, the amount of change (for example, the amount of increase, the amount of decrease).

As used herein, "level" means an approximate degree, even if the amount cannot be specified. These levels refer to the degree of effect (action), and include Movement Disorder Society-Unified Parkinson's Disease Rating Scale (MDS-UPDRS), UPDRS, Unified Dysk inesia Rating Scale (UDysRS), Clinical Dyskinesia Rating Scale (CDRS), The Rush Dyskinesia Rating Scale (Rush DRS), Abnormal Involuntary Movement Scale (AIMS), EuroQol 5 Dimensions (EQ- 5D-5L), PDQ-39 (Parkinson's Disease Questionnarie-39), Clinical Global Impressions (CGI), Patient Global It can be expressed using a clinical evaluation scale such as PGI (PGI), a patient diary, or a scale calculated from locomotion information acquired by a wearable device such as an accelerometer and/or angular velocity meter.

As used herein, "variation" refers to a change over time or a value that changes over time. The value that changes over time is also called the amount of change.

As used herein, "in vivo dopamine or a substance biologically equivalent to dopamine" refers to dopamine or a substance biologically equivalent to dopamine that exists in a certain organ, organ (including blood, etc.), or cell. refers to a substance that

As used herein, "dopamine in the brain or a substance biologically equivalent to dopamine" refers to dopamine existing in the brain or a substance biologically equivalent to dopamine. Dopamine in the brain or substances biologically equivalent to dopamine can be determined by measuring the amount of dopamine using microdialysis, mass spectrometry imaging (MSI), drugs that undergo metabolism similar to L-DOPA, and dopamine transporters. , Single Photon Emission Computed Tomography (SPECT) preparations and Positron Emission Tomography (PET) of drugs that have affinity for dopamine receptors. Measured by methods such as brain functional imaging using preparations can do.

As used herein, "eye-closing speed" refers to the average speed of upper eyelid movement from the moment the upper eyelid begins to move to close the eye during a blink to the moment when the upper eyelid movement is considered to have stopped. .

In this specification, "eye closing peak speed" refers to the maximum speed of upper eyelid movement from the moment the upper eyelid begins to move to close the eye during a blink to the moment when the upper eyelid movement is considered to have stopped. say.

As used herein, "eye opening speed" refers to the average speed of upper eyelid movement from the moment the upper eyelid begins to move to open the eye during a blink to the moment when the upper eyelid movement is considered to have stopped. .

In this specification, "peak eye opening speed" refers to the maximum speed of upper eyelid movement from the moment the upper eyelid begins to move to open the eye during a blink to the moment when the upper eyelid movement is considered to have stopped. say.

As used herein, "eye-closing time" refers to the time from the moment the upper eyelid finishes moving to close the eye during a blink to the moment the upper eyelid begins to move to open the eye.

As used herein, "eye-opening amplitude" refers to the distance that the upper eyelid moves from the moment the upper eyelid begins to move to open the eye during a blink to the moment when the upper eyelid movement is considered to have stopped. .

In this specification, "eye closure amplitude" refers to the distance that the upper eyelid moves from the moment the upper eyelid begins to move to close the eye during a blink to the moment when the upper eyelid movement is considered to have stopped. .

In this specification, "eye opening degree" is an index representing the relative position of the upper eyelid. For example, the relative position of the upper eyelid when the eyes are completely open is 0, and when the eyes are completely closed, the relative position of the upper eyelid is 0. The relative position of the upper eyelid can be expressed when the relative position of the upper eyelid is set to 1.

As used herein, "blink duration" refers to the time from the moment when the upper eyelid begins to move to close the eye during a blink to the moment when the upper eyelid is considered to have stopped moving to open the eye. say. "Blink duration" = (eye-opening amplitude/eye-opening speed) + (eye-closing amplitude/eye-closing speed).

In this specification, the "number of blinks" refers to the number of blinks in a certain time window.

In this specification, "parameters related to blinks" refer to parameters that can be obtained from blinks. Parameters related to blinking are typically among the following: eye closing speed, eye closing peak velocity, eye opening speed, eye opening peak velocity, eye closing time, eye opening amplitude, eye closing amplitude, degree of eye opening, blink duration, and number of blinks. at least one or a combination of the following. The combination of at least some of the following: eye-closing velocity, eye-closing peak velocity, eye-opening velocity, eye-opening peak velocity, eye-closing time, eye-opening amplitude, eye-closing amplitude, interblink time, blink duration, and number of blinks, for example, It can be expressed as blink confidence, blink interval, and blink energy. Preferably, the blink-related parameters include at least one of blink confidence, blink interval, blink energy, and blink depth. More preferably, the blink-related parameters include blink confidence and blink interval. The blink-related parameters may include calculated values of eye-closing speed, eye-closing peak speed, eye-opening speed, eye-opening peak speed, eye-closing time, eye-opening amplitude, eye-closing amplitude, degree of eye opening, blink duration, and number of blinks. . The calculated value is, for example, the calculated value of the value within each classification when data is divided by a predetermined threshold (for example, a time threshold (time window), an absolute value threshold, a relative value threshold). ) may be included. The intra-class calculation values include, but are not limited to, various statistics such as maximum value, minimum value, average value, median value, variance, 25% percentile value, 75% percentile value, and frequency analysis spectrum. The calculated values may include, for example, calculated values of intra-section calculated values (inter-section calculated values) of a plurality of sections. Inter-section calculation values include, but are not limited to, weighted sums, differences, time differentials, time integrals, ratios, correlation coefficients, covariances, and standard deviation parameters obtained by Poincaré plot analysis. Here, the standard deviation parameter obtained by Poincaré plot analysis is the one in which the method used for heartbeat analysis is applied to blink analysis. In other words, the horizontal axis represents the time between the nth blink and the n+1st blink (the "blink interval" described below), and the vertical axis represents the time between the n+1st blink and the n+2nd blink. The time BI(n+1) between BI(n+1) and BI(n+1) is plotted as a scatter diagram, and the standard deviations in the vertical direction and parallel direction to the straight line of BI(n)=BI(n+1) are defined as SD1 and SD2, respectively, as standard deviation parameters. Furthermore, ratios such as SD1/SD2 or SD2/SD1, and values obtained through arithmetic processing such as calculating S=SD1×SD2 as the area of a virtual ellipse drawn by a plot can also be used as parameters.

In this specification, "blink confidence" refers to the relative position of the upper eyelid when the eyes are completely open as 0, and the relative position of the upper eyelid when the eyes are completely closed as 1. The area on the curve / (area on the curve + area under the curve) when the horizontal axis is the time from the start to the end of one blink and the vertical axis is the relative position of the upper eyelid. It can be expressed as

In this specification, "blink interval" refers to the time between one blink and the next blink, and the relative position of the upper eyelid when the eyes are fully open is 0, completely If the relative position of the upper eyelid when the eyes are closed is set to 1, then the difference between the moment when the relative position of the upper eyelid reaches its maximum value and the next moment when the relative position of the upper eyelid reaches its maximum value. It can be expressed in terms of time.

In this specification, "blink energy" is a parameter that simulates the energy or amount of work used in a blink, and is the reciprocal of the number of blinks and the average blink duration or the power thereof in a certain time window. can be represented by the product of

In this specification, "blink depth" refers to a parameter obtained by dividing blink confidence by blink duration.

In this specification, "eyeball information" refers to information regarding eyeballs. The eyeball information includes, for example, information regarding eyeball movements. As used herein, "eyeball movement" refers to all movements related to the eyeballs. Eye movements may include, for example, eye movements, changes in eye conditions, changes in electrooculography, and the like. It is included in eye movement even when the eyeballs are stationary. Examples of eyeball information include, for example, information regarding blinks, information regarding pupil coordinates or the relative position of the eyeballs, and information regarding pupil diameter. The information regarding blinking is synonymous with the value of the parameter regarding blinking. Blinks include, for example, spontaneous blinks, voluntary blinks, and reflex blinks. Information regarding blinks includes, but is not limited to, for example, blink frequency or number of blinks, blink duration, inter-blink time, eye closure time, eye opening speed, eye closing speed, eyelid movement width, and eyelid opening degree. For example, information about pupillary coordinates or the relative position of the eyeballs may include movement distance, movement direction, velocity, acceleration, angular velocity, saccades, gliding eye movements, vestibulo-ocular reflexes, convergence/divergence, fixation micromovements (ocular tremor, drift). , microsaccades), amount, or level.

Information regarding blinks and information regarding pupil coordinates or relative positions of eyeballs may also include calculated values of these information. The calculated value may include, for example, a function output (power, logarithm, distribution, frequency analysis value, etc.) using the information as an input variable. The calculated value is, for example, the calculated value of the value within each classification when data is divided by a predetermined threshold (for example, a time threshold (time window), an absolute value threshold, a relative value threshold). ) may be included. The intra-class calculation values include, but are not limited to, various statistics such as maximum value, minimum value, average value, median value, variance, 25% percentile value, 75% percentile value, and frequency analysis spectrum. The calculated values may include, for example, calculated values of intra-section calculated values (inter-section calculated values) of a plurality of sections. Inter-section calculation values include, but are not limited to, weighted sums, differences, time differentials, time integrals, ratios, correlation coefficients, covariances, and standard deviation parameters obtained by Poincaré plot analysis.

As used herein, "Parkinson's disease patient being treated with L-DOPA" refers to a Parkinson's disease patient who is being treated with L-DOPA, including MDS-UPDRS, UPDRS, UDysRS, CDRS, Rush DRS, Symptoms are measured using clinical evaluation scales such as AIMS, EQ-5D-5L, PDQ-39, CGI, and PGI, patient diaries, and scales calculated from locomotion information acquired by wearable devices such as accelerometers and/or gyrometry. can be confirmed. On the other hand, "untreated Parkinson's disease patients" are Parkinson's disease patients who have not received treatment such as administration of L-DOPA, and although symptoms of Parkinson's disease can be confirmed, the symptoms have progressed. Refers to patients for whom it is decided not to start treatment until In addition, "patients suspected of having Parkinson's disease (prodromal PD)" are patients who have not been definitively diagnosed with Parkinson's disease, but who may have some of the symptoms or signs of Parkinson's disease. A patient who can confirm the

As used herein, the term "therapeutic drug" refers to a drug for curing a target disease or alleviating symptoms. In particular, when referring to patients with Parkinson's disease, it refers to drugs used to cure Parkinson's disease or alleviate symptoms. Clinically, clinical evaluation scales such as MDS-UPDRS, UPDRS, UDysRS, CDRS, Rush DRS, AIMS, EQ-5D-5L, PDQ-39, CGI, PGI, patient diary, accelerometer and/or gyro meter, etc. Symptom improvement can be confirmed using a scale calculated from movement information acquired by a wearable device.

As used herein, the term "prophylactic drug" refers to a drug for preventing aggravation of a subject's disease or preventing symptoms (such as dyskinesia) from occurring. In particular, when referring to patients with Parkinson's disease, it refers to drugs used to prevent Parkinson's disease from worsening or to prevent symptoms (such as dyskinesia) from occurring. Clinically, clinical evaluation scales such as MDS-UPDRS, UPDRS, UDysRS, CDRS, Rush DRS, AIMS, EQ-5D-5L, PDQ-39, CGI, PGI, patient diary, accelerometer and/or gyro meter, etc. The preventive effect can be confirmed by the scale etc. calculated from the movement information acquired by the wearable device.

As used herein, "medical technology" refers to technology for performing some medical treatment, and may be provided in any form such as pharmaceuticals, medical devices, regenerative medicine products, surgical therapy, etc.

As used herein, "effective amount" or "effective level" refers to an amount or level effective for treating or preventing a specific disease. In particular, it refers to an amount or level effective for the treatment or prevention of Parkinson's disease. In the present specification, the daily dosage of levodopa, which is the main drug, is the usual dose for levodopa treatment as described in the 2018 version of the Parkinson's Disease Clinical Practice Guidelines or the corresponding guidelines in the United States and Europe. Generally, the usual daily dose of levodopa is 50 to 1200 mg/day, preferably 100 mg to 600 mg/day, in combination or as a combination drug with peripheral dopa decarboxylase inhibitor (DCI). It is day. For example, SINEMET® (Carbidopa-Levodopa combination tablets) (New Drug Application (NDA) #017555) approved by the FDA has a 1:4 ratio combination tablet (Carbidopa 25mg-Levodopa 100mg), and 1:10 It is provided as a ratio combination tablet (Carbidopa 10mg - Levodopa 100mg, Carbidopa 25mg - Levodopa 250mg). SINEMET (registered trademark) is administered as a daily maintenance dose from 70 mg to 100 mg of Carbidopa, and SINEMET (registered trademark) is administered as a maximum daily dose of 200 mg as Carbidopa.

As used herein, "the effect of a drug that has an improving effect on L-DOPA, L-DOPA-related compounds, or dopamine agonists" refers to It refers to the effect that a drug has, and can be measured by various techniques such as the following.

That is, the improving effect on levodopa-induced dyskinesia of Parkinson's disease in the present disclosure is clinically based on MDS-UPDRS, UPDRS, UDysRS, CDRS, Rush DRS, AIMS, EQ-5D-5L, PDQ-39, CGI, PGI, etc. This can be confirmed using a clinical evaluation scale, a patient diary, a scale calculated from locomotion information acquired by a wearable device such as an accelerometer and/or a gyrometer, etc. Furthermore, in non-clinical PD-LID model rats or PD-LID model monkeys, the dyskinesia improving effect can be confirmed by evaluating dyskinesia-like abnormal involuntary movement behavior. By using this method, it is possible to measure the improvement, progression suppression, or prevention of levodopa-induced dyskinesia (PD-LID) symptoms, as well as the shortening of the onset time of levodopa-induced dyskinesia (PD-LID).

The states of Parkinson's disease that can be measured by the present disclosure include akinesia, bradykinesia, tremor, muscle rigidity, and loss of posture, which are the main motor symptoms. of postural reflexes, floxed posture, freezing, sleep disorders, mental/cognitive/behavioral disorders, autonomic nervous disorders, sensory disorders, and any motor complications.

As used herein, "motor complications" refers to any motor symptoms observed in patients with advanced stage Parkinson's disease that are a problem in treatment, and includes diurnal fluctuations in Parkinson's symptoms and problems associated with levodopa treatment. Examples include dyskinesia (levodopa-induced dyskinesia (PD-LID)), which is a voluntary movement. Motor complications are interpreted to be caused by abnormal release of dopamine in the brain, but the mechanism is not necessarily clear.

As used herein, "circadian fluctuations of parkinsonian symptoms" refers to wearing-off, on-off phenomenon, no-on phenomenon, delayed-on phenomenon, delayed on) phenomenon is known. Among these, wearing-off occurs when the ability to retain dopamine in the synaptic cleft decreases as the disease progresses, causing dopamine concentration in the brain to fluctuate in accordance with the blood concentration of levodopa, and as a result, the blood concentration drops below the safe treatment range. This is a symptom that occurs when the duration of the effect of levodopa is shortened.

As used herein, "levodopa-induced dyskinesia <involuntary movements> (PD-LID)" refers to involuntary movements (dyskinesia) induced by excessive administration of levodopa, which occurs within 5 to 10 days after starting treatment with levodopa. It occurs in more than half of patients with Parkinson's disease at older ages, and the proportion (%) of patients affected by PD-LID is increasing over time (for a review, see, for example, Encarnacion and Hauser, (2008), Levodopa-induced dyskinesias in Parkinson's disease: etiology, impact on quality of life, and tre ”, Eur Neurol, 60(2), pp. 57-66).

Dyskinesia is a condition in which a part of the body moves on its own and does not stop, bites the lip, has difficulty speaking, cannot sit still, or has difficulty moving the limbs as desired, including limbs and/or orofacial areas and/or body axis. It is a movement disorder in which involuntary movements of parts are seen. Peak-dose dyskinesia is known as a typical symptom of PD-LID, and symptoms appear on the face, tongue, neck, limbs, trunk, etc. when the blood concentration of levodopa is high. . PD-LID is more likely to appear if levodopa is continued to be taken in larger doses than necessary from the early stage of the disease, and once PD-LID appears, it is difficult to control it even if the dosage of levodopa is adjusted in various ways. It is known that it is difficult.

As used herein, "an index that can be used for clinical evaluation" refers to an index that can be used to evaluate the symptoms of a subject, and typically includes subjective evaluation items. "Indicators that can be used for clinical evaluation" include clinical evaluation scales such as MDS-UPDRS, UPDRS, UDysRS, CDRS, Rush DRS, AIMS, EQ-5D-5L, PDQ-39, CGI, PGI, and patient diaries. , but not limited to.

(Preferred embodiment)
Preferred embodiments of the present disclosure will be described below. It is understood that the embodiments provided below are provided for a better understanding of the present disclosure, and the scope of the present disclosure should not be limited to the following description. Therefore, it is clear that those skilled in the art can refer to the description in this specification and make appropriate modifications within the scope of the present disclosure. It is also understood that the following embodiments of the present disclosure can be used alone or in combination.

Note that the embodiments described below are comprehensive or specific examples. The numerical values, shapes, materials, components, arrangement positions and connection forms of the components, steps, order of steps, etc. shown in the following embodiments are merely examples, and are not intended to limit the scope of the claims. Further, among the constituent elements in the following embodiments, constituent elements that are not described in the independent claims indicating the most significant concept will be described as arbitrary constituent elements.

(Method of estimating or predicting the state of the target)
In one aspect of the present disclosure, a method for estimating or predicting the state of a target, a computer program for realizing the same or a storage medium storing the same, and a system or user equipment forming part of the system are provided. Ru. In this aspect, a method is provided for estimating or predicting the state of a subject.

FIG. 1A shows a flowchart of a method for estimating or predicting the state of a subject. As used herein, an "estimated" state refers to a past, present or future state. "Estimate" of the future state is sometimes expressed as "prediction", and "estimate", which is contrasted with "prediction", is sometimes interpreted narrowly as "estimation" of the past or present state.

The "state" estimated or predicted in this aspect may be any state associated with blinking. Conditions include, for example, conditions indicated by indicators that can be utilized in clinical evaluation, and typically include conditions associated with Parkinson's disease. The condition associated with Parkinson's disease is, for example, at least one of the conditions indicated by at least one of the presence or absence of dyskinesia, ON-OFF, MDS-UPDRS Part III score, UDysRS score, and plasma levodopa concentration. include. Clinical scores such as the MDS-UPDRS Part III score and the UDysRS score are shown as the sum of several subscores. For example, the MDS-UPDRS Part III score is based on individual symptoms such as rigidity, tremor, postural disturbance, and bradykinesia. It is composed of the total score of subscores indicating the severity of symptoms. Further, the UDysRS score is composed of the total score of subscores indicating the site of onset of dyskinesia and dystonia, severity, and degree of impact on daily life. Conditions associated with Parkinson's disease may include not only the total clinical score, but also at least one of these subscores. Since the "subject" targeted by the present disclosure is a human, it is possible to use an index that can be used for clinical evaluation. In particular, indicators that can be used for clinical evaluation that have subjective evaluation items are unique to the fact that the "subject" is a human.

In step S101, target eyeball information is acquired. Preferably, the value of a parameter related to the subject's blink is acquired.

In one embodiment, eyeball information (preferably the value of a parameter related to blinking) can be measured using a device capable of acquiring eyeball information of a target. Devices capable of acquiring eyeball information include, for example, eyeglass-type devices equipped with an image recording function, an eye tracking function, a three-point electrooculogram sensor, an acceleration sensor, and/or a gyro sensor, eye tracking glasses, a smartphone, a personal computer, This includes, but is not limited to, an electro-oculography measuring device and the like. Preferably, the device capable of acquiring eyeball information is a device capable of measuring the value of a parameter related to blinking, for example, an eye tracking device.

The measurement time per measurement is determined to be long enough for eyeball information to be measured relatively stably, and may range from 30 seconds to 16 hours per measurement. More preferably, it is in the range of 5 minutes to 8 hours. During the measurement, the subject may be engaged in unrestricted daily life, or may be performing a task to measure specific eye movements. Examples of tasks include following an indicator that suddenly disappears or appears on a display as a visual stimulus to induce a saccade, or following a smoothly moving indicator to induce a gliding eye movement. Examples include chasing. At this time, an accelerometer, etc. attached to the arm or head may be used in conjunction with the movement of the head or arm to evaluate eyeball information, or to remove the movement element of the head. A device to immobilize the head and jaw may also be used. Subjects may also complete a patient diary to record subjective symptoms.

In some embodiments, time intervals or "time windows" of digital data are extracted from time-series data. For example, the width of the time window is in the range of 10 seconds to 30 minutes, more preferably 1 minute to 10 minutes, and even more preferably 3 minutes.

In some embodiments, data from an eye-tracking device is filtered to remove noise and extract data of interest. For eyeblink data, it is possible to exclude abnormally long eyeblink events, which typically do not occur in human eyeblinks, such as eyeblink durations greater than 1000 ms, but can be excluded in patients with Parkinson's disease, etc. If it is assumed that the user has an abnormality in blinking, the exclusion criterion can be made more relaxed, for example, 2000 milliseconds. Regarding eye movement data, for example, when focusing on eye tremor, eye movements with large movement distances such as saccades become noise that interferes with frequency analysis, so time window data that includes data points with large movement distances are excluded in advance. be able to.

In some embodiments, eye movements are obtained as discrete pupil coordinate data, but data points may be smoothly interpolated with a spline curve or the like and upsampled to obtain sufficient temporal resolution.

In some embodiments, the measured values regarding the user's eyeballs may be standardized or corrected based on the user's past measured values or the measured values of other users.

In some embodiments, the measurements regarding the user's eyes may be based on body movements or activity changes due to the user's exercise estimated by an accelerometer or the like, or physiological states such as drowsiness estimated from eye movements or heart rate variability. may be standardized or corrected.

In a preferred embodiment, the blink-related parameters include eye-closing velocity, eye-closing peak velocity, eye-opening velocity, eye-opening peak velocity, eye-closing time, eye-opening amplitude, eye-closing amplitude, inter-blink time, eye-opening degree, blink duration, and At least one or a combination of blink counts is obtained. As a parameter related to blinking, at least one of the following: eye-closing speed, eye-closing peak speed, eye-opening speed, eye-opening peak speed, eye-closing time, eye-opening amplitude, eye-closing amplitude, inter-blink time, blink duration, and number of blinks. Two calculated values may be obtained. In a more preferred embodiment, at least one of blink confidence, blink interval, blink depth, and blink energy is acquired as the blink-related parameter. In a more preferred embodiment, a blink confidence and a blink interval are obtained as blink-related parameters.

In a preferred embodiment, in step S101, in addition to the subject's eyeball information, the elapsed time after administering the drug (more specifically, L-DOPA, L-DOPA-related compound, or dopamine agonist) to the subject is indicated. Information can also be obtained.

In step S102, the state of the target is estimated or predicted based on at least the eyeball information acquired in step S101. Preferably, the state of the object is estimated or predicted based on the value of a parameter related to blinking.

In one embodiment, the state of the subject can be estimated or predicted from the correlation between the eye information of the subject and the state of the subject. In particular, the state of the target can be estimated or predicted from the correlation between the value of the parameter related to the blink of the target and the state of the target. Specifically, by inputting the value of at least one appropriately selected parameter among blink-related parameters into a trained model as a feature, the state of the target can be estimated or Can be predicted. At this time, the learned model has learned the relationship between the value of at least one of the blink-related parameters and the state of the object.

In a preferred embodiment, the condition of the subject can be estimated or predicted from the correlation between the subject's eyeball information and the elapsed time after administering the drug to the subject, and the subject's condition. In particular, the condition of the subject can be estimated or predicted from the correlation between the value of the parameter related to the blink of the subject and the elapsed time after administering the drug to the subject, and the condition of the subject. Specifically, the value of at least one appropriately selected parameter related to blinking and the elapsed time after administering the drug to the subject are input as features into the trained model, and the learning is performed. The state of the target can be estimated or predicted from the output of the model. At this time, the learned model has learned the relationship between the value of at least one of the blink-related parameters, the elapsed time after administering the drug to the subject, and the state of the subject.

At least one of the blink-related parameters includes eye-closing speed, eye-closing peak speed, eye-opening speed, eye-opening peak speed, eye-closing time, eye-opening amplitude, eye-closing amplitude, blink duration, number of blinks, and blink confidence. , blink interval, and blink energy, preferably at least one of blink confidence, blink interval, and blink energy, and more preferably, blink confidence, blink interval, and blink energy. Can be confidence and blink interval.

At least one parameter among the blink-related parameters may be appropriately selected depending on the estimated or predicted state of the target. For example, when estimating or predicting the presence or absence of dyskinesia as a target condition, parameters that highly contribute to prediction accuracy, such as blink confidence, blink count, blink duration, etc., may be selected. If the elapsed time after administering the drug to the subject is also used for estimation or prediction, the selected blink-related parameters may vary. For example, when estimating or predicting the presence or absence of dyskinesia as a condition of a subject based on the elapsed time after administering a drug to the subject and parameters related to blinking, parameters that have a high contribution to prediction accuracy, such as blinking. Confidence, blink frequency, blink energy, etc. may be selected.

The condition of the subject estimated or predicted in step S102 can be applied to the subject's health management method. In the target health management method, after estimating or predicting the condition of the target in steps S101 and S102, based on the result, it is determined whether or not the target should be treated, and the process is performed on the target. This includes taking actions for the health management of the subject when it is determined that it is necessary.

Actions for health management of a target include, for example, administering a treatment to the target, issuing an alert that a treatment should be performed to the target, administering a prescribed drug or therapy to the target. Contains at least one of the following. Here, treatment refers to some kind of medical treatment, and typically may refer to intervention by a doctor.

For example, an alert that a treatment should be performed on a subject may be issued from a user terminal owned by the subject, or may be issued from a terminal device of a doctor who is examining the subject. However, it may also be emitted from another device. This allows the subject, physician, or other person to take action to perform treatment on the subject.

(Method of evaluating therapeutic or preventive drugs or medical technology for a target)
In another aspect of the present disclosure, a method for evaluating therapeutic or preventive drugs or other medical techniques for a subject, a computer program for realizing the same, a storage medium storing the same, and a system or a part of the system are provided. Configuring user equipment, etc. are provided. In this aspect, methods are provided for evaluating therapeutic or prophylactic agents or other medical techniques for a subject.

FIG. 1B shows a flowchart of a method for evaluating therapeutic or prophylactic drugs or other medical techniques for a subject.

The "therapeutic drug or preventive drug or other medical technology" evaluated in this aspect may be a therapeutic drug or preventive drug or other medical technology that may have some relation to blinking. Therapeutic or prophylactic drugs or other medical techniques include, for example, L-DOPA or L-DOPA-related compounds, adjuncts of L-DOPA, dopamine nerve function restoring agents, dopamine-producing cell medicines, or dopamine-producing gene therapy. , or surgical therapy (such as deep brain stimulation or stereotaxic destruction).

In step S111, target eyeball information is acquired. Preferably, the value of a parameter related to the subject's blink is acquired. Step S111 is a process similar to step S101, and its explanation will be omitted.

In step S112, an estimated effective amount or effective level of the therapeutic drug, preventive drug, or other medical technology is calculated based on at least the eyeball information acquired in step S111. Preferably, an estimated effective amount or level of a therapeutic or prophylactic drug or other medical technique is calculated based on the value of the blink-related parameter.

In one embodiment, the estimated effective amount or level of the therapeutic or prophylactic drug or other medical technology is determined from a correlation between the subject's ocular information and the estimated effective amount or level of the therapeutic or prophylactic drug or other medical technology. can be calculated. In particular, the estimated effective amount or level of a therapeutic or prophylactic drug or other medical technology is determined by the relationship between the value of the blink-related parameter in question and the estimated effective amount or level of the therapeutic or prophylactic drug or other medical technology. It can be calculated from the correlation. Specifically, by inputting the value of at least one appropriately selected parameter among blink-related parameters into a trained model as a feature quantity, therapeutic drugs, preventive drugs, or Estimated effective amounts or levels of other medical techniques can be calculated. At this time, the trained model has learned the relationship between the value of at least one of the blink-related parameters and the estimated effective amount or effective level of the therapeutic or preventive drug or other medical technology.

In a preferred embodiment, the estimated effective amount or level of the therapeutic or prophylactic drug or other medical technique is based on the subject's ocular information and the elapsed time since administering the drug to the subject, and the therapeutic or prophylactic drug or other medical technique. It can be calculated from a correlation with the estimated effective amount or level of effectiveness of the technology. In particular, the estimated effective amount or level of a therapeutic or prophylactic drug or other medical technique depends on the value of the subject's eyeblink parameters and the elapsed time after administering the drug to the subject and It can be calculated from a correlation with an estimated effective amount or level of medical technology. Specifically, the value of at least one appropriately selected parameter related to blinking and the elapsed time after administering the drug to the subject are input as features into the trained model, and the learning is performed. From the output of the model, an estimated effective amount or level of a therapeutic or prophylactic drug or other medical technology can be calculated. At this time, the trained model calculates the value of at least one of the blink-related parameters, the elapsed time after administering the drug to the subject, and the estimated effective amount or effective amount of the therapeutic or preventive drug or other medical technology. Learning the relationship with level.

At least one of the blink-related parameters includes eye-closing speed, eye-closing peak speed, eye-opening speed, eye-opening peak speed, eye-closing time, eye-opening amplitude, eye-closing amplitude, degree of eye opening, blink duration, number of blinks, It may be at least one of blink confidence, blink interval, blink depth, and blink energy, preferably at least one of blink confidence, blink interval, blink depth, and blink energy. more preferably blink confidence and blink interval.

In step S112, in addition to or in place of the estimated effective amount or level, the effective usage or dosage of the therapeutic or prophylactic drug or other medical technology may be calculated as well. .

The estimated effective amount or effective level or usage or dosage of the therapeutic or preventive drug or other medical technique calculated in step S112 determines the therapeutic or preventive drug or other medical technique recommended for the subject. can be used for.

For example, an estimated effective amount or effective level is calculated for the first therapeutic agent, an estimated effective amount or effective level is calculated for the second therapeutic agent, and an estimated effective amount or effective level is calculated for the nth therapeutic agent. When the effective level is calculated, each estimated effective amount or effective level is compared, and the therapeutic drug with the most appropriate estimated effective amount or effective level for the subject is determined as the therapeutic drug recommended for the subject. Can be done. The most appropriate estimated effective amount or level can be, for example, the lowest amount or level. Similarly, for prophylactic drugs or other medical techniques, the prophylactic drug or other medical technique that has the most appropriate estimated effective amount or level for the subject is the recommended prophylactic drug or other medical technique for the subject. can be determined.

In this way, by evaluating therapeutic drugs, preventive drugs, or other medical techniques for the subject, it is possible to estimate and propose effective therapeutic drugs, preventive drugs, or other medical techniques for the subject. Can be done.

(System for estimating or predicting the state of a target)
The method of estimating or predicting the state of a target described above is performed using, for example, the system 10 of the present disclosure.

The system 10 is configured to estimate or predict the state of the target based on eyeball information of the target, more preferably parameters related to the blink of the target.

FIG. 2 shows an example of the configuration of the system 10. The system 10 includes at least an acquisition means 11 and an estimation/prediction means 12.

The acquisition means 11 is configured to acquire eyeball information of the target. The information acquisition means 11 can acquire eyeball information by any means. The information acquisition means 11 may acquire eyeball information stored in a storage means external or internal to the system 10, or may acquire eyeball information by extracting eyeball information from an eyeball information source. You may also do so. The ocular information source includes optical information (image or reflected light), physical information (vibration, etc.) or electrical information (potential, current, etc.), chemical information, or a combination of multiple thereof, of the subject's eyeball. The ocular information source may be, for example, at least one image taken of the subject's eyeball, electrooculography measurements, or motion information (e.g., acceleration measurements) of the subject's body part. It may be reflected light information from illumination light irradiated onto the eyeball (cornea and/or sclera), current information obtained by a search coil method, or information directly applied to the eyeball. The amount of electromotive force information obtained from a piezo element probe in contact may be used. The reflected light information may be information obtained by a simple sensor such as a photodiode, or may be information obtained by image analysis. The motion information may be, for example, information obtained by image analysis or information obtained by distance measurement.

Preferably, the acquisition means 11 can acquire the value of the parameter related to the subject's blink. The acquisition means 11 can acquire the value of a parameter related to the subject's blinking, for example, from a moving image or a plurality of still images of the subject's eyeball. For this purpose, the acquisition means 11 can communicate with a device capable of acquiring an eyeball image of the object, such as an eye tracking device. Alternatively, the acquisition means 11 itself may be implemented by a device capable of acquiring an eyeball image of a target, such as an eye tracking device.

In an eyeball image, the value of the parameter related to blinking can be obtained based on the pupil or eyelid in the image. For example, when an eyeball is photographed by irradiating the eyeball with infrared rays and then binarized using an appropriate threshold value, the pupil is extracted as an ellipse with relatively low brightness information in the eyeball image. Ellipse fitting is performed on the binarized eyeball image, and if the ellipse fitting is achieved with a certain level of accuracy or higher (an ellipse is recognized in the image), when the eyes are opened, the ellipse fitting is performed with less than a certain level of accuracy (an ellipse is not recognized in the image). ), it can be inferred that the eyes are closed. A blink can be expressed as an event in which the accuracy of the ellipse fitting decreases below a certain level and increases again above the certain level in a time window of 0.25 seconds, for example. Information about blinking can be obtained by using electrodes attached to the top and bottom and/or left and right sides of the eye to capture and analyze the potential difference between the retina and cornea (electrooculography) during eye movement and eyelid movement. is also possible. Information regarding blinks can also be obtained by extracting an electro-oculography waveform characteristic of blinks. Specifically, in the differential waveform of the electrooculogram during blinking, a sharp peak on the negative side first appears, and then a peak on the positive side occurs beyond the baseline. This corresponds to the reciprocating movement of the eyelids, and such a positive/negative reversal of the differential waveform is not seen in typical eye movements.

In a preferred embodiment, the acquisition means 11 includes, in addition to the eyeball information, information indicating the elapsed time after administering the drug (more specifically, L-DOPA, an L-DOPA-related compound, or a dopamine agonist) to the subject. Also get. Information indicating the elapsed time after administering a drug to a target can be obtained, for example, when a user inputs a medication administration time into a user device (e.g., a smartphone, tablet, smart watch, etc.) and receives the input medication administration time from the user device. The elapsed time can be obtained by deriving the elapsed time from the medication administration time. Alternatively, the information indicating the elapsed time after administering the drug to the target can be obtained by, for example, inputting the timing at which the user took the drug into the user device (for example, by pressing a button indicating that the drug has been taken), and It can be obtained by receiving the information from the device and deriving the elapsed time from the timing of taking the medication. Alternatively, information indicating the elapsed time after administering the drug to the subject may be obtained, for example, by receiving a signal indicating the timing of taking the drug from a sensor incorporated in the drug, and deriving the elapsed time from the received signal. can be done.

The eyeball information (values of parameters related to blinking) acquired by the acquisition means 11 is passed to the estimation/prediction means 12. In a preferred embodiment, the information acquired by the acquisition means 11 indicating the elapsed time after administering the drug to the subject is also passed to the estimation/prediction means 12 .

The estimation/prediction means 12 is configured to estimate or predict the state of the target based on the eyeball information. In a preferred embodiment, the estimating/predicting means 12 is configured to estimate or predict the state of the object based on the value of the blink-related parameter.

The estimation/prediction means 12 can estimate or predict the state of the target, for example, using a learned model. The learned model has learned, for example, the relationship between the value of a parameter related to blinking and the state of the object.

FIG. 3 shows an example of the structure of a neural network 20 for constructing a learned model that can be used by the estimation/prediction means 12.

The neural network 20 has an input layer, at least one hidden layer, and an output layer. The number of nodes in the input layer of the neural network 20 corresponds to the number of dimensions of input data. The number of nodes in the output layer of the neural network 20 corresponds to the number of dimensions of output data. For example, when outputting the presence or absence of dyskinesia as the target state, the number of nodes in the output layer is 1 (for example, the output node outputs a value between 0 and 1, and "0" indicates the absence of dyskinesia. ``1'' may indicate the presence of dyskinesia). The hidden layers of neural network 20 can include any number of nodes.

The weighting coefficient of each node of the hidden layer of the neural network 20 may be calculated based on data obtained in advance. The process of calculating this weighting coefficient is the learning process. The learning process may be supervised learning or unsupervised learning.

In the case of supervised learning, for example, the weighting coefficient of each node is calculated so that when the value of a parameter related to the blink of a certain subject is input to the input layer, the value of the output layer will be the value indicating the condition of the patient. can be done. This can be done, for example, by backpropagation. Although it may be preferable to have a large amount of training datasets used for learning, if the amount is too large, overfitting is likely to occur.

For example, the set of (input supervised data, output supervised data) for supervised learning is (value of the parameter related to the blink of the first object, value indicating the state of the first object), (second (value of the parameter related to the blink of the object, value indicating the state of the second object), ... (value of the parameter related to the blink of the i-th object, value indicating the state of the i-th object), ...・etc. Here, the value of the parameter related to blinking may have any dimension. For example, the value of the blink-related parameter may be a value for one blink-related parameter (i.e., one-dimensional) or a value for two blink-related parameters (i.e., two-dimensional). Alternatively, it may be a value for three or more blink-related parameters (that is, three or more dimensions). When a value of a parameter related to blinking acquired from a new object is input to the input layer of such a trained neural network model, a value indicating the state of the object is output to the output layer.

The estimation/prediction means 12 may, for example, utilize a plurality of trained models. For example, the first trained model among the plurality of trained models can output a value indicating the first state of the subject (for example, the presence or absence of dyskinesia), and the first trained model among the plurality of trained models The second trained model can output a value indicating the second state of the patient (for example, plasma levodopa concentration), and the third trained model among the plurality of trained models can output a value indicating the second state of the patient. A value (for example, MDS-UPDRS Part III score) indicating the status 3 can be output. The estimation/prediction means 12 can estimate or predict the state of the object by integrating outputs from a plurality of trained models.

In a preferred embodiment, the estimation/prediction means 12 can estimate or predict the state of the subject based on the value of the parameter related to blinking and the elapsed time after administering the drug to the subject. The estimation/prediction means 12 can estimate or predict the state of the target using the learned model, and at this time, the trained model can estimate or predict the state of the target, for example, based on the values of parameters related to blinking and whether a drug has been administered to the target. It learns the relationship between the elapsed time and the state of the target.

For example, the set (input supervised data, output supervised data) for supervised learning is ((value of the parameter related to the blink of the first subject, elapsed time after administering the drug to the first subject). ), value indicating the state of the first subject), ((value of parameter related to blinking of the second subject, elapsed time after administering the drug to the second subject), indicating the state of the second subject value), ... ((value of the parameter related to the blink of the i-th object, elapsed time after administering the drug to the i-th object), value indicating the state of the i-th object), ..., etc. It can be. If you input the blink-related parameter values obtained from a new subject and the elapsed time after administering the drug to the new subject into the input layer of such a trained neural network model, the state of that subject will be shown. The value is output to the output layer.

The trained model described above is not limited to the neural network model as described, but any other machine learning model can be used. For example, random forests can be used. For example, it is also possible to use a simple model constructed by simply setting weighting coefficients obtained through prior learning.

For example, the estimation/prediction means 12 may estimate or predict the state of the target based on a rule without relying on machine learning.

In one example, a system that predicts changes in a subject's state caused by a drug based on the amount and time course of drug administration based on a pharmacokinetic model, or a system that estimates a subject's state solely from motor symptoms. The techniques of this disclosure can be implemented without using a learning process. Alternatively, you can collect data from a system that was implemented without using a machine learning process to train the model as virtual ground truth values, or use that data as variables in the model or in corrections to model output values. For example, the present disclosure can be implemented by using these systems together and using a trained model.

The state of the object estimated or predicted by the system 10 can be used for any purpose. For example, a doctor can use the condition of the subject estimated or predicted by the system 10 as an index for diagnosis. For example, the subject itself can use the state of the subject estimated or predicted by the system 10 as an index for understanding its own state. Alternatively, the condition of the subject estimated or predicted by the system 10 can be used for health management of the subject. For example, after estimating or predicting the state of the target, the system 10 determines whether or not the target should be processed based on the result, and when it is determined that the target should be processed. , taking actions for the health management of the subject.

Actions for health management of the target include, for example, administering treatment to the target. The system 10 may include means for administering a treatment to a subject, or may issue an instruction to a means external to the system 10 to administer a treatment to a subject.

Actions for health management of a subject include, for example, issuing an alert that a treatment should be performed on the subject. The system 10 may include a means for issuing an alert that a treatment should be taken against the object, or may issue an instruction to a means external to the system 10 to issue an alert. Good too.

Actions for health management of a subject include, for example, administering a predetermined drug or therapy to the subject. The system 10 may include means for administering a predetermined drug or therapy to a subject, or means external to the system 10 may be configured to administer a predetermined drug or therapy to a subject. You may also issue instructions like this.

In one embodiment, the system 10A of the present disclosure can evaluate therapeutic or preventive drugs or other medical techniques for a subject. Evaluating a therapeutic or prophylactic drug or other medical technology includes calculating an estimated effective amount or level of effectiveness of the therapeutic or prophylactic drug or other medical technology for a subject. The estimated effective amount or level allows one to assess whether a therapeutic or prophylactic drug or other medical technique will be useful to a subject.

FIG. 4 shows an example of the configuration of the system 10A.

The system 10A includes at least an acquisition means 11 and a calculation means 13. The configuration of the system 10A is the same as the configuration of the system 10, except that the estimation/prediction unit 12 is replaced by a calculation unit 13. In FIG. 4, the same reference numerals are given to the same components as those described above with reference to FIG. 2, and the description thereof will be omitted here.

The calculation means 13 is configured to calculate an estimated effective amount or effective level of a therapeutic or preventive drug or other medical technique for the subject based on the eyeball information. In a preferred embodiment, the estimating/predicting means 12 is configured to calculate an estimated effective amount or effective level of a therapeutic or prophylactic drug or other medical technique for the subject based on the value of the blink-related parameter. .

The calculation means 13 can calculate the estimated effective amount or effective level of a therapeutic or preventive drug or other medical technology for a subject, for example, using a learned model. The learned model has learned, for example, the relationship between the value of a parameter related to blinking and the estimated effective amount or effective level.

The calculation means 13 can utilize a neural network similar to the neural network 20 described above with reference to FIG.

The weighting coefficient of each node of the hidden layer of the neural network may be calculated based on previously obtained data. The process of calculating this weighting coefficient is a learning process. The learning process may be supervised learning or unsupervised learning.

In the case of supervised learning, for example, when the value of a parameter related to blinking of a certain subject is input to the input layer, the value of the output layer is the estimated effective amount or effective amount of a therapeutic or preventive drug or other medical technology for that patient. A weighting factor for each node may be calculated to be a level. This can be done, for example, by backpropagation. Although it may be preferable to have a large amount of training datasets used for learning, if the amount is too large, overfitting is likely to occur.

For example, the set of (input supervised data, output supervised data) for supervised learning is (value of parameter related to blink of the first target, therapeutic or preventive drug or other medical treatment for the first target). (estimated effective amount or effective level of the technology), (value of parameter related to blinking of the second subject, estimated effective amount or effective level of the therapeutic or prophylactic drug or other medical technology for the second subject),... (the value of the parameter related to the blink of the i-th subject, the estimated effective amount or effective level of the therapeutic or prophylactic drug or other medical technology for the i-th subject), etc. Here, the value of the parameter related to blinking may have any dimension. For example, the value of the blink-related parameter may be a value for one blink-related parameter (i.e., one-dimensional) or a value for two blink-related parameters (i.e., two-dimensional). Alternatively, it may be a value for three or more blink-related parameters (that is, three or more dimensions). When the values of blink-related parameters obtained from a new subject are input into the input layer of such a trained neural network model, the estimated effective amount or effective level of a therapeutic or preventive drug or other medical technology for that subject can be calculated. is output to the output layer.

In a preferred embodiment, the calculation means 13 calculates the estimated effective amount or effective amount of the therapeutic or prophylactic drug or other medical technique for the subject based on the value of the blink-related parameter and the elapsed time after administering the drug to the subject. The level can be calculated. The estimation/prediction means 12 can calculate the estimated effective amount or effective level of a therapeutic or preventive drug or other medical technology for a subject by using the trained model, and at this time, the trained model can calculate, for example, , the relationship between the values of blink-related parameters and the elapsed time after administering the drug to the subject and the estimated effective amount or level of the therapeutic or prophylactic drug or other medical technique for the subject.

For example, the set (input supervised data, output supervised data) for supervised learning is ((value of the parameter related to the blink of the first subject, elapsed time after administering the drug to the first subject). ), estimated effective amount or effective level of the therapeutic or prophylactic drug or other medical technique for the first subject), ((value of the blink-related parameter of the second subject, after administering the drug to the second subject) ), the estimated effective amount or effective level of the therapeutic or prophylactic drug or other medical technology for the second subject), ... the estimated effective amount or effective level of the therapeutic or prophylactic drug or other medical technique for the i-th subject), etc. By inputting the blink-related parameter values obtained from a new subject and the elapsed time since administering the drug to the new subject into the input layer of such a trained neural network model, the therapeutic agent or An estimated effective amount or level of the prophylactic drug or other medical technology is output to the output layer.

For example, the calculating means 13 may calculate the estimated effective amount or effective level of a therapeutic drug, preventive drug, or other medical technology for a subject based on a rule without relying on machine learning.

The estimated effective amount or effective level calculated by the system 10A can be used for any purpose. For example, a doctor can use the estimated effective amount or effective level calculated by the system 10A as an index for prescribing a therapeutic or preventive drug or determining the administration of a medical technique. Alternatively, the estimated effective amount or level calculated by system 10A can be utilized to determine recommended therapeutic or prophylactic agents or other medical techniques for the subject. For example, after calculating the estimated effective amount or level, system 10A can determine a recommended therapeutic or prophylactic drug or other medical technique for the subject based on the results. For example, the calculation means 13 of the system 10A calculates the estimated effective amount or effective level for the first therapeutic agent, calculates the estimated effective amount or effective level for the second therapeutic agent, and... When the estimated effective amount or effective level of the therapeutic drug is calculated, the system 10A compares the estimated effective amount or effective level and selects the therapeutic drug with the most appropriate estimated effective amount or effective level for the subject. can be determined as the recommended treatment for. Similarly for prophylactic drugs or other medical techniques, the system 10A determines which prophylactic drugs or other medical techniques are recommended for the subject, and which have the most appropriate estimated effective amount or level for the subject. can be determined as a medical technology.

(System implementation example)
The systems 10 and 10A described above may be implemented, for example, in a system 1000 that includes a user device and a server device connected via a network.

FIG. 5 is a diagram showing an example of the configuration of the system 1000 of the present disclosure.

The system 1000 includes at least one user device 100, a server device 200 connected to the at least one user device 100 via a network 400, and a database unit 300 connected to the server device 200.

The user device 100 may be any terminal device such as a smartphone, a tablet computer, smart glasses, a smart watch, a laptop computer, a desktop computer, an eye tracker, etc. User device 100 can communicate with server device 200 via network 400. Here, the type of network 400 does not matter. For example, the user device 100 may communicate with the server device 200 via the Internet or may communicate with the server device 200 via a LAN. Although three user devices 100 are depicted in FIG. 5, the number of user devices 100 is not limited thereto. The number of user devices 100 may be any number greater than or equal to one.

The server device 200 can communicate with at least one user device 100 via the network 400. Further, the server device 200 can communicate with a database unit 300 connected to the server device 200.

For example, the database unit 300 connected to the server device 200 may store eyeball information of a plurality of targets acquired in advance. A plurality of pieces of stored eyeball information can be used, for example, to construct a learned model. For example, the database unit 300 may store constructed learned models.

FIG. 6A shows an example of the configuration of the user device 100.

The user device 100 includes a communication interface section 110, an input section 120, a display section 130, a memory section 140, and a processor section 150.

The communication interface unit 110 controls communication via the network 400. The processor unit 150 of the user device 100 can receive information from outside the user device 100 via the communication interface unit 110, and can transmit information to the outside of the user device 100. For example, the processor unit 150 of the user device 100 can receive information from the server device 200 via the communication interface unit 110, and can transmit information to the server device 200. Communication interface section 110 can control communication in any manner.

The input unit 120 allows the user to input information into the user device 100. It does not matter in what manner the input unit 120 allows the user to input information into the user device 100. For example, if the input unit 120 is a touch panel, the user may input information by touching the touch panel. Alternatively, if the input unit 120 is a mouse, the user may input information by operating the mouse. Alternatively, if the input unit 120 is a keyboard, the user may input information by pressing keys on the keyboard. Alternatively, if the input unit 120 is a microphone, the user may input information by voice.

The display unit 130 may be any display for displaying information.

The memory unit 140 stores programs for executing processes in the user device 100, data required for executing the programs, and the like. The memory unit 140 stores, for example, a program for estimating or predicting the state of the object (for example, a program for performing processing including the steps shown in FIG. 1A or a program for implementing the processing shown in FIGS. 7 and 8, which will be described later). ), and/or a part or all of a program for evaluating a therapeutic or preventive drug or other medical technology for a subject (for example, a program for performing processing including the steps shown in FIG. 1B) is stored. There is. The memory unit 140 may store an application that implements an arbitrary function. The memory unit 140 may store, for example, an application for acquiring eyeball information by tracking a pupil in an eyeball image (eye tracking). Thereby, the user device 100 has a portion that implements the eye tracking function. The memory unit 140 may store, for example, an application for acquiring eyeball information from electro-oculogram measurements. As a result, the user device 100 has a portion that implements the function of measuring electro-oculography. In addition, the memory unit 140 may store applications for acquiring eyeball information by other means. Here, it does not matter how the program is stored in the memory unit 140. For example, the program may be preinstalled in the memory unit 140. Alternatively, the program may be installed in the memory unit 140 by being downloaded via the network 400. Memory section 140 may be implemented by any storage means.

The processor unit 150 controls the operation of the user device 100 as a whole. The processor unit 150 reads a program stored in the memory unit 140 and executes the program. This allows the user device 100 to function as a device that executes desired steps. The processor unit 150 may be implemented by a single processor or by multiple processors.

In addition to the configuration described above, the user device 100 can include an imaging means 160. Imaging means 160 is any means for acquiring images. The imaging means 160 is, for example, a camera. Here, "image" includes still images and moving images. The imaging means 160 may be, for example, a built-in camera in the user device 100 or an external camera attached to the user device 100. For example, when the user device 100 is a smartphone, a camera built into the smartphone can be used as the imaging means 160.

In addition to the above-described configuration, or in place of the above-described imaging device 160, the user device 100 can include an electro-oculography acquisition device 170. The electro-oculogram acquisition means 170 is any means for acquiring electro-oculography. The electro-oculography acquisition means 170 is, for example, a plurality of electrodes that can be placed on the skin around the eyes. For example, the electro-oculography acquisition means 170 may be an electrode built into the user device 100 or an external electrode attached to the user device 100. For example, when the user device 100 is a glasses-type eye tracker, electrodes built into the glasses-type eye tracker can be used as the electro-oculogram acquisition means 170.

In the example shown in FIG. 6A, each component of the user device 100 is provided within the user device 100, but the present disclosure is not limited thereto. It is also possible for any of the components of user device 100 to be provided outside of user device 100. For example, if the input section 120, display section 130, memory section 140, processor section 150, and imaging means 160 are each composed of separate hardware components, each hardware component may be connected via an arbitrary network. may be done. At this time, the type of network does not matter. Each hardware component may be connected via a LAN, wirelessly, or wired, for example. User device 100 is not limited to a specific hardware configuration. For example, it is also within the scope of the present disclosure to configure the processor section 150 with an analog circuit rather than a digital circuit. The configuration of the user device 100 is not limited to that described above as long as its functions can be realized.

FIG. 6B shows an example of the configuration of the server device 200.

The server device 200 includes a communication interface section 210, a memory section 220, and a processor section 230.

The communication interface unit 210 controls communication via the network 400. The communication interface section 210 also controls communication with the database section 300. The processor unit 230 of the server device 200 can receive information from outside the server device 200 via the communication interface unit 210, and can transmit information to the outside of the server device 200. The processor section 230 of the server device 200 can receive information from the user device 100 via the communication interface section 210, and can transmit information to the user device 100. Communication interface unit 210 may control communication in any manner.

The memory unit 220 stores programs required for executing processes of the server device 200, data required for executing the programs, and the like. For example, a program for estimating or predicting the state of a target (for example, a program for performing processing including the steps shown in FIG. 1A or a program for implementing the processing shown in FIGS. 7 and 8 described later), and/or Part or all of a program for evaluating a therapeutic drug or preventive drug or other medical technology for a subject (for example, a program for performing processing including the steps shown in FIG. 1B) is stored. The memory unit 220 stores, for example, an application for acquiring eyeball information by tracking the pupil in an eyeball image (eye tracking), and an application for acquiring eyeball information from electrooculography measurement values. Good too. The memory unit 220 may store, for example, an application for acquiring eyeball information by other means. Memory section 220 may be implemented by any storage means.

The processor unit 230 controls the operation of the server device 200 as a whole. The processor unit 230 reads a program stored in the memory unit 220 and executes the program. This allows the server device 200 to function as a device that executes desired steps. The processor unit 230 may be implemented by a single processor or by multiple processors.

In the example shown in FIG. 6B, each component of the server device 200 is provided within the server device 200, but the present disclosure is not limited thereto. It is also possible for any of the components of the server device 200 to be provided outside the server device 200. For example, if the memory section 220 and the processor section 230 are each composed of separate hardware components, each of the hardware components may be connected via an arbitrary network. At this time, the type of network does not matter. Each hardware component may be connected via a LAN, wirelessly, or wired, for example. Server device 200 is not limited to a specific hardware configuration. For example, it is also within the scope of the present disclosure to configure the processor section 230 with an analog circuit rather than a digital circuit. The configuration of the server device 200 is not limited to that described above as long as its functions can be realized.

In the examples shown in FIGS. 5 and 6B, the database unit 300 is provided outside the server device 200, but the present disclosure is not limited thereto. It is also possible to provide the database unit 300 inside the server device 200. At this time, the database unit 300 may be implemented by the same storage unit that implements the memory unit 220, or may be implemented by a different storage unit from the storage unit that implements the memory unit 220. In any case, the database unit 300 is configured as a storage unit for the server device 200. The configuration of the database unit 300 is not limited to a specific hardware configuration. For example, the database unit 300 may be composed of a single hardware component or a plurality of hardware components. For example, the database unit 300 may be configured as an external hard disk device of the server device 200, or may be configured as a storage on a cloud connected via a network.

The components of the system 10 may be included in the user device 100, the server device 200, or distributed in both the user device 100 and the server device 200, for example. When the components of the system 10 are distributed to both the user device 100 and the server device 200, the user device 100 can be provided with the acquisition means 11, and the server device 200 can be provided with the estimation/prediction means 12.

Similarly, the components of the system 10A may be included in the user device 100, the server device 200, or distributed in both the user device 100 and the server device 200, for example. good. When the components of the system 10A are distributed to both the user device 100 and the server device 200, the user device 100 may include the acquisition means 11, and the server device 200 may include the calculation means 13.

In one embodiment, in the system 10, the user device 100 includes the acquisition unit 11, and the server device 200 includes the estimation/prediction unit 12.

FIG. 7 is a data flow diagram showing data exchange between the user device 100 and the server device 200 in this embodiment.

In step S701, the user device 100 acquires an eyeball information source. The eyeball information source may be, for example, at least one image taken of the subject's eyeball, or may be an electrooculography measurement value. Preferably, the eyeball information source may be a plurality of images taken of the subject's eyeball or a time series of electro-oculography measurements. This is because a plurality of images or time-series electro-oculography measurement values can include a time component of eyeball information.

For example, the imaging unit 160 of the user device 100 can acquire at least one image by photographing the eyeball. For example, the imaging means 160 of the user device 100 may photograph the eyeball of the target when the target is aware, such as when the target is staring at the imaging means 160, or when the target is within the field of view of the imaging means 160. The subject's eyeballs may be photographed when the subject is not recognized, such as when the subject is in the room. By photographing the subject's eyeballs when the subject is not aware of the situation, it is possible to obtain an eyeball information source without placing a burden on the subject.

For example, the electro-oculogram acquisition unit 170 of the user device 100 can acquire an electro-oculogram measurement value by measuring the electric potential around the eyes. For example, the electro-oculogram acquisition means 170 of the user device 100 may acquire the electro-oculogram measurement value when the subject is aware of the situation, such as by attaching electrodes around the eyes, or when the subject is wearing them. Electro-oculography measurements may be obtained using electrodes or the like built into a glasses-type device, when the object is not recognized. By photographing the subject's eyeballs when the subject is not aware of the situation, it is possible to obtain an eyeball information source without placing a burden on the subject.

In step S702, the user device 100 acquires eyeball information. For example, the acquisition unit 11 that may be included in the processor unit 150 of the user device 100 acquires eyeball information from the eyeball information source acquired in step S701. The eyeball information includes, for example, information regarding blinks, information regarding pupil coordinates or relative positions of the eyeballs, and information regarding eyeball movements. The acquisition means 11 can acquire eyeball information from an eyeball information source, for example, by any known method. The acquisition means 11 preferably acquires values of parameters related to blinking.

In step S703, the user device 100 transmits eyeball information to the server device 200. For example, the processor unit 150 of the user device 100 transmits eyeball information to the server device 200 via the communication interface unit 110. Server device 200 receives the transmitted eyeball information. For example, the processor unit 230 of the server device 200 receives eyeball information from the user device 100 via the communication interface unit 210.

In step S704, the server device 200 estimates or predicts the state of the target based on the eyeball information. For example, the estimation/prediction means 12 that may be included in the processor unit 230 of the server device 200 estimates or predicts the state of the target based on the eyeball information received in step S703.

The calculation means 12 can estimate or predict the state of the target by using, for example, a trained model. The learned model has learned the relationship between the eyeball information and the state of the target, and when the eyeball information received in step S703 is input to the trained model, the corresponding state of the target is output.

In step S705, the server device 200 transmits the estimation or prediction result in step S704 to the user device 100. For example, the processor unit 230 of the server device 200 transmits the estimation or prediction result in step S704 to the user device 100 via the communication interface unit 210. User equipment 100 receives the transmitted estimation or prediction results. For example, the processor unit 150 of the user device 100 receives the estimation or prediction result from the server device 200 via the communication interface unit 110.

In step S706, the user device 100 presents the estimation or prediction result received in step S705 to the target. The mode of presentation does not matter. For example, the display unit 130 of the user device 100 can display the results. Alternatively, a microphone that the user device 100 may include or be connected to may present the results audibly. Alternatively, a printer that the user device 100 may include or be connected to can print the results.

This allows the subject (or the doctor, nurse, or other medical worker who is examining the subject) to know the condition of the subject.

In this embodiment, the user device 100 can obtain an estimated or predicted result from the eyeball information simply by acquiring the eyeball information and transmitting it to the server device 200. The user device 100 also does not have to bear the calculation load on the estimation/prediction means 12.

In another embodiment, the server device 200 can include the acquisition means 11 and the estimation/prediction means 12 of the system 10.

FIG. 8 is a data flow diagram showing data exchange between the user device 100 and the server device 200 in this embodiment.

In step S801, the user device 100 acquires an eyeball information source. Step S801 is a step similar to step S701.

In step S802, the user device 100 transmits the eyeball information source acquired in step S801 to the server device 200. For example, the processor unit 150 of the user device 100 transmits the eyeball information source to the server device 200 via the communication interface unit 110. The server device 200 receives the transmitted eyeball information source. For example, the processor unit 230 of the server device 200 receives the transmitted eyeball information source from the user device 100 via the communication interface unit 210.

In step S803, the server device 200 acquires eyeball information. For example, the acquisition unit 11 that may be included in the processor unit 230 of the server device 200 acquires eyeball information from the eyeball information source received in step S802. The eyeball information includes, for example, information regarding blinks, information regarding pupil coordinates or relative positions of the eyeballs, and information regarding eyeball movements. The acquisition means 11 can acquire eyeball information from an eyeball information source, for example, by any known method. The acquisition means 11 preferably acquires values of parameters related to blinking.

In step S804, the server device 200 estimates or predicts the state of the target based on the eyeball information. Step S804 is a step similar to step S704.

In step S805, the server device 200 transmits the estimation or prediction result in step S804 to the user device 100. Step S805 is a step similar to step S705.

In step S806, the user device 100 presents the estimation or prediction result received in step S805 to the target. Step S806 is a step similar to step S706.

In this embodiment, the user device 100 can obtain a fixed or predicted result from the eyeball information source by simply acquiring the eyeball information source and transmitting it to the server device 200. The user device 100 does not have to bear the calculation load for acquiring eyeball information in addition to the calculation load placed on the estimation/prediction means 12.

In another embodiment, the user device 100 can include the acquisition means 11 and the estimation/prediction means 12 of the system 10. At this time, the user device 100 does not need to communicate with the server device 200 and can operate standalone. The user device 100 performs, for example, steps S701 to S706 other than steps S703 and S705.

In the examples described above with reference to FIGS. 7 and 8, it was explained that the processes are performed in a specific order, but the order of each process is not limited to that described, and can be performed in any logically possible order. It can be done.

In the example described above with reference to FIGS. 7 to 8, at least one of the processes in each step shown in FIGS. , the processor unit 230 and the memory unit 220 of the server device 200, but the present disclosure is not limited thereto. At least one of the processes in each step shown in FIGS. 7 to 8 may be realized by a hardware configuration such as a control circuit.

The system of the present disclosure can be used in conjunction with, for example, a means capable of outputting indicators related to Parkinson's disease.

In one example, the means capable of outputting indicators related to Parkinson's disease is a wearable device configured to be worn on the patient's body.

The wearable device may be configured to be attached to at least one limb of the patient and to measure limb movements (eg, tremors) due to Parkinson's disease. For example, the wearable device can measure the movement of a limb using an inertial sensor mounted therein, and can output velocity data, acceleration data, angular velocity data, and/or angular acceleration data. Such a wearable device whose output data is analyzed and used as an index regarding Parkinson's disease is described, for example, by Rob Powers1 et al. , “Smartwatch inertial sensors continuously monitor real-world motor fluctuations in Parkinson’s disease ”, Sci. Transl. Med. 13, eabd7865 (2021) 3 February 2021.

By using such a wearable device to obtain an index related to a patient's Parkinson's disease in advance, this index can be used to calculate a threshold for estimating the patient's condition.

For example, the wearable device and the information acquisition means 11 are used together to continuously acquire movement measurement data and eyeball information. Eyeball information when movement measurement data indicates that symptoms of Parkinson's disease are present can be set as a threshold value regarding symptoms of Parkinson's disease.

The systems, apparatuses, and devices of the present disclosure may optionally be configured with one or more sensors that collect health data of the user, other than indicators related to Parkinson's disease. Health data may include any suitable data related to the user's health. In some examples, the device may be configured to capture health data from the user. Such health data may include information about the user, such as pulse rate, heart rate, heart rate variability measurements, temperature data, number of steps, standing and sitting time, number of calories burned, number of minutes exercised, and/or any other suitable information. Data may be specified. A device may be configured with one or more input devices that allow a user to interact with the device. The device may also be configured with one or more output devices for outputting any suitable output information. For example, the device may be configured to output visual, audio, and/or tactile information. In some examples, the output information can be presented to the user in a manner that guides the user to perform one or more actions related to breathing. For example, the output information may include a fluctuating progress indicator (eg, some type of visual information). A progress indicator can be presented on the device's graphical user interface and can be configured to guide the user through a series of breathing movements included in a breathing sequence, as further described herein. Output information may be presented by an application running on the device.

The device may be associated with a second device (e.g., a paired device (e.g., a wearable device such as a wristwatch or glasses) or a host device). In some examples, this may include a device paired with the second device in any suitable manner. Pairing two devices optionally allows the second device to function as a replacement for this device. The device, the second device, or any suitable combination of the device and the second device can generate output information based at least in part on the health data. Further parameters may be obtained by arranging arbitrary electrodes. Health indicators that can be calculated using electrodes include, but are not limited to, cardiac function (ECG, EKG), water content, body fat percentage, galvanic skin resistance, and combinations thereof.

The present disclosure may apply systems, apparatus, and methods for operation of a wearable device that depends on whether the wearable device utilized by the present disclosure is being worn. Such a wearable device can be attached to a portion of a user's body by an attachment member and can operate in at least a connected state and a disconnected state. One or more sensors disposed within the wearable device and/or the attachment member can detect when the device is attached to or in close proximity to a subject. One or more sensors disposed within the wearable device and/or the attachment member can detect that the attached object, if present, is a part of the user's body. Such a technique is publicly known, and for example, the technique described in WO2015/116163 can be used.

It may be linked or used in conjunction with an electronic medical record application such as Welby. For example, you can periodically take pictures of your eye movements at home for about 5 minutes using a smartphone camera or a stationary eye tracker and record the parameters. ) to optimize the relationship between symptoms and eye movement parameters. Data can be transmitted remotely to help doctors optimize drug prescriptions. Here, a contactless eye tracker or an eye tracker available from Tobii Technology (Sweden) can be used.

In one embodiment, the present disclosure provides a system for implementation as an application, the system being used for treatment of a treatable disease, the system comprising a server and a user terminal. wherein medical characteristics representing medical characteristics of a patient related to a disease are clustered into various medical characteristics (e.g., behavioral medical characteristics, knowledge medical characteristics, and cognitive medical characteristics), and the server are stored in association with one or more treatments, each medical characteristic is further associated with another type of medical characteristic (e.g., any one of a behavioral medical characteristic, a knowledge medical characteristic, and a cognitive medical characteristic), and the server further includes a treatment method associated with a specific medical characteristic selected as a treatment target from the plurality of medical characteristics and a treatment method associated with each of the various medical characteristic information associated with the selected medical characteristic. selecting a therapy to perform and transmitting therapy information for the selected therapy, the user terminal presenting information for a therapy based on the received therapy information; Can be adopted. In some embodiments, the selection of a treatment by a medical professional becomes training information and changes the probability of selecting a treatment, thereby increasing the probability of selecting a treatment considered to be more effective, thereby increasing the probability of selecting a treatment that is considered to be more effective. It becomes possible to do so. Alternatively, cluster factors for individual patients can be modified based on efficacy information. For example, the cluster factor of a cluster to which a medical characteristic associated with a treatment method that has been found to be effective belongs is increased, and when it is ineffective, it is decreased. Which cluster is more effective for treatment may vary depending on the patient. A cluster to which medical characteristics associated with an effective treatment belongs is likely to be an effective cluster for that patient, so by increasing the cluster factor, the probability of selecting a treatment for that cluster increases. increases, making more effective treatment possible.

(Note)
In this specification, "or" is used when "at least one or more" of the items listed in the sentence can be employed. The same goes for "or." In this specification, when "within a range of two values" is specified, the range includes the two values themselves.

All references, including scientific literature, patents, patent applications, etc., cited herein are incorporated by reference in their entirety to the same extent as if each were specifically indicated.

The present disclosure has been described above by showing preferred embodiments for ease of understanding. The present disclosure will now be described based on examples, but the above description and the following examples are provided for illustrative purposes only and are not provided for the purpose of limiting the present disclosure. The present invention will be explained in more detail below using Reference Examples, Examples, and Test Examples, but these are provided for the purpose of illustration only, and the present disclosure is not limited thereto. Note that the compound names shown in the following Reference Examples and Examples do not necessarily follow IUPAC nomenclature. Note that abbreviations may be used to simplify the description, but these abbreviations have the same meanings as in the above description. The scope of the disclosure is not limited to the embodiments or examples specifically described herein, but is limited only by the claims.

Examples are described below. The following examples were carried out in compliance with the Declaration of Helsinki.

(Example 1)
An uncontrolled, open-label, exploratory clinical trial was conducted on 20 patients with Parkinson's disease. The 20 Parkinson's disease patients attending or being hospitalized at the Department of Neurology, Juntendo Hospital, Juntendo University School of Medicine (Clinical Diagnosis Criteria for Parkinson's Disease (PD) (2015)) Patients who have been diagnosed with levodopa (confirmed or clinically probable), whose Hoehn & Yahr severity classification at the time of ON is Stage III or below, who have been treated with levodopa for at least 6 months (26 weeks) and who are not eligible for levodopa. Patients for whom efficacy was observed at the beginning of treatment were targeted. Eyeglass-type devices were worn on 20 patients with Parkinson's disease, parameters related to blinking were obtained, and the patient's condition at that time was also measured. By learning the relationship between the measured patient condition and blink-related parameters, a trained model capable of estimating or predicting the patient's condition was created, and estimation/prediction was performed.

(Procedure of Example 1)
FIG. 9 schematically shows the test procedure of Example 1.

A screening test was conducted 28 to 8 days before the test to confirm eligibility for this test.

A patient diary was recorded from 7 days to 1 day before the test. The patient's diary mainly records the patient's chief complaint (subjective symptoms, etc.), but it also records the presence or absence of dyskinesia and whether the medication is working or not, every 30 minutes. , and also included sleep time and levodopa taking time. At the same time, the patient was made to wear the glasses-type device in order to get the patient used to wearing the glasses-type device. Pupil Core eye-tracking glasses were used as the eyeglass-type device (https://pupil-labs.com). These Pupil Core eye tracking glasses can output an eyeball image whose photographing position is fixed and the time at which it was photographed. Pupil Core eye-tracking glasses and a smartphone (Android OS) were connected by wire, and the Pupil Core smartphone was used to supply power to the eye-tracking glasses, control ON/OFF, and record data collection.

On the day of the test, the patient took a levodopa preparation, and ocular information and condition before and after was recorded.

For levodopa preparations, patients were given the drugs they normally take in the usual dosage and administration. However, when evaluating eyeball information, etc., it was decided that the levodopa preparation should be taken after the patient was in the OFF state, and that the levodopa preparation should not be taken for a maximum of 4 hours after taking the levodopa preparation. However, if OFF was observed twice in a row in the MDS-UPDRS Part III score every 30 minutes within 4 hours after taking the levodopa preparation, the evaluation was terminated and the patient was allowed to take the levodopa preparation.

Two hours before taking the levodopa preparation, a glasses-type device was worn on the patient and eyeball information was measured. Parameters related to blinking were derived from the measured eyeball information.

30 minutes before the start of levodopa administration, every 30 minutes after levodopa administration, a total of 9 times up to 4 hours after administration, movements performed in daily life necessary for evaluation of MDS-UPDRS Part III score, UDysRS Part 3 and Part 4 (moving their hands, getting up from a chair, walking, etc.) and recording the movements on video. UDysRS scores and MDS-UPDRS scores were determined by the researcher and an assistant judge from the recorded videos. The evaluation ended when the MDS-UPDRS Part III score became OFF twice in a row. Immediately before administration of the levodopa preparation, every 15 minutes for 1 hour after administration of the levodopa preparation, and then every 30 minutes up to 4 hours after administration, 4 mL each time, up to a total of 11 times (maximum 44 mL), blood samples were collected for a total of 11 times (maximum 44 mL), and plasma levodopa concentration was determined. was measured. In addition, an indwelling intravenous catheter was used to reduce the burden on the patient. During the catheter placement, saline was used to prevent blood backflow and clotting within the catheter. Plasma levodopa concentration was measured using high performance liquid chromatography (HPLC).

(Example of one patient data)
FIG. 10 shows an example of data measured from one patient.

The top graph shows the time series fluctuation of the total score of MDS-UPDRS Part III score. The vertical axis is the total score of the MDS-UPDRS Part III score, and the horizontal axis is the time axis.

The second graph from the top shows time-series fluctuations in ON/OFF bedside scores. The vertical axis is the ON/OFF bedside score axis, and the horizontal axis is the time axis. The actual measured values are discrete quantities from 1 to 4, but they are linearly interpolated to make them continuous quantities. A score above 2.5 indicates ON, and a score below 2.5 indicates OFF.

The third graph from the top shows time-series fluctuations in plasma levodopa concentration. The vertical axis is the axis of plasma levodopa concentration (ng/mL), and the horizontal axis is the time axis.

The bottom graph shows time-series fluctuations in the number of blinks among the blink-related parameters. The vertical axis is the axis of the number of blinks, and the horizontal axis is the time axis. The color coding represents the blink duration of each blink, and the darker the color, the longer the blink duration.

From these graphs, after the levodopa preparation is taken at time 0, the plasma levodopa concentration increases, motor symptoms are affected, the total score of the MDS-UPDRS Part III score decreases, and the state shifts to ON. Accordingly, the number of blinks also increased.

From this, it was confirmed that blink-related parameters (at least the number of blinks), the total score of the MDS-UPDRS Part III score, the ON/OFF state, and the plasma levodopa concentration can be correlated.

(Results of data analysis)
Data from 19 of the 20 participants were analyzed, excluding data from one patient who had poor measurements.

Three patterns were observed as changes in the number of blinks in response to levodopa preparations: "increase," "decrease," and "indistinct." That is, a patient has a pattern in which the number of blinks gradually increases after taking a levodopa preparation at time 0, and a patient has a pattern in which the number of blinks gradually decreases after taking a levodopa preparation at time 0. After the levodopa preparation was taken at time 0, patients were classified as having an unclear pattern of change in the number of blinks.

Dyskinesia was observed in 7 of the 19 cases.

(Creation of machine learning model)
Using the acquired data, we constructed a machine learning model that can estimate clinical information using blink-related parameters as features.

A machine learning model was created using the machine learning platform "DataRobot."

Parameters related to blinking for 3 minutes were used as feature quantities. The blink-related parameters include 666 parameters including the number of blinks, blink duration, blink confidence, blink energy, and blink interval, and some of these parameters were used for learning.

The corresponding 3 minutes of clinical information was used as the objective variable. As clinical information, presence or absence of dyskinesia, ON/OFF, total score of MDS-UPDRS Part III score, and plasma levodopa concentration were used. Although the clinical information sampling interval is 15 to 30 minutes, linear interpolation was performed to obtain a value every 3 minutes. Specifically, we created a machine learning model that can estimate the presence or absence of dyskinesia using the presence or absence of dyskinesia as the objective variable. Using ON/OFF as the objective variable, we created a machine learning model that can estimate ON/OFF. Using the total score of the MDS-UPDRS Part III score as an objective variable, a machine learning model capable of estimating the total score of the MDS-UPDRS Part III score was created. Using plasma levodopa concentration as the objective variable, we created a machine learning model that can estimate plasma levodopa concentration.

Furthermore, a machine learning model was created that also used the time elapsed since taking the levodopa preparation as a feature.
For example, the following parameters were cited as feature quantities that had a relatively large contribution in any of the above machine learning models.
Here, the baseline is defined as the blink parameter at the measurement time before taking levodopa or at the OFF time after the effect of levodopa disappears.
・Number of blinks with high blink confidence ・Absolute value of change from baseline in number of blinks with high blink confidence ・Percentage of blinks with high blink confidence ・Number of blinks with low blink confidence - Percentage of blinks with low blink confidence - Maximum value of blink confidence - Average value of blink confidence - Square value of the average value of blink confidence - Square root of the average value of blink confidence - Average value of blink confidence・Absolute value of the amount of change from the baseline in the logarithm of the average value of blink confidence ・Ratio of the number of blinks with high blink confidence to the number of blinks with average blink confidence ・Blink Absolute value of the change from the baseline in the ratio of the number of blinks with high blink confidence to the number of blinks with average eye confidence ・Blink confidence relative to the number of blinks with average eye blink confidence The absolute value of the change from the baseline in the logarithm of the ratio of the number of blinks with high ・Median value of blink confidence ・Standard deviation of blink confidence ・Standard deviation of blink depth ・Maximum value of blink depth ・Absolute value of the amount of change from the baseline in the maximum value of blink depth - Absolute value of the amount of change from the baseline in the minimum value of blink depth - Percentage of blinks with large blink depth - Blinks with large blink depth Logarithm of the percentage of eyes ・Absolute value of the change from the baseline in the percentage of blinks with large blink depth ・Absolute value of the change from the baseline in the number of blinks with small blink depth ・Blink depth is the average number of blinks - Average value of the blink interval - Absolute value of the amount of change from the baseline in the average value of the blink interval - Absolute value of the amount of change from the baseline in the minimum value of the blink interval・Median value of blink interval ・Absolute value of change from baseline in median value of blink interval ・Number of blinks with short blink duration ・Number of blinks with short blink duration from baseline・Percentage of blinks with short blink duration ・Absolute value of change from baseline in percentage of blinks with short blink duration ・Minimum value of blink duration ・Blink duration Median time ・Percentage of blinks with average blink duration ・Absolute value of change from baseline in percentage of blinks with average blink duration ・Blinks with long blink duration Number of eyes - Absolute value of change from baseline in number of blinks with long blink duration - Percentage of blinks with long blink duration - Blink energy - Amount of change in blink energy from baseline - Logarithmic value of blink energy - Absolute value of change from baseline in logarithmic value of blink energy - Total number of blinks - Logarithmic value of total blink count - Base of logarithm of total blink count Absolute value of the amount of change from the baseline - Absolute value of the amount of change from the baseline in the square root of the total number of blinks - Absolute value of the amount of change from the baseline in the square value of the total number of blinks

(Determination of presence or absence of dyskinesia)
A machine learning model was created using the presence or absence of dyskinesia as the objective variable and 15 blink-related parameters as features.

Furthermore, another machine learning model was created using the presence or absence of dyskinesia as the objective variable and the 15 blink-related parameters and the time elapsed since taking the levodopa preparation as feature quantities.

For comparison, a machine learning model was created using the presence or absence of dyskinesia as the objective variable and the plasma levodopa concentration as the only feature. Furthermore, for comparison, a machine learning model was created using the presence or absence of dyskinesia as the objective variable and only the time elapsed since taking the levodopa preparation as a feature.

A first group of machine learning models was created using data from all patients, and a second group of machine learning models was created using data from patients who had a pattern of increased blink frequency, and a second group of machine learning models was created using data from patients who had a pattern of increased blink frequency. A third group of machine learning models was created using data of patients with patterns, and a fourth group of machine learning models was created using data of patients with patterns in which changes in blink frequency were unclear. Here, the machine learning model group refers to a machine learning model created with the presence or absence of dyskinesia as an objective variable and 15 eyeblink-related parameters as features, a machine learning model that uses dyskinesia as an objective variable, 15 eyeblink-related parameters, and levodopa It includes a machine learning model created using the elapsed time from the time of taking the drug as a feature. In addition to this, the first machine learning model group includes a machine learning model created using the presence or absence of dyskinesia as an objective variable and only the concentration of plasma levodopa as a feature quantity, and a machine learning model that uses the presence or absence of dyskinesia as an objective variable and is based on the time of taking levodopa preparations. machine learning model created using only the elapsed time as a feature.

To create the first machine learning model group, data from all patients (number of patients N=19) was used. There were 7 patients with dyskinesia (N _dys = 7), and the number of data points was n = 1451. Training: test: holdout = 16:4:5. In the following, for example, the following feature amounts made a large contribution to the estimation accuracy of the first machine learning model group created using the 15 blink-related parameters as feature amounts.
・Number of blinks with high blink confidence ・Standard deviation of blink confidence ・Total number of blinks ・Standard deviation of blink depth ・Average value of blink interval ・Maximum value of blink depth

To create the second machine learning model, data from patients (number of patients N=8) with a pattern of increased blink frequency was used. There were 4 patients with dyskinesia (N _dys = 4), and the number of data points was n = 596.

To create the third machine learning model, data from patients (number of patients N=5) with a pattern of decreased blink frequency was used. There was one patient with dyskinesia (N _dys = 1), and the number of data points was n = 369.

In order to create the fourth machine learning model, data of patients (number of patients N=6) with unclear patterns of changes in blink frequency were used. There were 2 patients with dyskinesia (N _dys = 2), and the number of data points was n = 486.

The results of the first to fourth machine learning models were compared.

Table 1 shows the results of comparing the results of the first to fourth machine learning models. Recall (true positive rate) in cross-validation data is an index indicating whether dyskinesia was estimated without missing anything, and is expressed as TP/(TP+FN). Precision (positive predictive value) is an index indicating how accurate the estimation of dyskinesia was, and is expressed as TP/(TP+FP). Here, TP is true positive, FP is false positive, and FN is false negative.

In the third machine learning model and the fourth machine learning model, the number of samples was too small to obtain sufficient results for comparison.

As can be seen from Table 1, the presence or absence of dyskinesia could be estimated with high accuracy by using parameters related to blinking as feature quantities. In contrast, it was difficult to determine the presence or absence of dyskinesia using a machine learning model based only on plasma levodopa concentration and a machine learning model created using only the time elapsed since taking the levodopa preparation as a feature.

The effects of adding parameters related to patient stratification based on the pattern of increase/decrease in the number of blinks and the time to take levodopa preparations were limited.

Next, we examined the estimation performance for unknown data.

From the data of 19 patients, data of one patient with dyskinesia and data of one patient without dyskinesia were excluded, and a machine learning model was created using data of 17 patients. The presence or absence of dyskinesia was used as the objective variable, and 15 types of blink-related parameters and the elapsed time from the time of taking the levodopa preparation were used as the feature quantities. Training (N = 17 patients, including patients with dyskinesia N _dys = 6, number of data points n = 1291), evaluation (N = 2 patients, including patients with dyskinesia N _dys = 1, number of data points n = 160).

FIG. 11 shows the results of estimation performance evaluation for unknown data.

The graph on the left shows the results of estimating the presence or absence of dyskinesia for data from patients without dyskinesia (unknown data derived from patients not used for learning). The graph on the right shows the results of estimating the presence or absence of dyskinesia for data from patients without dyskinesia (unknown data derived from patients not used for learning). The vertical axis shows the output from the machine learning model, and indicates the likelihood of dyskinesia as a value from 0 to 1. 0 indicates no dyskinesia and 1 indicates dyskinesia. The horizontal axis shows time. The graph on the right shows the actually measured duration of dyskinesia (about 90 minutes to about 220 minutes).

For the two cases of unknown data, the true positive rate (Recall)/positive precision rate (Precision) = 0.77/0.88 (threshold value = 0.10 for the moving average of 5 points (15 minutes)), and the unknown data It was also possible to estimate the presence or absence of dyskinesia.

(ON/OFF judgment)
A machine learning model was created using ON/OFF as the objective variable and 15 blink-related parameters as feature quantities.

Furthermore, another machine learning model was created with ON/OFF as the objective variable and the 15 blink-related parameters and the elapsed time from the time of taking the levodopa preparation as the feature quantities.

For comparison, a machine learning model was created using ON/OFF as the objective variable and only the plasma levodopa concentration as the feature quantity. Furthermore, for comparison, a machine learning model was created using ON/OFF as the objective variable and using only the elapsed time from the time of taking the levodopa preparation as the feature quantity.

A first set of machine learning models was created using data from all patients, and a second set of machine learning models was created using data from patients who had a pattern of increased blink frequency, and a second set of machine learning models was created using data from patients who had a pattern of increased blink frequency. A third group of machine learning models was created using data of patients with patterns, and a fourth group of machine learning models was created using data of patients with patterns in which changes in blink frequency were unclear. Here, the machine learning model group is a machine learning model created with ON/OFF as the objective variable and 15 blink-related parameters as features, and a machine learning model created with ON/OFF as the objective variable and 15 blink-related parameters. , and a machine learning model created using the elapsed time from the time of taking the levodopa preparation as a feature quantity. In addition to this, the first machine learning model group includes a machine learning model created with ON/OFF as the objective variable and only the plasma levodopa concentration as a feature quantity, and a machine learning model that uses ON/OFF as the objective variable and uses levodopa preparations as the objective variable. This includes a machine learning model created using only the elapsed time as a feature.

To create the first machine learning model group, data from all patients (number of patients N=19) was used. The number of data points was n=1451. Training: test: holdout = 16:4:5. For example, the following feature amounts made a large contribution to the estimation accuracy of the first machine learning model created using the 15 blink-related parameters as feature amounts.
・Percentage of blinks with high blink confidence ・Median value of blink interval ・Total number of blinks ・Percentage of blinks with high blink confidence ・Number of blinks with short blink duration ・Maximum value of blink depth ,minimum value

For example, the following feature values made a large contribution to the estimation accuracy of the first machine learning model, which was created using the 15 blink-related parameters and the elapsed time from the time of taking the levodopa preparation as feature values.
・Time elapsed since taking the levodopa preparation ・Percentage of blinks with high blink confidence ・Total number of blinks ・Maximum value of blink confidence ・Number of blinks with long blink duration ・Blinks with short blink duration Number of times ・Maximum blink depth

To create the second machine learning model, data from patients (number of patients N=8) with a pattern of increased blink frequency was used. The number of data points was n=596.

To create the third machine learning model, data from patients (number of patients N = 5) with a pattern of decreased blink frequency was used. The number of data points was n=369.

In order to create the fourth machine learning model, data from patients (number of patients N = 6) with unclear patterns of changes in blink frequency were used. The number of data points was n=486.

The results of the first to fourth machine learning models were compared.

Table 2 shows the results of comparing the results of the first to fourth machine learning models.

As can be seen from Table 2, ON/OFF is estimated with a certain degree of accuracy using a machine learning model based only on plasma levodopa concentration and a machine learning model created using only the time elapsed since taking the levodopa preparation as a feature. We were able to. A machine learning model that uses blink-related parameters as features was also able to estimate with similar accuracy. The most accurate estimation was possible by using parameters related to blinking and the time elapsed since the time of taking the levodopa preparation as features.

Estimation accuracy was somewhat improved by stratifying patients based on the pattern of increase/decrease in the number of blinks.

Next, we examined the estimation performance for unknown data.

A machine learning model was created using the data of 17 patients, excluding 2 data from the 19 patients' data. ON/OFF was used as the objective variable, and 15 types of blink-related parameters and the elapsed time from the time of taking the levodopa preparation were used as the feature quantities. Training (patient N=17, number of data points n=1291), evaluation (patient N=2, number of data points n=160).

FIG. 12 shows the results of estimation performance evaluation for unknown data.

The graph on the left shows the results of estimating ON/OFF for data from patient #3 (unknown data derived from the patient not used for learning). The graph on the right shows the result of estimating ON/OFF for data from patient #20 (unknown data derived from the patient not used for learning). The vertical axis shows the output from the machine learning model, and indicates the likelihood of being ON as a value from 0 to 1. 0 indicates that it is not ON, and 1 indicates that it is ON. The horizontal axis shows time. The graphs on the left and right show the actually measured ON periods.

For the two cases of unknown data, the true positive rate (Recall)/positive precision rate (Precision) = 0.61/0.80 (threshold value = 0.40 for the moving average of 5 points (15 minutes)), and the unknown data It was also possible to determine whether it was ON or OFF.

(Determination of total score of MDS-UPDRS Part III score)
A machine learning model was created using the total score of the MDS-UPDRS Part III score as the objective variable and the 15 blink-related parameters as the feature quantities.

Furthermore, another machine learning model was created using the total score of the MDS-UPDRS Part III score as the objective variable, and using the 15 blink-related parameters and the time elapsed since taking the levodopa preparation as feature quantities.

For comparison, a machine learning model was created using the total score of the MDS-UPDRS Part III score as the objective variable and using only the plasma levodopa concentration as the feature quantity. Furthermore, for comparison, a machine learning model was created using the total score of the MDS-UPDRS Part III score as an objective variable and using only the time elapsed from the time of taking the levodopa preparation as a feature quantity.

A first group of machine learning models was created using data from all patients, and a second group of machine learning models was created using data from patients who had a pattern of increased blink frequency, and a second group of machine learning models was created using data from patients who had a pattern of increased blink frequency. A third group of machine learning models was created using data of patients with patterns, and a fourth group of machine learning models was created using data of patients with patterns in which changes in blink frequency were unclear. Here, the machine learning model group refers to a machine learning model created using the total score of MDS-UPDRS Part III score as the objective variable and 15 blink-related parameters as the feature quantity, and a machine learning model created using the total score of MDS-UPDRS Part III score as the objective variable. It includes a machine learning model created with 15 blink-related parameters as variables and the elapsed time from the time of taking the levodopa preparation as features. In addition to this, the first machine learning model group includes a machine learning model created using the total score of the MDS-UPDRS Part III score as an objective variable and only the plasma levodopa concentration as a feature quantity, and the MDS-UPDRS Part III score The model includes a machine learning model created with the total score as the objective variable and only the time elapsed since taking the levodopa preparation as the feature quantity.

To create the first machine learning model group, data from all patients (number of patients N=19) was used. The number of data points was n=1451. Training: test: holdout = 16:4:5. For example, the following feature amounts made a large contribution to the estimation accuracy of the first machine learning model created using the 15 blink-related parameters as feature amounts.
・Median blink duration ・Percentage of blinks with long blink duration ・Median value of blink interval ・Standard deviation of blink depth ・Percentage of blinks with large blink depth ・Blinks with high blink confidence number of eyes

For example, the following feature values made a large contribution to the estimation accuracy of the first machine learning model, which was created using the 15 blink-related parameters and the elapsed time from the time of taking the levodopa preparation as feature values.
・Time elapsed since taking the levodopa preparation ・Percentage of blinks with high blink confidence ・Blink energy ・Total number of blinks ・Standard deviation of blink depth

The results of the first to fourth machine learning models were compared.

Table 3 shows the results of comparing the results of the first to fourth machine learning models.

As can be seen from Table 3, with respect to the total score of the MDS-UPDRS Part III score, it was found that there was a moderate correlation between the output and the actual measured value in the machine learning model that uses blink-related parameters as the feature quantity. In addition to blink-related parameters, a high correlation between the output and the measured value was obtained by using the time elapsed since taking the levodopa preparation as a feature.

It was difficult to determine the total score of the MDS-UPDRS Part III score using a machine learning model based only on plasma levodopa concentration and a machine learning model created using only the time elapsed since taking the levodopa preparation as a feature quantity. .

It was possible to estimate with higher accuracy for the patient group where the number of blinks increased.

Next, we examined the estimation performance for unknown data.

A machine learning model was created using the data of 17 patients, excluding 2 data from the 19 patients' data. The total score of the MDS-UPDRS Part III score was used as the objective variable, and 15 types of blink-related parameters and the elapsed time from the time of taking the levodopa preparation were used as the feature quantities. Training (patient N=17, number of data points n=1291), evaluation (patient N=2, number of data points n=160).

FIG. 13 shows the results of estimation performance evaluation for unknown data.

The graph (a) on the left shows the results of estimating the total score of the MDS-UPDRS Part III score for data from patient #3 (unknown data derived from a patient not used for learning). The graph (a) on the right shows the result of estimating the total score of the MDS-UPDRS Part III score for data from patient #20 (unknown data derived from the patient not used for learning). The vertical axis indicates the total score of the MDS-UPDRS Part III score. The horizontal axis shows time. The light solid line indicates the actual measured value of the total score of the MDS-UPDRS Part III score, and the dark solid line indicates the output value from the machine learning model.

For two examples of unknown data, we calculated the correlation between the actual measured values and the output from the machine learning model. (b) shows the results. In either case, the coefficient of determination was 0.77, and it was possible to estimate temporal changes in clinical scores with strong correlation even for unknown data.

(Determination of plasma levodopa concentration)
A machine learning model was created using plasma levodopa concentration as the objective variable and 15 blink-related parameters as features. As the plasma levodopa concentration, the plasma levodopa concentration measured almost simultaneously with the blink measurement and the plasma levodopa concentration measured about 30 minutes before the blink measurement were used, respectively.

Furthermore, another machine learning model was created using the plasma levodopa concentration as the objective variable and the 15 blink-related parameters and the elapsed time from the time of taking the levodopa preparation as feature quantities.

For comparison, a machine learning model was created with plasma levodopa concentration as the objective variable and only the time elapsed since taking the levodopa preparation as a feature.

A first set of machine learning models was created using data from all patients, and a second set of machine learning models was created using data from patients who had a pattern of increased blink frequency, and a second set of machine learning models was created using data from patients who had a pattern of increased blink frequency. A third group of machine learning models was created using data of patients with patterns, and a fourth group of machine learning models was created using data of patients with patterns in which changes in blink frequency were unclear. Here, the machine learning model group refers to a machine learning model created with plasma levodopa concentration as an objective variable and 15 blink-related parameters as features, and a machine learning model created with plasma levodopa concentration as an objective variable and 15 blink-related parameters. It includes a machine learning model created using parameters and the elapsed time from the time of taking the levodopa preparation as features. In addition to this, the first machine learning model group includes a machine learning model created using plasma levodopa concentration as an objective variable and using only the elapsed time from the time of taking the levodopa preparation as a feature quantity.

To create the first machine learning model group, data from all patients (number of patients N=19) was used. The number of data points was n=1451. Training: test: holdout = 16:4:5.

To create the third machine learning model, data from patients (number of patients N=5) with a pattern of decreased blink frequency was used. The number of data points was n=369.

The results of the first to fourth machine learning models were compared.

Table 4 shows the results of comparing the results of the first to fourth machine learning models.

As can be seen from Table 4, in addition to the blink-related parameters, the plasma levodopa concentration could be estimated with high accuracy by using the elapsed time from the time of taking the levodopa preparation as a feature quantity. In addition, in order for the levodopa taken to affect blinking, the levodopa taken into the peripheral blood must move into the brain, so the blink parameter is a function of the peripheral blood levodopa concentration with a certain time lag. There is a possibility that it can be strongly estimated. Therefore, we estimated the plasma levodopa concentration at the same time as the blink measurement and uniformly 30 minutes before the blink measurement, and using only the blink-related parameters as feature quantities, we calculated the plasma levodopa concentration standardized for each patient. In the case of estimation, higher accuracy was obtained when estimating the plasma levodopa concentration 30 minutes before the blink measurement. Therefore, as hypothesized, blinking changes with a certain time lag relative to the plasma levodopa concentration, suggesting that it is more strongly correlated with the brain levodopa concentration than with the plasma levodopa concentration.

(Determination of signs of dyskinesia)
FIG. 14A(a) shows the results of estimating the presence or absence of dyskinesia from unknown data of patient #20.

It can be seen that the output from the machine learning model fluctuates about 15 minutes before the actual measurement time of about 90 minutes when dyskinesia is present (▼). This shows that the machine learning model that was created was able to determine even the signs of dyskinesia.

FIG. 14A(b) shows one of the results of determining the presence or absence of dyskinesia in the learning data. The dark solid line indicates the output value from the machine learning model, and the light solid line indicates the presence or absence of actually measured dyskinesia.

There were places where the output value from the machine learning model started to fluctuate several minutes (about 3 minutes or more) before the actual measurement of dyskinesia (▼). We were able to detect such a precursor in five of the seven cases of dyskinesia.

By using the machine learning model according to the present disclosure, it is possible to predict the appearance of dyskinesia during the day in advance and intervene therapeutically by taking medication, and even in patients who do not develop dyskinesia, levodopa treatment may be excessive. This is inferred from a machine learning model, suggesting that it may be possible to optimize the amount of therapeutic drugs before dyskinesia develops.

(Judgment of ON/OFF signs)
FIG. 14B(a) shows the results of ON/OFF determination for unknown data of patient #3 and patient #20.

It can be seen that the output from the machine learning model fluctuates at the timing before ON switches to OFF (▼). From this, it can be seen that the created machine learning model is able to determine even the signs of ON/OFF switching.

FIG. 14B(b) shows one of the results of determining whether the learning data is ON/OFF. A dark solid line indicates the output value from the machine learning model, and a light solid line indicates the presence or absence of actually measured ON/OFF.

It was observed that the output value from the machine learning model started to fluctuate several minutes (approximately 6 minutes or more) before the ON was switched to OFF (▼). We were able to detect such a precursor in 14 of the 19 cases' data.

By using the machine learning model according to the present disclosure, it is possible to predict in advance the appearance of OFF in diurnal symptom fluctuations, to intervene therapeutically by taking medication, and to diagnose patients in the prodromal stage before motor symptoms appear. This suggests the possibility of early intervention.

(Implementation example 2: Application example)
As shown in these examples, measurements that correlate with clinical evaluations, such as the presence or absence of dyskinesia in Parkinson's disease patients, ON/OFF, and the total score of MDS-UPDRS Part III scores, from eye movement measurements such as blink parameters. It is possible to objectively estimate the value. This could be a means to solve the problems faced by today's treatments for Parkinson's disease. An example is shown below.

(patient)
This implementation example shows an application method related to Parkinson's disease. The users of this implementation example may include all Parkinson's disease patients from immediately after diagnosis to advanced and late stages of Parkinson's disease. Furthermore, patients who have not been diagnosed with Parkinson's disease but are suspected of having it, or who are considered to be at high risk of developing Parkinson's disease, could be targeted. People who are considered to have a high risk of developing Parkinson's disease include, for example, people who have mutations in genes that cause hereditary Parkinson's disease (such as SNCA, PARK2, PINK1, etc.), people who have anosmia, and people who have REM sleep behavior disorder. These include people, and relatively elderly people aged 60 to 70 years or older.

(installed)
In one embodiment, a user can measure eye movements on a daily basis by wearing smart glasses having an eye movement measuring function, an electro-oculography measuring device, or glasses with an external eye movement measuring device attached. . In some embodiments, non-wearable devices may also be used, such as stationary eye-tracking devices or camera-equipped smartphones, tablets, or personal computers to measure eye movements over relatively short periods of time. You can also do that.

(Daily life/measurement)
For example, users are expected to obtain the most abundant data by wearing the eye movement measurement device all day long, except when sleeping or bathing, but data is obtained relatively stably and continuously. Therefore, it may be worn only at certain times of the day, for example, for about 1 to 3 hours after taking the medication after waking up. For non-wearable devices, for example, eye movements can be measured for 3 to 5 minutes at a time using an installed eye tracking device, a smartphone with a camera function, a tablet terminal, a personal computer, etc. at the patient's home or medical institution. It is also possible to carry out measurements approximately 1 to 5 times a day to observe changes over time. Eye movement data may be measured not only on a daily basis but also periodically. When measuring regularly, it is most desirable to measure every day, but in reality, it is also possible to measure periodically, such as once to three times a week, once every January to March, or once every six months to a year. Changes over time can also be detected through measurements. The measurement environment is not particularly important as long as it does not affect the performance of the measurement device, but it is assumed that the measurement environment is the user's home, a medical institution, or a regular health checkup.

(Diagnosis and administration by a doctor)
The eye movement data thus acquired is accumulated in the user device and/or the server device, and various estimation algorithms are used to output an index of the amount of dopamine in the brain, an index of the severity of motor symptoms, dyskinesia, etc.

Such an embodiment could be a means to solve the problems of today's treatments for Parkinson's disease.

For example, symptoms of Parkinson's disease patients are evaluated using subjective evaluation scales used by clinicians and patients, such as MDS-UPDRS, UdysRS, and patient diaries. However, the accuracy of such subjective evaluation indicators is influenced by many factors, such as the clinician's experience and training, the patient's temperament, and the interpersonal relationship between the provider and the patient, and the results may vary depending on the evaluator. obtain. Furthermore, subjective rating scales have various costs and limitations due to the fact that clinicians and patients must actively perform the evaluation. Evaluations by clinicians are usually performed during hospital visits once every January to March, and are only measured at one point in the day. However, since Parkinson's disease has large diurnal and day-to-day fluctuations, it is difficult to obtain measured values that truly reflect the patient's condition with such infrequent and short-term evaluations.

An example of this embodiment provides a new index for objectively evaluating the motor symptoms and motor complications of Parkinson's disease and the amount of dopamine in the body by applying eye movements. Although the frequency and duration of assessment may vary depending on the embodiment, in all cases the measurements are convenient for the patient and provide a good indication of the patient's condition even in the absence of active observation of symptoms by the clinician or patient. , providing a means to measure, analyze, and report objectively and easily, remotely and far more frequently and continuously than clinically. These processes are almost automatic. The output results are automatically converted into a report format and can be used as a clinical tool for clinicians to refer to at any time. Objective and continuous data collection allows for more accurate capture of patient symptoms, including diurnal and day-to-day variations. From that information, clinicians can more effectively adjust and set drug doses and types to reduce the frequency and severity of motor symptoms and complications, such as dyskinesia and ON/OFF. Become. This makes it possible to provide precision medicine that provides treatments optimized to each individual's symptoms while reducing the burden on clinicians and patients, improving the quality of life of patients. A broader benefit of this embodiment is that it reduces the frequency of in-person patient visits to health care facilities, reducing the financial burden of health care on patients, health care providers, and the nation, and helping to optimize health economics. It is hoped that it will get closer.

In one example of the present embodiment, in addition to the above information, peripheral information such as measurement records based on past eye movements of the user and other users, patient diaries, evaluation records by clinicians, and medication status are simultaneously referred to, By analyzing and reporting, it is possible to provide even more useful information to help clinicians make more appropriate decisions.

In addition, in one example of this embodiment, when a patient is receiving device-assisted treatment such as deep brain stimulation therapy or administration of an L-DOPA preparation through a transdermal or enteral device, motor symptoms and motor complications are output. Alternatively, by using the estimated amount of dopamine in the body as an input value to a device-assisted treatment algorithm, the degree of device treatment can be determined in real time in response to fluctuations in symptoms, such as the dose and speed of administration of L-DOPA preparations, and the It is possible to construct a closed-loop system that controls the stimulation frequency and intensity of stimulation therapy within a safe range.

Furthermore, in some embodiments, some items in a report that reports the patient's condition can be referenced not only by medical personnel but also by the patient, and by visualizing the patient's symptoms and their degree of improvement, it is possible to be more proactive in treatment. It is possible to encourage people's participation. Furthermore, it is also possible to provide alerts and reminder functions that prompt patients to take medication based on the worsening of symptoms, including the current situation or in the future from several minutes to several tens of minutes, or when time has elapsed since taking the medication.

In recent years, objective evaluation systems for Parkinson's disease symptoms using wristwatch-type acceleration and angular velocity measurement devices (IMUs) have been considered, but the symptoms that can be evaluated are limited to dyskinesia chorea and resting tremor. However, we have only been able to grasp one aspect of the motor symptoms and motor complications of Parkinson's disease, which manifest in a variety of ways. Furthermore, since it is ineffective against symptoms that do not occur in the arm where the device is attached, it cannot be applied to all patients and does not necessarily solve the problem. The present disclosure focuses on eye movements that directly reflect dopamine function in the brain, and is different from IMU-based systems that directly quantify motor symptoms and motor complications of Parkinson's disease. As a result, in the embodiments of the present disclosure, changes in brain dopamine function, which are the essence of pathological conditions, can be acutely grasped.

In one example of this embodiment, the degree of progression of Parkinson's disease can also be estimated by measuring the in-vivo dopamine level and/or responsiveness to a therapeutic agent such as L-DOPA over time using eye movements. . Therefore, by regularly performing measurements according to the present embodiment on patients who are taking or being treated with disease-modifying drugs, preventive drugs, candidate drugs and treatment methods, regenerative cell medicines, etc. for Parkinson's disease, The effect of the treatment method on inhibiting the progression of Parkinson's disease can be evaluated. Today, to evaluate the effects of disease-modifying drugs for Parkinson's disease, we use invasive and high-cost imaging diagnostic methods such as PET and SPECT, which measure dopamine receptors and dopamine transporters in the brain, to quantify neurological deficits, as well as the methods described above. The mainstream is the evaluation of motor symptoms by clinicians, which is subjective and cannot always be expected to be accurate. Alpha-synuclein-related substances, which are thought to be causative agents of Parkinson's disease, are also being investigated as biomarkers in the blood, but no established indicators necessarily exist. The lack of biomarkers for evaluating the progression of the disease makes it difficult to evaluate disease-modifying drugs, and is one of the reasons why no radical treatment for Parkinson's disease has yet been established. In this situation, eye movement indicators directly evaluate dopamine function in the brain and can be easily measured frequently, continuously, and objectively, making it possible to assess the degree of progression of Parkinson's disease. It can be a means of quantitative evaluation. Therefore, such embodiments are expected to be useful for evaluating the effects of disease-modifying drugs.

Furthermore, in one example of the present embodiment, the user whose eye movements are measured does not need to be a patient with a confirmed diagnosis of Parkinson's disease. For example, patients suspected of having Parkinson's disease or people considered to be at high risk of developing Parkinson's disease may be targeted. Cases suspected of having Parkinson's disease include, for example, in addition to Parkinson's disease, patients who actually have parkinsonism syndromes such as multiple system atrophy, corticobasal degeneration, and supranuclear progressive palsy, as well as patients with essential tremor. Examples include war patients. It is known that the reactivity to L-DOPA is limited for diseases other than Parkinson's disease. In such users, we can help diagnose the disease by estimating in-vivo dopamine levels and motor symptoms by measuring eye movements, and by measuring responsiveness to Parkinson's disease treatment drugs such as L-DOPA. It can be applied as a biomarker to promote appropriate treatment. In addition, people who are considered to have a high risk of developing Parkinson's disease include, for example, people with mutations in genes that cause hereditary Parkinson's disease (such as SNCA, PARK2, PINK1, etc.), people with anosmia, and people with REM sleep behavior disorder. These include people who have , and relatively elderly people aged 60 to 70 years or older. In such users, dopamine function can be estimated before the onset of Parkinson's disease by estimating dopamine function in the brain by measuring eye movements, and by measuring responsiveness to Parkinson's disease treatment drugs such as L-DOPA. By detecting the decline, it can be applied as a biomarker that enables early diagnosis. It is said that motor symptoms appear in Parkinson's disease patients after about 60% of dopaminergic neurons have been lost. Its effectiveness is limited after significant neurological deficits occur. In other words, in order to realize a radical treatment for Parkinson's disease, we are faced with the challenge of finding and treating patients before the onset of Parkinson's disease. Although it is not yet clear to what extent dopamine nerve loss causes abnormalities in eye movement, it is thought to reflect dopamine function more directly, so it is possible that abnormalities can be identified earlier than general motor symptoms. is high, and it is thought that the possibility of discrimination will be further increased by performing a thorough analysis as shown in the present disclosure. That is, there is a possibility that users who have eye movement abnormalities themselves are in a high-risk group for Parkinson's disease.

Therefore, in such an example, it is possible to screen patients from an early stage when Parkinson's disease has not manifested as motor symptoms, and it may be possible to prescribe disease-modifying drugs at a stage when they can be expected to be effective. In addition, a broader advantage of such an embodiment is that if a treatment method that can be used as a preventive measure against Parkinson's disease is developed, such as a vaccine therapy against alpha-synuclein, an antibody drug, or a nucleic acid drug, it will be available to all elderly people. Although it is not realistic from a medical economic point of view to administer Parkinson's disease to patients in the early stages of Parkinson's disease, if it is possible to screen high-risk patients or patients with very early stages of Parkinson's disease based on their eye movements, the most optimal prescription can be achieved from a medical economic point of view. .

This application claims priority to Japanese Patent Application No. 2022-144047 filed with the Japan Patent Office on September 9, 2022, and the contents thereof may be cited as reference in this application as necessary. Ru.

The present disclosure provides technology for estimating or predicting a subject's condition and providing treatment that improves the quality of life.

10, 10A, 1000 System 11 Acquisition means 12 Estimation/prediction means 13 Calculation means 100 User device 200 Server device 300 Database section 400 Network

Claims

A method of estimating a state of a target, the method comprising:
A) obtaining the value of a parameter related to the subject's blink;
B) estimating the state of the object based on at least the value of the parameter related to the blink.
The parameter related to blinking is at least one of the following: eye-closing speed, eye-closing peak speed, eye-opening speed, eye-opening peak speed, eye-closing time, eye-opening amplitude, eye-closing amplitude, degree of eye opening, blink duration, and number of blinks. or a combination.
The method according to claim 1 or 2, wherein the blink-related parameters include at least one of blink confidence, blink interval, blink depth, and blink energy.
The method according to claim 1, wherein the blink-related parameters include blink confidence and blink interval.
The step B) includes estimating the state of the subject based on the elapsed time after administering L-DOPA, an L-DOPA-related compound, or a dopamine agonist to the subject and the blink-related parameter. , the method according to any one of claims 1 to 4.
The method according to any one of claims 1 to 5, wherein the condition includes a condition indicated by an index that can be used for clinical evaluation.
Any one of claims 1 to 6, wherein the condition includes a condition indicated by at least one of the presence or absence of dyskinesia, ON-OFF, MDS-UPDRS Part III score, UDysRS score, and plasma levodopa concentration. The method described in.
The method according to any one of claims 1 to 7, wherein the subject is a Parkinson's disease patient being treated with L-DOPA, an L-DOPA-related compound, or a dopamine agonist.
A method of estimating a state of a target, the method comprising:
A) acquiring eyeball information of the subject;
B) A method comprising the step of estimating the condition of the subject based on the eyeball information and the elapsed time after administering L-DOPA, an L-DOPA-related compound, or a dopamine agonist to the subject.
A method for estimating the condition of a Parkinson's disease patient being treated with L-DOPA, an L-DOPA-related compound, or a dopamine agonist, the method comprising:
A) obtaining a blink confidence value of the patient;
B) estimating the condition of the patient based on the blink confidence value and the elapsed time after administering L-DOPA, an L-DOPA-related compound, or a dopamine agonist to the patient; The method, wherein the condition includes at least one of the presence or absence of dyskinesia, ON-OFF, MDS-UPDRS Part III score, UDysRS score, and plasma levodopa concentration.
The step A) further includes obtaining at least one value of the patient's blink interval, blink energy, blink duration, and number of blinks,
The step B) is based on the value of the blink confidence, the elapsed time, and the value of at least one of the blink interval, blink energy, blink duration, and number of blinks. 11. The method of claim 10, comprising estimating the state of.
A system for estimating the state of a target,
means for acquiring the value of a parameter related to the blink of the subject;
A system comprising: means for estimating a state of the object based on at least a value of a parameter related to the blink.
A program for estimating a state of a target, the program being executed in a computer including a processor, the program comprising:
A) obtaining the value of a parameter related to the subject's blink;
B) A program that causes the processor to perform a process including: estimating the state of the object based on at least the value of the parameter related to the blink.
A method of evaluating a therapeutic or preventive drug or other medical technology for a subject, the method comprising:
A) obtaining the value of a parameter related to the subject's blink;
B) calculating an estimated effective amount or effective level or usage or administration of the therapeutic or prophylactic drug or other medical technique based on at least the value of the blink-related parameter.
15. The method according to claim 14, further comprising determining a therapeutic or prophylactic drug or other medical technique recommended for the subject based on the calculated estimated effective amount or effective level or usage or dosage. Method.
A system for evaluating therapeutic or preventive drugs or other medical techniques for a subject,
means for acquiring the value of a parameter related to the blink of the subject;
and means for calculating an estimated effective amount or effective level or usage or administration of the therapeutic or prophylactic drug or other medical technique based on at least the value of the blink-related parameter.
A program for evaluating therapeutic or preventive drugs or other medical techniques for a subject, the program being executed on a computer including a processor, the program comprising:
A) obtaining the value of a parameter related to the subject's blink;
B) Calculating an estimated effective amount or effective level or usage or dosage of the therapeutic or prophylactic drug or other medical technology based on at least the value of the blink-related parameter. ,program.
A target health management method,
carrying out the method according to any one of claims 1 to 11;
determining whether or not to treat the subject based on the results of the method;
If it is determined that treatment should be performed on the subject, taking action for health management of the subject.
The action includes performing the treatment on the target, issuing an alert that the treatment should be performed on the target, and administering a predetermined drug or therapy to the target. 19. The method of claim 18, comprising at least one of:
A target health management system,
an estimation system configured to perform the method according to any one of claims 1 to 11;
means for determining whether or not to treat the subject based on the output from the estimation system;
A system comprising means for taking action for health management of the subject when it is determined that the subject should be treated.
A target health management program, the program being executed on a computer comprising a processor, the program comprising:
carrying out the method according to any one of claims 1 to 11;
determining whether or not to treat the subject based on the results of the method;
A program that causes the processor to perform processing including, when it is determined that treatment should be taken for the subject, issuing an instruction to take action for health management of the subject.