WO2021206046A1

WO2021206046A1 - Method for objective sleep evaluation of mentally disordered patient

Info

Publication number: WO2021206046A1
Application number: PCT/JP2021/014471
Authority: WO
Inventors: 紀夫尾崎; 邦弘岩本; 聖子宮田; 淳一江口; 智史中田; 邦明加賀
Original assignee: 国立大学法人東海国立大学機構; 株式会社三菱ケミカルホールディングス
Priority date: 2020-04-10
Filing date: 2021-04-05
Publication date: 2021-10-14
Also published as: WO2021205648A1; JPWO2021206046A1

Abstract

The present invention relates to a method for evaluating sleep in a subject and to support of treatment of a mentally disordered patient based on said evaluation. Specifically, the present invention relates to a method for objectively evaluating sleep in a subject, the evaluation method including a step for acquiring sleep information after determining a sleep stage by processing a brain wave signal of the subject using a device that electrically processes bioelectric activity acquired from an electrode affixed to the skin on a temporal bone, and a step for analyzing the sleep information to evaluate the sleep of the subject.

Description

Objective sleep evaluation method for psychiatric patients

Related application:
This specification includes the contents described in the specification of PCT / JP2020 / 016112 (filed on April 10, 2020), which is the basis of the priority of the present application.
Technical field:
The present invention relates to a method for evaluating sleep of a mentally ill patient, and therapeutic support for the mentally ill patient based on the evaluation.

Sleep disorders are known to cause various illnesses. For example, those with insomnia have a higher incidence of depression than those without insomnia. In addition, shorter subjective sleep times increase the intracerebral deposition of β-amyloid associated with Alzheimer's disease.

In mental disorders such as depression, bipolar disorder, schizophrenia, anxiety, dementia, and neurodevelopment (developmental disorder), sleep problems such as insomnia are often observed. For example, in depression patients, insomnia is one of the most frequent complaints from the beginning, and insomnia is also the most common residual symptom of depression. Changes in sleep often precede other clinical symptoms, and their exacerbations and improvements are considered clinically useful indicators of the course of treatment for depression. On the other hand, sleep disorders such as sleep apnea syndrome often coexist, and it is essential to distinguish between insomnia and hypersomnia, and psychiatric symptoms and psychiatric disorders. Among dementia, Lewy body dementias, which presents with depressive symptoms similar to depression, shows morbid sleep in which muscle tension appears during REM sleep, and it is possible to confirm this type of sleep. It is useful for distinguishing from.

Methods for objectively evaluating sleep include polysomnography (PSG), which comprehensively evaluates sleep depth and physiological phenomena during sleep, actigraphy, which evaluates sleep-wake rhythm, and sleep scope (Sleepwell: Simple sleep evaluation devices such as Patent Document 1) and Sleep Profiler (Advanced Brain Monitoring Co., Ltd .: Patent Document 2) are known.

The PSG test is a test that measures biological activities such as electroencephalogram, eye movement, electrocardiogram, electromyogram, respiratory curve, snoring, and arterial oxygen saturation overnight. The PSG test is a medically recognized method for objectively assessing sleep, but it requires large-scale measuring equipment, is limited to measurements in a special laboratory after hospitalization, and is subject to testing. Since it is necessary to equip a person with many sensors and electrodes, continuous measurement for a long period of time is difficult. In addition, the PSG test has the inconvenience that the physical burden on the test subject is heavy and the measurement is performed in an environment different from the normal sleeping environment. Extraordinary effects called first night effects such as decreased sleep, decreased sleep efficiency, and increased arousal in patients with mental disorders, decreased total sleep time, decreased sleep efficiency, decreased REM sleep, and increased awakening time. The effect appears. Therefore, the PSG test is not suitable for early detection and follow-up of sleep disorders in psychiatric patients who are sensitive to environmental changes.

It is known that in psychiatric patients who complain of insomnia, poor sleep quality and abnormal sleep construction are observed even in PSG examination. However, in psychiatric disorders such as bipolar disorder, depression, and schizophrenia, it has been reported that subjective sleep assessment often does not match sleep status objectively measured by polysomnography.

Actigraphy is a small accelerometer and logger with a wristwatch structure, which is attached to the non-interested arm or waist of the person to be measured to detect the sleep-wake rhythm. The measurement subject wearing the actigraphy was bedtime, wake-up time, time spent on the bed without sleeping, nap, mood when waking up, taking sleeping pills, activities different from daily life and time when the device was removed. Record such things in a sleep diary. The data recorded in the actigraphy is processed by a computer, and by comparing it with the sleep diary, information such as total sleep time, sleep time ratio, total awakening time, awakening time ratio, awakening frequency and sleep onset latency can be obtained. .. Although actigraphy is effective for observing the general tendency of sleep-wake rhythm, since it is an evaluation based on body movement, it is compared with the PSG test in the evaluation of total sleep time, mid-wake time, sleep efficiency, and sleep onset latency. The correlation is very poor and it is not suitable for grasping sleep depth and accurate sleep time.

The sleep scope is a device that simply measures sleep electroencephalograms from electrodes attached to the forehead and nape of the neck (behind the ears). A method for determining the presence or absence of mental disorders by acquiring sleep electroencephalogram information using a sleep scope and analyzing information indicating the appearance status of α wave, δ wave, or β wave has been published (patented). Document 1). This method is essentially intended for the detection of psychiatric disorders, not for sleep assessment in psychiatric patients.

The sleep profiler is a device that simply measures brain waves, eye movements, pulse rate, chin EMG, etc. during sleep. Evaluation of sleep in healthy adults and patients with sleep-breathing disorders such as obstructive sleep apnea syndrome has been reported using a sleep profiler (Patent Document 1, Non-Patent Documents 1 to 3). However, sleep profilers are not suitable for psychiatric patients who are often sensitive to hyperesthesia and environmental changes because they wear a relatively large device on the forehead of the face.

Although there is a report of a method of diagnosing depression using brain waves (Patent Document 3), it is necessary to attach a plurality of electrodes to the scalp, which imposes a heavy burden on the subject. There is also a report on a device (Patent Document 4) that measures biological signals such as brain waves using an auricle wearing device, but wearing it during sleep causes a large resistance when the ear is placed down, which hinders sleep, especially. It is not acceptable to hypersensitive mentally handicapped patients. As described above, the existing simple sleep evaluation devices and methods are not sufficient as a method for objectively evaluating sleep, particularly for evaluating changes in sleep of mentally handicapped patients.

WO2012 / 176790 US8,355,769B2 US2016 / 0113567A1 JP 2011-005176

An object of the present invention is to provide a method for easily and accurately evaluating the sleep of a subject and treatment support for a mentally handicapped patient using the method.

The inventors have attached a small electrode to the skin on the temporal bone of a mentally ill patient (for example, behind the left and right ears) to acquire an electroencephalogram, and analyze the information contained in the electroencephalogram to analyze the existing PSG method. We found that sleep can be evaluated with an accuracy comparable to that of. Furthermore, as a result of analyzing various combinations of a plurality of information (parameters), it was found that the combination of three parameters of sleep efficiency, REM sleep latency, and light sleep appearance rate is useful for sleep evaluation.

The present invention is based on the above findings and relates to the following (1) to (24).
(1) A method for objectively evaluating a subject's sleep, in which an electroencephalogram signal of the subject is used to electrically process bioelectric activity acquired from an electrode attached to the skin on the temporal bone. An evaluation method including a step of acquiring sleep information after processing and determining a sleep stage, and a step of analyzing the sleep information to evaluate the sleep of a subject.
(2) Sleep parameters include sleep efficiency, REM sleep time, REM sleep appearance rate, REM sleep latency, light sleep time, deep sleep time, light sleep time appearance rate, deep sleep time appearance rate, and sleep stage. The evaluation method according to (1), which is selected from the group consisting of the number or frequency of transitions and the time series of transitions of sleep stages.
Among these parameters, (i) sleep efficiency, (ii) REM sleep time or its appearance rate or REM sleep latency, and (iii) light sleep and / or deep sleep time or its appearance rate, (iv) sleep onset It is preferable to use latency, (v) the number or frequency of sleep stage transitions, and the time series of sleep stage transitions, with (i)-(iii) being more preferred. Is preferably used.
(3) The method according to (1) or (2), wherein the subject is a mentally ill patient.
(4) The method according to any one of (1) to (3), which comprises a step of comparing sleep information of the patient at two or more different time points.
(5) The method according to (4), wherein two or more different time points include before and after treatment, before and after changing treatment, or during remission and recurrence / relapse.
(6) The method according to any one of (1) to (5), which comprises a step of diagramming and analyzing sleep information in a sleep progress chart.
(7) Input the subject's sleep information into a machine learning model determined by the sleep information of the healthy person and the mentally handicapped patient acquired in advance, and analyze whether the subject's sleep is closer to that of the healthy person or the mentally handicapped patient. The method according to any one of (1) to (6), which comprises the step of performing.
(8) The step of comparing the probability (p) calculated by using the machine learning model with a predetermined threshold value is included, and if the probability (p) is smaller than the threshold value, it is evaluated that the subject's sleep is close to that of a healthy person. The method according to (7).
(9) The method according to (8), wherein the machine learning model is a logistic regression model and the probability (p) is represented by the following equation.

The coefficients a1, a2, a3 and the constant term b are appropriately determined according to the population. In one embodiment, a1 = -1.299, a2 = 0.591, a3 = 0.531, b = -0.635.
(10) The method according to (8), wherein the machine learning model is a logistic regression model and the probability (p) is represented by the following equation.

The coefficients a1, a2, a3, a4 and the constant term b are appropriately determined according to the population. In one embodiment, a1 = -1.263, a2 = 0.604, a3 = 0.532, a4 = 0.064, b = -0.635.
(11) The method according to any one of (8) to (10), wherein the threshold value is 0.3 to 0.5, preferably 0.3 to 0.4.
(12) The method according to (7), wherein the machine learning model is XGBoost.
(13) The method according to (7), wherein the machine learning model is a neural network, preferably deep learning.
(14) A method for supporting (assisting) the treatment of a psychiatric patient, based on the step of evaluating the sleep of the patient according to the method according to any one of (1) to (13), and the evaluation result. The method comprising assessing the symptoms or extent of a patient's mental illness.
(15) A method for supporting (assisting) the treatment of a mentally handicapped patient, based on a step of evaluating the sleep of the patient according to the method according to any one of (1) to (13), and the evaluation result. The method comprising the step of selecting a preferred treatment.
(16) The orexin receptor antagonist is presented as the preferred treatment option when the sleep onset latency is 30 minutes or more and the bedtime or less, or the sleep efficiency is 75% or less, according to (15). Method.
(17) The method according to any one of (14) to (16), wherein the step of evaluating the symptom or the degree of mental disorder of the patient or the step of selecting a preferable treatment is carried out using machine learning.
(18) A treatment support system for a mentally handicapped patient, which includes an electroencephalogram signal processing device for electrically processing an electroencephalogram signal and an information analysis device.
The electroencephalogram signal processing device is a device that electrically processes electroencephalogram signals acquired from electrodes attached to the skin on the temporal bone of a patient, preferably behind the left and right ears, and more preferably to the mastoid process.
The information analysis device is
A storage unit that stores sleep information of the patient processed by a device that electrically processes an electroencephalogram signal, and a storage unit.
An analysis unit that analyzes and evaluates sleep information stored in the storage unit,
Equipped with an output unit that outputs analysis / evaluation results
The analysis unit evaluates the sleep of the patient according to the method according to any one of (1) to (13), or the patient according to any one of (14) to (17). The treatment support system for assessing the symptoms of mental illness or its degree, or selecting a preferred treatment.
(19) An information analysis device for supporting the treatment of mentally handicapped patients.
A storage unit that stores sleep information of the patient processed by a device that electrically processes an electroencephalogram signal, and a storage unit.
An analysis unit that analyzes and evaluates sleep information stored in the storage unit,
Equipped with an output unit that outputs analysis / evaluation results
The analysis unit evaluates sleep according to the method according to any one of (1) to (13), or the mentality of the patient according to any one of (14) to (17). The information analyzer for assessing the symptoms of a disorder or its degree, or selecting a preferred treatment.
(20) A program to support the treatment of mentally ill patients
Processing that acquires sleep information of the patient processed by a device that electrically processes an electroencephalogram signal, stores the sleep information, evaluates / analyzes the stored sleep information, and outputs the evaluation / analysis result. To let the computer do
The evaluation / analysis evaluates sleep according to the method according to any one of (1) to (13), or mental disorder of a patient according to any one of (14) to (17). The program for assessing the symptoms or extent of the disease, or selecting the preferred treatment.
(21) A method of supporting the treatment of mentally ill patients.
In the step of processing the brain wave signal of the patient to acquire sleep information using a device that electrically processes the brain wave signal, the step of analyzing the sleep information to evaluate the sleep of the patient, and the evaluation result. Including the step of selecting the preferred treatment based on
The electroencephalogram signal was obtained from an electrode attached to the skin on the patient's temporal bone.
The sleep information includes (i) sleep efficiency, (ii) REM sleep time or rate thereof or REM sleep latency, and (iii) light sleep and / or deep sleep time or rate thereof.
The method described above, wherein an orexin receptor antagonist is presented as the preferred treatment option when sleep onset latency is 30 minutes or more and bedtime or less, or sleep efficiency is less than 75%.
(22) A method for assisting the diagnosis of a psychiatric patient, the step of evaluating sleep according to the method according to any one of (1) to (13), and the psychiatric disorder of the patient based on the evaluation result. The method comprising a step of diagnosing.
(23) A method for selecting a subject for a clinical trial, in which the step of evaluating sleep according to the method according to any one of (1) to (13), and according to the evaluation result (insomnia therapeutic agent, etc.) ) The method comprising selecting a subject to be the subject of a clinical trial. According to this method, in a hypnotic clinical trial, a subject can be easily selected before a definitive diagnosis by a PSG test.
(24) A method for treating a psychiatric patient, the step of evaluating the patient's sleep according to the method according to any one of (1) to (13), the symptom of the patient's psychiatric disorder or the patient's psychiatric disorder based on the evaluation result. The method comprising assessing the degree and determining treatment. For example, orexin receptor antagonists are offered as preferred treatment options when sleep onset latency is 30 minutes or more and bedtime or less, or sleep efficiency is 75% or less.
The method may include a step of assessing the therapeutic effect and / or a step of assessing changes in the patient's symptoms over time by assessing sleep over time.
The step of evaluating a patient's sleep can be performed using the treatment system according to (18) or the information analysis device according to (19).

The method of the present invention uses small electrodes attached to the skin on the temporal bone (for example, behind the left and right ears), unlike the conventional method that requires the attachment of a large device and the restriction of physical freedom. Because of this, the subjects are well tolerated and are also suitable for sleep evaluation in psychiatric patients who are often hypersensitive. Therefore, there are few measurement errors and failures due to the burden associated with the measurement, and accurate sleep information can be obtained.
As will be described later, the waveforms of the EEG acquired from the forehead and the EEG acquired from the skin on the temporal bone (for example, behind the left and right ears) are different, but the EEG acquired from the electrodes attached to the skin on the temporal bone. It was also demonstrated by the present invention that sleep can be easily evaluated with an accuracy comparable to that of the PSG method.

FIG. 1 shows an example of an electroencephalogram signal processing device used in the method of the present invention. (A) Schematic configuration diagram, (b) Block diagram, (c) A state in which the device is attached to the body. FIG. 2A shows one form of the information analysis device of the present invention. FIG. 2B shows a processing example of the processor of the information analysis device. FIG. 2C is a schematic configuration diagram of one form of the treatment support system (electroencephalogram signal processing device + information analysis device) of the present invention. The treatment support system consists of an electroencephalogram signal processing device and an information analysis device. FIG. 3 shows the correlation of the evaluation results when the PSG test and the method of the present invention are applied to the same mentally ill patient at the same time. (A) total sleep time, (b) sleep efficiency, (c) REM sleep time, (d) non-REM sleep time, (e) mid-sleep awakening time, (f) sleep onset latency, (g) light sleep time (stage N1) And N2). FIG. 4 shows the results of comparing the sleep quality of healthy adults and the first night and the second night for each disease when the PSG test was performed (●: sleep apnea disorder group, ×: insomnia disorder group, ▲: exercise.・ Behavioral disorders, ○: healthy adults). (A) sleep efficiency, (b) mid-wake time, (c) deep sleep time, (d) REM sleep time. FIG. 5 compares the correlation of the evaluation results when the PSG test and the actigraphy are applied to the same mentally handicapped patient at the same time with the correlation of the evaluation results when the PSG test and the method of the present invention are applied at the same time. Shown (〇: actigraphy, ●: method of the present invention). (A) total sleep time, (b) mid-wake time, (c) sleep efficiency, (d) sleep onset latency. FIG. 6 shows the correlation of the evaluation results when the PSG test and the sleep profiler are applied to the same mentally ill patient at the same time. (A) total sleep time, (b) sleep efficiency, (c) non-REM sleep time, (d) REM sleep time, (e) light sleep time. FIG. 7 shows a typical sleep progress chart of a healthy person. FIG. 8 shows a sleep progress chart and sleep variables of a bipolar I disorder patient (case 1). (A) At the time of admission, (b) At the time of remission, (c) At the time of relapse. FIG. 9 shows a sleep progress chart and sleep variables of a bipolar disorder patient (case 2). (A) 1st time, (b) 2nd time. FIG. 10 shows a sleep progress chart and sleep variables of a depressed patient (case 3). (A) Before changing the therapeutic agent, (b) After changing the therapeutic agent. FIG. 11 shows a sleep progress chart and sleep variables of a hypnotic-dependent patient (case 4). (A) Before drug reorganization, (b) After drug reorganization. FIG. 12 shows a sleep progress chart and sleep variables of a schizophrenia patient (case 5). (A) Before the treatment change, (b) After the treatment change. FIG. 13 shows a sleep progress chart and sleep variables of a depressed patient (case 6). (A) Before the treatment change, (b) After the treatment change. FIG. 14 shows a sleep progress chart and sleep variables of a bipolar disorder (outpatient) patient (case 7). (A) Simultaneous measurement with PSG at hospitalization, (b) and (c) are home measurement. FIG. 15 shows the waveforms of the electroencephalograms acquired from the forehead and the electrodes attached to the skin on the temporal bone (posterior ear (mastoid process)). Upper: Forehead (40 μV / scale), Lower: Mastoid process (20 μV / scale), Measuring device: ZA-X (Pro Assist), Bandpass filter: 0.3-35 Hz (Recommended value by the American Society of Sleep Medicine)

1. 1. Sleep Evaluation Method The present invention is a method for evaluating sleep of a mentally handicapped patient, which includes a step of acquiring sleep information from an electroencephalogram signal of a patient using a device that electrically processes an electroencephalogram signal, and the sleep information. It includes a step of evaluating the patient's sleep by analyzing the above, and is characterized by using an electroencephalogram signal acquired from the skin on the patient's temporal bone, particularly the electrodes attached to the back of the left and right ears.

1.1 Electroencephalogram signal Electroencephalogram (EEG) is an electrical activity that occurs from the brain. EEG (electroencephalogram signal) is recorded by electrodes placed on the scalp, forehead, earlobe, and the like. In PSG examination, multiple electrodes are placed in the center of the head and the back of the head. In the sleep profiler, which is a simple electroencephalogram measuring device, electrodes are attached to the head, and in the sleep scope, electrodes are attached to the forehead (center of the forehead) and under the ears to measure the electroencephalogram.

In the present invention, electrodes are placed on the skin on the temporal bone, preferably behind the left and right ears, and more preferably on the mastoid process to acquire electroencephalogram signals. Attaching a small electrode to the above-mentioned site is less uncomfortable for the subject (patient) than the conventional placement of multiple electrodes on the scalp, and is shinobi even in patients with mental disorders who are sensitive to hyperesthesia and environmental changes. It is highly tolerant and enables accurate measurement of brain waves. In fact, in experiments conducted by the inventors, a sleep profiler with a relatively large device attached to the forehead of the face had a high patient dropout rate (59%), but with the method of the present invention, the dropout rate was 10 minutes. It decreased to 1 (6%), and it was possible to measure and evaluate in patients with mental disorders (Example 6).

As shown in Example 5 (FIG. 15) described later, the waveform of the electroencephalogram differs depending on the electrode mounting site, but the inventors obtained it from the electrodes attached to the skin on the temporal bone (behind the left and right ears). It was confirmed that the sleep information derived from the brain waves is comparable to the sleep information of the PSG test, which is medically recognized as a method for measuring sleep brain waves.

1.2 Device for Electrically Processing EEG Signals In the present invention, sleep information is acquired from a subject's EEG signal using a device for electrically processing EEG signals. Since the electroencephalogram signal is obtained from the skin on the subject's temporal bone, preferably behind the left and right ears, and more preferably from electrodes attached to the mastoid process, the subject is subject to physical restraint when acquiring the electroencephalogram with this device. No. Moreover, since the electrodes are small, the subject does not feel uncomfortable.

FIG. 1 is an example of a device that electrically processes an electroencephalogram signal. The device (device that electrically processes an electroencephalogram signal) 1 includes a display 2, an input operation button 3, electrodes 4 (4a, 4b, 4c), a visual warning sensor 5, and an auditory warning sensor 6. The device 1 includes a means for acquiring a brain wave signal from the subject, an analysis means for outputting an output measurement signal indicating the sleep stage of the subject in response to the brain wave signal acquired from the subject, and an analysis means in response to the output measurement signal. It has a thresholding means for outputting an output signal corresponding to the sleep stage, and the thresholding means is a means for determining a threshold value associated with the output measurement signal and corresponds to the physiological state of the sleep stage of the subject. Comparison of outputting the output signal indicating that the subject is in a physiological state by comparing the threshold value determining means that defines the range of the output measurement signal value with the threshold value determined by the output measurement signal and the threshold value determining means. Has means. The device can be rechargeable or battery-powered, does not need to be plugged into an electrical outlet for measurement, and does not require complex transmitters and receivers.

The electrode 4 composed of the "+ electrode" 4a, the "-electrode" 4b and the "com electrode" 4c is connected to the amplifier 11. Using the "com electrode" 4c, the reference point of the amplifier 11 is set to the same potential as the subject. The output of the subject's electroencephalogram signal is input to the amplifier 11 via the "+ electrode" 4a and the "-electrode" 4b. Reference numeral 12 denotes a device for digitizing the analog output signal of the amplifier 11. Reference numeral 13 denotes a single board computer, and all the computer hardware necessary for carrying out the method of the present invention is incorporated in the single board computer 13. Reference numeral 14 denotes a liquid crystal display with a keypad. The liquid crystal display 14 is used as both an output device (display and backlight) and an input device (keypad). Reference numeral 15 is a speaker, 16 is a main power source, and 17 is an amplifier power source. Since many devices are connected to the main power supply 16, electrical noise tends to occur in the system. Therefore, it is preferable to provide a power supply for the amplifier because it is not necessary to worry about the influence of electrical noise on the amplifier 11. The amplifier 11, the digitizing device 12, the single board computer 13, the liquid crystal display 14, the speaker 15, the main power supply 16, and the amplifier power supply 17 are incorporated in the device 1.

As shown in FIG. 1 (c), the “+ electrode” 4a and the “-electrode” 4b are the skin on the temporal bone, which is highly tolerated by the subject, preferably the left and right posterior ears, and more preferably the left and right mastoids. It is attached to the protrusion, and the "com electrode" 4c is attached to the back neck. An example of a suitable electrode is a self-adhesive electrode comprising a fixed gel adhesive hydrogel and a pre-mounted lead wire having a safety socket end. When the

electrodes

4a and 4b are attached to the above-mentioned sites, the subject's electroencephalogram signal is input to the device 1. Then, by inserting the key dedicated to the device 1 into a predetermined place of the device 1 and performing a necessary operation, the sleep information described later can be obtained by the algorithm installed in the device 1. These various sleep information can be displayed on the monitor screen of the personal computer by processing with special software or by using an Excel macro or the like created for analysis.

As an example of the device for electrically processing the electroencephalogram signal of the present invention, a Z machine manufactured by General Sleep can be mentioned.

1.3 Sleep information (sleep parameters)
Various types of sleep information can be obtained by a device that electrically processes an electroencephalogram signal. In the present specification, this quantified sleep information is referred to as a "sleep variable". That is, "sleep variable" means sleep parameters used in sleep science research and sleep clinical practice, for example, total bedtime, sleep latency, total sleep time, sleep time, sleep efficiency, light sleep time or the like. Appearance rate, deep sleep time or its appearance rate, REM sleep time or its appearance rate and latency, non-REM sleep time or its appearance rate, mid-wake time and its appearance rate, awakening response number, arousal response index (times / hour) , Sleep cycle, REM sleep interval, ratio of each sleep stage, appearance rate and power of each frequency band of brain wave, number of transitions of sleep stage, and other variables, but these are within the scope of the present invention. Not limited to. In the present specification, these "sleep variables" and non-quantified information such as the time series of transitions of sleep stages are collectively referred to as "sleep information (sleep parameters)".

Hereinafter, sleep information (sleep parameters) used in the present invention will be described, but sleep information is not limited thereto.
・ "Total bedtime" is the time from bedtime to waking up.
・ "Sleep onset latency" is the time required from awakening after bedtime to falling asleep. It is an indicator of drowsiness and good or bad sleep.
-The "total sleep time" is the actual sleep time, which is the time from falling asleep to awakening the next morning excluding halfway awakening.
-"Sleep efficiency" is the ratio of total sleep time to total bedtime (time from bedtime to waking up).
・ "REM sleep" is sleep accompanied by rapid eye movement (REM), in which the body is in a resting state, but the brain is in a wakeful state. The time from falling asleep to the appearance of REM sleep is called "REM sleep", and the appearance rate of REM sleep with respect to the total bedtime (Time in Bed: TIB) is called "REM sleep appearance rate". In addition to TIB, Total Sleep Time (TST) or Sleep Period Time (SPT) may be used as the denominator of the appearance rate.
-"Non-rem sleep" is sleep that does not involve rapid eye movement, and the appearance rate of non-rem sleep with respect to the total bedtime is called "non-rem sleep appearance rate". The denominator of the appearance rate is as described above.
-Non-rem sleep is divided into the following four stages 1 to 4 (judgment criteria by Rechtschaffen & Kales), stage 1 and stage 2 are called "light sleep", and stages 3 and 4 are called "deep sleep".
Stage 1: "Alpha wave is 50% or less" or "Waves of various frequencies with low amplitude are mixed" or "No bump wave is present"
Stage 2: "There is a low-amplitude irregular θ-δ wave or a high-amplitude slow wave (-)" or "There is a bump wave, a spindle wave or a K complex"
Stage 3: "Slow wave 20-50% below 2Hz and above 75μV" or "Spindle wave slows down and appears more widely"
Step 4: "Slow wave 50% or more of 2Hz or less and 75μV or more" or "Spindle wave (±)"
Every 30 seconds is determined as one unit (epoch). The longest sleep stage in 30 seconds is the epoch stage. For example, in a certain epoch, when 10 seconds is light sleep and the remaining 20 seconds is deep sleep, the epoch is determined to be deep sleep. The appearance rates of light sleep and deep sleep with respect to the total bedtime are referred to as "light sleep appearance rate" and "deep sleep appearance rate", respectively. The denominator of the appearance rate is as described above.
-"Mid-time awakening" means a state of waking up after falling asleep, and is indicated by the awakening time within the sleep time.
・ "Early morning awakening" means a state of waking up early in the morning and then unable to sleep.
-The "sleep cycle" is the time from falling asleep to the end of the first REM sleep, and then from the end of the REM sleep to the end of the next REM sleep.
・ "REM sleep interval" is the time from the end of REM sleep to the start of the next REM sleep.
・ "Brain wave frequency band" is a classification based on the frequency of brain waves, and is defined as δ wave: 0.5 to less than 4 Hz, θ wave: 4 to less than 8 Hz, α wave: 8 to less than 13 Hz, β wave: 13 Hz or more, etc. However, it is not limited to this.
-"Power" refers to the sum of each frequency band or all frequency bands among the power obtained by power spectrum analysis of brain waves.

Among the above sleep information (sleep variables), in particular, (i) sleep efficiency, (ii) REM sleep time or its appearance rate or REM sleep latency, and (iii) light sleep and / or deep sleep time or its appearance rate. , (Iv) sleep latency, (v) the number or frequency of sleep phase transitions, and the timeline of sleep phase transitions are suitable for sleep evaluation in patients with mental disorders, with (i)-(iii) being more preferred. Is. These sleep variables well reflect the sleep disorders characteristic of mentally ill patients. Example 4 using logistic regression analysis also shows that these sleep variables are useful for evaluation in sleep evaluation of psychiatric patients.

Since there is no index that can be quantitatively evaluated for mental disorders, doctors evaluate the medical condition based on the subjective evaluation of patients. On the other hand, sleep disorders are inevitable in psychiatric patients, but sleep can be objectively quantitatively evaluated, suggesting that it may serve as a surrogate marker.

1.4 Evaluation by sleep progress chart A sleep progress chart can be used for sleep analysis. The sleep progress chart is a diagram of the progress (transition) of the sleep stage in the sleep time, and the overall image (profile) of sleep can be easily visually grasped as an image. As shown in FIGS. 8 to 13, the sleep profile of a mentally handicapped patient is clearly different from that of a healthy person (FIG. 7), and the sleep profile and sleep variables change significantly depending on the course of symptoms and the therapeutic effect. .. Therefore, it is useful to use the sleep progress chart for analysis and evaluation of the whole picture of sleep.

For example, in the case of patients with insomnia, treatment was successful by extending total sleep time and deep sleep time, reducing awakening time and / or awakening frequency, increasing sleep efficiency, and reducing the number of sleep stage transitions. On the contrary, it is judged that the symptoms worsened due to shortening of total sleep time and deep sleep time, increase of awakening time and / or number of awakenings, decrease of sleep efficiency, increase of number of transitions of sleep stage, etc. These can be grasped visually and intuitively. In addition, it is possible to similarly grasp whether the sleep pattern of a healthy person is approached, for example, whether the cycle of light sleep, deep sleep, and REM sleep is regular, or whether there is deep sleep in the first half of sleep and REM sleep is increasing in the second half. ..

1.5 Evaluation by Logistic Regression Analysis Psychiatric patients often have sleep disorders, and their sleep profiles are distinguished from those of healthy individuals. The inventors analyze the sleep information of a mentally ill patient and a healthy person acquired in advance by logistic regression analysis, and input the sleep information of the subject (patient) into the determined logistic regression model to obtain the subject's sleep. We found that it is possible to easily evaluate whether it is closer to a healthy person or a mentally ill patient.

The logistic regression model is shown by the following general formula, and the value of the probability (p) obtained by inputting the subject's sleep information (sleep variable) into this formula is between 0 and 1, and the closer it is to 0, the healthier it is. The closer it is to a person, the closer to a mentally ill patient.

The sleep variables used are preferably information that is not easily affected by the environment such as hospitalization in the sleep evaluation of mentally handicapped patients. For example, sleep efficiency, REM sleep appearance rate, light sleep appearance rate, deep sleep appearance rate, and arousal. Ratios such as response index and latency such as REM sleep latency and sleep onset latency can be mentioned. In the usual case, a combination of sleep efficiency, REM sleep latency, and light sleep appearance rate, or a combination of sleep efficiency, REM sleep latency, light sleep appearance rate, and indwelling sleep is preferable.

The combination of sleep variables used can be changed as appropriate by selecting population information according to the target mental disorder or sleep disorder, and the logistic regression model (correlation coefficient and correlation coefficient) can be changed accordingly. Correlation function) is also set as appropriate.

The coefficient a_i and the constant term b of the sleep variable are determined according to each model and population. If the coefficient is positive, the larger the sleep variable, the closer the subject's sleep is to the sleep of the mentally ill patient, and if the coefficient is negative, the smaller the sleep variable, the more the subject's sleep. It contributes to the judgment that it is close to sleep of a mentally ill patient. The magnitude of the absolute value of the coefficient indicates the contribution of the sleep variable to the sleepiness of the mentally handicapped patient.

Regarding the probability (p), the value that is the boundary between the group of healthy subjects and the group of mentally handicapped patients is called the "threshold value". The "threshold" is determined according to each model and population. If the probability (p) is smaller than the threshold value, it can be evaluated that the patient's sleep is close to that of a healthy person. Also, when comparing the probabilities (p) of the same patient at two or more time points, if the probabilities (p) decrease, and as a result of the decrease, if the threshold is crossed or approaches the threshold, the patient's symptoms are closer to those of a healthy person. In other words, it can be evaluated as improved.

A preferable example for distinguishing between a mentally ill patient and a healthy person is a combination of sleep efficiency, REM sleep latency, and light sleep appearance rate (model 1). Another preferred example is the combination of four sleep variables (model 2), which is the sleep information plus the sleep onset latency. Furthermore, as a model for distinguishing the sleep of a mentally handicapped patient showing a special sleep profile from a healthy person, a combination of sleep efficiency, REM sleep appearance rate, and deep sleep appearance rate can be cited (Model 7). can.

(2) Model 1
Model 1 is represented by the following equation.

In model 1, the coefficients, a1, a2, a3 and the constant term b of each sleep variable are appropriately determined according to the population. In one embodiment, a1 = -1.299, a2 = 0.591, a3 = 0.531, b = -0.635.

In model 1, the threshold is 0.3 to 0.5, preferably 0.3 to 0.4, more preferably 0.3 to 0.36, 0.3 to 0.35, 0.3 to 0.34, for example 0.339. If the probability (p) obtained by inputting the patient's sleep information (sleep variable) into the model 1 is higher than the threshold value, it can be said that the patient's sleep is close to that of a mentally ill patient. If the probability (p) becomes lower than the threshold value as a result of the treatment, it can be said that the sleep of the patient approaches that of a healthy person and is normalized.

(3) Model 2
Model 2 is represented by the following equation.

In model 2, the coefficients, a1, a2, a3, a4 and the constant term b of each sleep variable are appropriately determined according to the population. In one embodiment, a1 = -1.263, a2 = 0.604, a3 = 0.532, a4 = 0.064, b = -0.635.

In model 2, the threshold is 0.3 to 0.5, preferably 0.3 to 0.4, more preferably 0.3 to 0.36, 0.3 to 0.35, 0.3 to 0.34, for example 0.335. If the probability (p) obtained by inputting the patient's sleep information (sleep variable) into the model 2 is higher than the threshold value, it can be said that the patient's sleep is close to that of a mentally ill patient. If the probability (p) becomes lower than the threshold value as a result of the treatment, it can be said that the sleep of the patient approaches that of a healthy person and is normalized.

(4) Model 7
For a group of patients showing a specific sleep profile or a therapeutic drug, it is possible to set a model according to the sleep profile. For example, unlike benzodiazepine hypnotics that increase light sleep, orexin receptor inhibitors increase deep sleep and REM sleep. Patients treated with orexin receptor inhibitors have increased deep and REM sleep if they respond to treatment. Examples of orexin receptor inhibitors include, but are not limited to, suvorexant and lemborexant. Model 7 is a logistic regression model in which the incidence of deep sleep and REM sleep was determined as sleep information in order to evaluate the sleep of patients treated with orexin receptor inhibitors that increase deep sleep and REM sleep. be.

Model 7 is represented by the following equation.

In model 7, the coefficients, a1, a2, a3 and the constant term b of each sleep variable are appropriately determined according to the population. In one embodiment, a1 = -0.742, a2 = -0.744, a3 = -0.307, b = -0.681.

In model 7, the threshold is 0.3 to 0.5, preferably 0.35 to 0.45, more preferably 0.35 to 0.44, 0.36 to 0.43, 0.37 to 0.42, 0.38 to 0.42, 0.39 to 0.42, for example 0.409. Model 7 is preferred for patients treated with orexin receptor inhibitors. In patients treated with orexin receptor inhibitors, if the probability (p) is lower than the threshold, the patient can be said to have been successfully treated.

1.6 Evaluation by machine learning model XGBoost The inventors also found that more accurate evaluation is possible by using the number of transitions of each sleep stage as an index showing the characteristics of the sleep profile.

The value of the probability (p) obtained by inputting the subject's sleep information (sleep variable) is between 0 and 1, and the closer it is to 0, the closer it is to a healthy person, and the closer it is to 1, the closer it is to a mentally ill patient.

The sleep variables used are preferably information that is not easily affected by the environment such as hospitalization in the sleep evaluation of mentally handicapped patients. For example, sleep efficiency, REM sleep appearance rate, light sleep appearance rate, deep sleep appearance rate, and arousal. Examples include ratios such as response index, number or frequency of sleep stage transitions, and latency such as REM sleep latency and sleep onset latency.

It is preferable that the number or frequency of transitions of the sleep variable between sleep stages is evaluated for each stage. In the present embodiment, the frequency of transitions between sleep stages is assumed to be the number of transitions between each sleep stage divided by the total bedtime, and 12 types (awakening → REM sleep, awakening → light sleep, awakening → deep sleep, REM sleep). → Awakening, REM sleep → Light sleep, REM sleep → Deep sleep, Light sleep → Awakening, Light sleep → REM sleep, Light sleep → Deep sleep, Deep sleep → Awakening, Deep sleep → Rem sleep, Deep sleep → Light sleep) Although it is used, the number of transitions between each sleep stage may be divided by the total sleep time.

The combination of sleep variables to be used can be appropriately changed by selecting population information according to the target mental disorder or sleep disorder, and the machine learning model is appropriately set accordingly. .. In this embodiment, XGBoost is used as a machine learning model.

FIG. 2A shows an example of the information analysis device of the present invention. FIG. 2B is a flowchart showing a processing example of the processor 21. In step S11, the processor 21 acquires a set of data each composed of a plurality of data. That is, the processor 21 acquires a plurality of data via the storage device 22 or the communication circuit 23.

In step S12, the processor 21 calculates the sleep variable, which is a feature amount of each of the plurality of data acquired in step S11, and obtains a probability (p) by a machine learning model using the calculated sleep variable as an input. If the data acquired in step S11 is for a healthy person, it is set to 0, and if it is for a mentally handicapped patient, it is set to 1, and a loss function determined by an error from the probability (p) calculated by the machine learning model is calculated. Then, the parameters of the machine learning model are modified (the weights of each parameter are updated) so that the loss function is reduced by the gradient boosting method. In this embodiment, a gradient boosting tree is used with gbtree as a Booster, and parameters are updated based on the loss function binary cross entropy error, but the loss function may be a mean square error or the like. The binary cross entropy error E can be expressed by the following equation, where tk is the teacher data and pk is the predicted value by the machine learning model.

The model performance is evaluated by AUC (Area Under the Curve), the model parameter that maximizes the evaluation index is determined, and stored via the storage device 22 or the communication circuit 23.

Regarding the probability (p), the value that is the boundary between the group of healthy subjects and the group of mentally handicapped patients is called the "threshold value". The "threshold value" is determined in step S13 according to each model and population. If the probability (p) is smaller than the threshold value, it can be evaluated that the patient's sleep is close to that of a healthy person. Also, when comparing the probabilities (p) of the same patient at two or more time points, if the probabilities (p) decrease, and as a result of the decrease, if the threshold is crossed or approaches the threshold, the patient's symptoms are closer to those of a healthy person. In other words, it can be evaluated as improved.

In step S14, the processor 21 reads the machine learning model saved in step S12 via the storage device (storage unit) or the communication circuit.

In step S15, the processor 21 acquires the data (input data) to be the identification target data from the storage device (storage unit) or the like, and the processor 1 extracts the feature amount from the identification target data and reads it in step S14. The probability (p) is calculated by the machine learning model. The closer p is to 0, the closer to a healthy person, and the closer to 1 is, the closer to a mentally ill patient.

In step S16, the processor 21 compares the probability calculated in step S15 with the threshold value set in step S13, and makes a determination. At this time, if the probability is equal to or higher than the threshold value, the patient is determined to be a mentally ill patient. In steps S17 and S18, the processor 21 outputs the determination result obtained in step S16 (for example, it is displayed on the display 25).

1.7 Evaluation by Other Machine Learning Models In the present invention, analysis / evaluation of sleep information (including analysis of sleep progress chart) may be performed by machine learning other than logistic regression analysis or XGBoost (supervised learning). .. Machine learning may be unsupervised learning, supervised learning, or semi-supervised learning. As unsupervised learning, for example, k-means clustering, hierarchical clustering, neural network, and the like can be preferably used. Supervised learning includes, for example, k-nearest neighbor method, support vector machine, decision tree, ensemble learning (Random Forest, XGBoost, LightGBM, Ada Boost, etc.), linear discrimination, quadratic discrimination, naive bayes, logistic regression, neural network. As semi-supervised learning such as a self-organizing map, for example, a neural network or the like can be preferably used.

1.8 Evaluation of time-series sleep information by machine learning model Mental illness patients often have sleep disorders, and their sleep profiles are distinguished from those of healthy people. By inputting the sleep information of the subject (patient) into the machine learning model determined from the sleep information of the mentally ill patient and the healthy person acquired in advance, the inventors can make the subject's sleep of the healthy person and the mentally ill patient. We found that it is possible to easily evaluate which one is closer.

In this embodiment, a neural network encoder and classifier are used as a machine learning model. In step S11 of FIG. 2B, the processor 21 acquires a set of data each composed of a plurality of data. That is, the processor 21 acquires a plurality of data from the network via the storage device 22 or the communication circuit 23.

In step S12, the processor 21 has a probability (p) calculated from the feature quantity obtained by the encoder from the input data using the classifier, 0 when the input data is that of a healthy person, and a mental disorder. If it belongs to a patient, it is set to 1, and the loss function is calculated from the difference from these. The processor 1 corrects the parameters of the encoder and the classifier (updates the weights of each parameter) so as to reduce the loss function by the error backpropagation method (backpropagation), determines the parameters of the machine learning model, and determines the parameters of the machine learning model. It is stored via the storage device 22 or the communication circuit 23.

When the cross entropy error is used as the loss function, the cross entropy error E can be expressed by the following equation, where tk is the teacher data and pk is the predicted value by the machine learning model.

In step S14, the processor 21 reads the machine learning model stored in step S12 via the storage device 22 or the communication circuit 23.

In step S15, the processor 21 acquires the data (input data) to be the identification target data from the storage device 2, the network, or the like, and the processor 21 extracts the feature amount from the identification target data by the encoder from the feature amount. Calculate the probability (p) with the classifier. The closer p is to 0, the closer to a healthy person, and the closer to 1 is, the closer to a mentally ill patient.

In step S16, the processor 21 compares the probability calculated in step S15 with the threshold value set in step S13, and makes a determination. At this time, if the probability is equal to or higher than the threshold value, the patient is determined to be a mentally ill patient. In steps S17 and S18, the processor 21 outputs the determination result obtained in step S16 (for example, it is displayed on the display 25 or transmitted to the network).

In this embodiment, Transformer, which is one of the deep learning models, is used as the encoder, but CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), etc. can also be used. The error back propagation method is performed by calculating a loss function that is determined by the error between the output value and the input data obtained by inputting the feature quantity obtained from the encoder into the classifier and whether it is a healthy person or a mentally handicapped person. The parameters of the encoder and the classifier are modified to reduce the loss function. The loss function can use the cross entropy error, the mean square error, and the like.

In order to realize the classification of overnight sleep as a time series, the judgments of one or more consecutive epochs are collectively digitized (vectorized) into a encoder consisting of one or more Transformer layers. input. A feature amount for the input data is obtained from the encoder, and the probability (p) is calculated from the feature amount by the classifier. The classifier consists of one or more fully connected layers.

In sleep evaluation, time-series information (pattern of sleep progress chart) of sleep progress in which each sleep stage appears and shifts is considered to be important. Therefore, when the sleep information acquired in advance is the transition of the sleep stage of the overnight time series, more accurate sleep evaluation becomes possible by constructing a machine learning model that handles it as time series data.

2. Treatment Support for Psychiatric Patients The present invention also provides a method for supporting (assisting) the treatment of a mentally ill patient. The method includes a step of evaluating the sleep of a patient according to the method described in 1 above, and a step of evaluating the symptom of mental disorder of the patient or the degree thereof based on the evaluation result.

Alternatively, it includes a step of evaluating the sleep of the patient according to the method described in 1 above, and a step of selecting a preferable treatment based on the evaluation result.

For example, if the arousal level is high and there are many awakenings and light sleep, specifically, if the sleep onset latency is 30 minutes or more and the bedtime or less, or if the sleep efficiency is less than 75%, orexin is accepted. Orexins are presented as the preferred treatment option. If the patient's sleep profile does not change despite treatment, change to a treatment with a different effect or consider modified electroconvulsive therapy.

According to the present invention, by evaluating sleep before dosing, a hypnotic is administered to a patient (Shiro Endo, Psychiatry and Neurology Magazine 64, 173-707, 1962) who complains of excessive insomnia by subjective evaluation. It can be avoided. In addition, by evaluating the degree of improvement in sleep after administration, long-term administration when there is no effect can be avoided, and by evaluating the degree of improvement in sleep after administration, an appropriate dose can be selected.

3. 3. Treatment Support System for Mental Illness Patients The present invention also provides a treatment support system for mentally ill patients. FIG. 2C shows an outline of the treatment support system of the present invention. As shown in FIG. 2, the treatment support system of the present invention includes an electroencephalogram signal processing device (a device that electrically processes an electroencephalogram signal) and an information analysis device. The electroencephalogram signal processing device is a simple device independent of the information analysis device, and is used at the home of the patient (subject), etc., and as described in 1 above, the skin on the temporal bone of the patient, preferably the left and right ears. The electroencephalogram signal obtained from the electrodes attached posteriorly, more preferably on the mastoid process, is electrically processed. The information analysis device includes a storage unit that acquires and stores sleep information (sleep parameters) of a patient processed by the electroencephalogram signal processing device, an analysis unit that analyzes and evaluates sleep information stored in the storage unit, and an analysis unit. It is equipped with an output unit that outputs analysis / evaluation results. The analysis unit evaluates the patient's sleep according to the present invention, or evaluates the symptoms or degree of the patient's psychiatric disorder, or selects a preferable treatment, and together with or separately from the patient's sleep information. The evaluation result is output to. The analysis / evaluation may be carried out by machine learning. The output data is sent to a printer or display for printing or display. The printer and the display may be a part of the information analysis device or may be connected to the outside.

4. Information Analyst Device for Supporting Treatment of Psychiatric Disorders The present invention also provides an information analyzer used in the treatment support system of the present invention. FIG. 2C shows an outline of the treatment support system of the present invention. As shown in FIG. 2C, the information analysis device of the present invention is stored in a storage unit (storage device) for storing sleep information of the patient processed by a device that electrically processes a brain wave signal, and a storage unit. It is equipped with an analysis unit (processor) that analyzes / evaluates sleep information and an output unit that outputs analysis / evaluation results. Does the analysis unit perform sleep evaluation according to the sleep evaluation method described in 1 above? , Or, evaluate the symptoms or degree of mental disorders of the patient, or select the preferred treatment, and output the evaluation results together with or separately from the sleep information of the patient. The analysis / evaluation may be carried out by machine learning. As described above, the printer and the display for printing or displaying the output data may be a part of the information analysis device or may be connected to the outside.

5. Treatment Support Program for Mental Illness Patients The present invention also provides a program for executing processing in the information analyzer of the present invention. Specifically, the program of the present invention acquires the sleep information of the patient processed by the device that electrically processes the brain wave signal, stores the sleep information, and evaluates / analyzes the stored sleep information. The evaluation / analysis result is output, and the processing is executed by the computer. The evaluation / analysis evaluates the sleep of the patient according to the present invention, or the symptom of the patient's mental disorder. Or it involves assessing the extent of the drug, assessing the effects of drug withdrawal, or selecting the preferred treatment.

6. Other Applications According to the method of the present invention, the sleep of a mentally handicapped patient can be easily and objectively evaluated with almost the same accuracy as the PSG test. Unlike the PSG test, the method of the present invention does not restrain the body of the subject, so that the evaluation can be performed under the same conditions as in a normal sleeping environment without showing extraordinary effects such as the first night effect. There is an advantage that it can be done. Therefore, it can be used not only for sleep evaluation and treatment support for psychiatric patients but also for various clinical purposes.

For example, the method of the present invention can be used as a diagnostic aid (method for assisting diagnosis) of a mentally handicapped patient using sleep. For example, by using the method of the present invention to objectively evaluate sleep in a psychiatric patient, the psychiatric disorder can be diagnosed and appropriate intervention can be performed earlier. In the present invention, since the electroencephalogram is measured using a small electrode attached on the temporal bone, there is little burden associated with the measurement and measurement errors and failures due to the measurement, and accurate sleep information can be easily obtained even at home. Can be done. Therefore, outpatient sleep can be assessed at home without the need for hospitalization.

In addition, the method of the present invention can be used to select subjects for clinical trials such as sleeping pills. For example, in a clinical trial of a hypnotic drug for insomnia patients, by using the method of the present invention to objectively evaluate sleep before a definitive diagnosis by a PSG test, a subject suitable for the clinical trial can be quickly and easily selected. Can be selected for.

Hereinafter, the present invention will be specifically described with reference to Examples, but the present invention is not limited to these Examples.

Example 1: Verification of validity of sleep evaluation by the method of the present invention 1. Correlation between sleep evaluation by the method of the present invention and sleep evaluation by the PSG test In order to verify the validity of the sleep evaluation method by the present invention, a comparison with the PSG test was performed. Syndrome, bipolar disorder, depression, Lewy body dementias / REM sleep behavior disorder, sleep disorder, attentionlessness / hyperactivity disorder, autism spectrum disorder, etc. (adjustment disorder, panic disorder, dissection disorder, delusional disorder, For 37 mentally ill patients with at least one condition (such as adjustment disorder), sleep evaluation by PSG test and sleep evaluation by the method of the present invention were simultaneously performed, and the correlation was observed.

In order to carry out the method of the present invention, a Z machine (General Sleep) was used as an electroencephalogram signal processing device. Electrodes were attached to the skin on the patient's temporal bone (behind the left and right ears (mastoid process)) and the neck, and brain waves were measured according to the protocol recommended by the manufacturer to obtain sleep information.

The PSG test and the method of the present invention have a good correlation in all of total sleep time, sleep efficiency, REM sleep time, non-REM sleep time, mid-wake time, sleep onset latency, and light sleep time (stages N1 and N2). (Fig. 3 and Table 1)

2. Correlation between sleep evaluation by actigraphy and sleep evaluation by PSG test For the same 37 cases as in Example 1, sleep evaluation by PSG test and sleep evaluation by actigraphy (small acceleration sensor with a wristwatch structure) were performed at the same time. I saw the correlation.

The results were plotted in comparison with the correlation between the PSG test and the method of the present invention (Example 1). The correlation between the PSG test and polysomnography is as follows: (a) total sleep time, (b) mid-wake time, (c) sleep efficiency, (d) as compared to the correlation between the PSG test and the method of the present invention. It was confirmed that it was inferior in all of the sleep onset latency (Fig. 5).

3. 3. Correlation between sleep evaluation by sleep profiler and sleep evaluation by PSG test For 4 cases, sleep evaluation by PSG test and sleep evaluation by sleep profiler (Advanced Brain Monitoring Co., Ltd.) were performed at the same time, and the correlation was observed.

Shows the correlation between the PSG test and sleep evaluation by the sleep profiler. The correlation between the PSG test and sleep evaluation by the sleep profiler is as follows in any of (a) total sleep time, (b) sleep efficiency, (c) non-REM sleep time, (d) REM sleep time, and (e) light sleep time. , It was confirmed that it was not good (Fig. 6).

Example 2: Examination of the first night effect in the method of the present invention 1. First night effect by the method of the present invention Generally, in a PSG test, the quality of sleep is affected by measurement in an environment different from the normal sleeping environment, such as many sensors and electrodes being attached to the test subject. Often appears. This is called the first night effect. For example, "increased arousal, decreased total sleep time, decreased deep sleep, decreased sleep efficiency" in healthy adults, "increased arousal, decreased sleep efficiency, decreased REM sleep" in insomnia patients, depression. For sick patients, "decrease in total sleep time, decrease in sleep efficiency, increase in awakening time, decrease in REM sleep", and in patients with autism spectrum disorder, "decrease in stage N2, increase in awakening, decrease in sleep efficiency", etc. The first night effect often appears. Therefore, for 14 healthy adults, brain waves were measured for 6 days from electrodes attached to the skin on the temporal bone (posterior left and right ears (mastoid process)) using a Z machine, and total sleep time, sleep efficiency, and so on. We observed changes in the parameters of light sleep time ratio, deep sleep time ratio, and REM sleep time ratio.

As shown in Table 2, according to the method of the present invention, the quality of sleep (total sleep time, sleep efficiency, light sleep time ratio, deep sleep time ratio, REM sleep time ratio) from the first night to the sixth night. It can be seen that there is no change in and the first night effect does not appear.

2. First night effect by PSG test PSG test was conducted for healthy adults (1 case) and mentally handicapped patients (sleep / breathing disorder group, insomnia disorder group, motor / behavior disorder group, 1 case each), and (a) sleep The sleep qualities of the first and second nights were compared for efficiency, (b) mid-wake time, (c) deep sleep time, and (d) REM sleep time.
As shown in FIG. 4, it can be seen that the quality of sleep changes between the first night and the second night. Thus, in PSG tests, the first night effect often appears.

Example 3: Sleep evaluation in psychiatric patients In 6 psychiatric patients, sleep information was acquired using a Z machine, a sleep progress chart was created, and the relationship between symptoms and treatment transitions and sleep was analyzed. For comparison, FIG. 7 shows a typical sleep profile (sleep progress chart) of a healthy person.

The abbreviations used in this example are shown below.
BDI: Beck Depression Inventory
HA: Harm Avoidance
PSQI: Pittsburgh Sleep Quality Index
ESS: Epworth Sleepiness Scale
HAMD: Hamilton Depression Scale
YMRS: Young Mania Rating Scale
BPRS: Brief Psychiatric Rating Scale

1. 1. Case 1: Bipolar Disorder Type I Bipolar Disorder Type I patients (female, 40s) were evaluated for sleep at the following three time points.
(a) At the time of admission (June X)
BDI 43 points, HA 20 points, PSQI 11 points, ESS 3 points, HAMD 17 points, YMRS 3 points Subjective sleep evaluation: sleep onset latency 50 minutes, total sleep time 480 minutes, awakening time 20 minutes Dosing: Lithium carbonate 1000 mg / day, clonazepam 2 mg / day, quetiapine fumarate 300 mg / day, quetiapine 50 mg / day, sodium valproate 600 mg / day
(b) At remission (14 days after (a))
BDI 43 points, HA 20 points, PSQI 4 points, ESS 1 point, HAMD 4 points, YMRS 0 points Subjective sleep evaluation: sleep onset latency 20 minutes, total sleep time 510 minutes, awakening time 3 minutes Dosing: Lithium carbonate 600 mg / day, clonazepam 2 mg / day, quetiapine fumarate 300 mg / day, sodium valproate 800 mg / day
(c) Relapse (February X + 1)
BDI 34 points, HA 19 points, PSQI 15 points, ESS 2 points, HAMD 20 points, YMRS 2 points Subjective sleep evaluation: sleep onset latency 40 minutes, total sleep time 300 minutes, awakening time 10 minutes Dosing: Ribotril (clonazepam) ) 2 mg, Na valproate 600 mg, Remus (lithium carbonate) 600 mg, Versomura (svorexant) 20 mg, Caronal (acetaminophen) 400 mg, Bioslee, Coronel (polycarbofyl calcium) 1500 mg, Strattera capsule (atomoxetine) 120 mg, Carbocysteine 1500 mg, FAD enteric coated tablets 30 mg, vibramycin (doxycycline) 200 mg, bepotastine besilic acid 20 mg, bamboo cab (vovorexant fumarate) 20 mg

The patient presented with insomnia, loss of appetite, decreased motivation, etc. and was hospitalized in the depressive phase. At the time of admission, the severity of depressive symptoms (HAMD) was 17 points and the severity of manic symptoms (YMRS) was 3 points. The severity of the symptom was 0 points, leading to remission, and the patient was discharged from the hospital. However, the patient was re-hospitalized due to relapse of symptoms due to interpersonal stress and worsening of insomnia, loss of appetite, and malaise. Depressive symptoms on readmission worsened to 20 points and manic symptoms to 2 points.

Figure 8 shows the sleep progress chart and sleep variables of Case 1. From admission to remission, along with improvement of symptoms, the sleep progress chart also shows that total sleep time, deep sleep, and sleep efficiency increased, approaching the sleep profile of healthy subjects. At the time of relapse, total sleep time, sleep efficiency, and deep sleep are all reduced. As symptoms worsen, total sleep time and deep sleep time decrease, sleep onset latency increases, sleep efficiency decreases, and psychotic symptoms and changes in sleep profile are consistent.

These results show that by combining the evaluation of sleep by the method of the present invention with treatment, not only the therapeutic effect at the time of intervention can be visualized, but also changes in pathological conditions such as hatred, relapse, and recurrence can be quantified and visualized.

2. Case 2: Bipolar disorder Sleep was evaluated in patients with bipolar disorder (male, 60s) at the following two time points.
(a) First time (October X)
BDI 30 points, HAMD 15 points, YMRS 6 points, HA 15 points, PSQI 17 points, ESS 8 points Dosing: Valproic acid 800 mg, Alprazolam 0.4 mg
(b) 28 days after the second ((a))
BDI 18 points, HAMD 6 points, YMRS 2 points, HA 17 points, PSQI 12 points, ESS 9 points Dosing: Valproic acid 800 mg, alprazolam 0.4 mg, suvorexant 20 mg

The patient was hospitalized in the depressive phase of bipolar disorder and presented with depressive symptoms such as insomnia, decreased motivation, and malaise. In addition to rest and improvement of life rhythm, insomnia symptoms persisted. Therefore, by adding suvorexant (hypnotic), the symptoms gradually improved and sleep also improved. The severity of depressive symptoms (HAMD) improved from 15 points to 6 points, and the severity of manic symptoms (YMRS) also improved from 6 points to 2 points.

Figure 9 shows the sleep progress chart and sleep variables of Case 2. As the symptoms improve, total sleep time and deep sleep time increase, and sleep efficiency also improves. These results indicate that the sleep profile obtained by the method of the present invention may sensitively reflect changes in clinical symptoms due to treatment.

3. 3. Case 3: Depression For depressed patients (female, 50s), sleep was evaluated at the following two time points.
(a) Before change of therapeutic drug (April X)
YMRS 26 points (most severe)
Mirtazapine 30mg / day, olanzapine 2.5mg / day, flunitrazepam 1mg / day
(b) After changing the therapeutic drug (May X)
YMRS 3 points (normal range)
Escitalopram 20 mg / day, olanzapine 5 mg / day, trazodone 50 mg / day

Olanzapine (antipsychotic) and flunitrazepam (benzodiazepine hypnotic) were prescribed for insomnia, centering on mirtazapine (antidepressant), but the symptoms did not improve, and sleep onset latency was long and deep. There was little sleep.
After changing the main antidepressant drug to escitalopram, which has a different mechanism of action, discontinuing the benzodiazepine hypnotic that increases light sleep, and introducing trazodone (antidepressant) that increases deep sleep, the severity of depression is severe. Depression (HAMD) improved from 26 points to 3 points.

The sleep progress chart and sleep variables of Case 3 are shown in FIG. Before the prescription change, the sleep onset latency was long and the deep sleep was small, but after the prescription change, the sleep onset latency was shortened, the deep sleep increased, and a sleep profile close to that of a healthy person was observed. The sleep evaluation according to the present invention is consistent with changes in clinical symptoms and is useful for selecting a therapeutic agent and confirming its effect.

4. Case 4: Hypnotic use disorder Patients (female, 30s) with hypnotic use disorder were evaluated for sleep at the following two time points.
(a) Before drug reorganization (August X)
Medication: Bromazepam (bromovaleryl uric acid) 1.8 g (monoureid hypnotic), isomitar (amobarbital) 0.4 g (barbituric acid hypnotic), arten (trihexifenidil) 6 mg (anticholinergic drug), tofranil (imiplamine) 75 mg ( Tricyclic antidepressant), Sepazone (cloxazolam) 6 mg and Lexotan (bromazepam) 5 mg (all are benzodiazepine anxiolytics).
(b) After drug arrangement (September X)
Dosing: trazodone 200 mg, tofranil (imipramine) 25 mg, sertraline hydrochloride 50 mg (above, antidepressant), sodium valproate 600 mg (antiepileptic drug)

In addition to panic disorder, a case in which strong insomnia persisted, falling into a polypharmacy prescription of sleeping pills, and becoming dependent on sleeping pills. At the time of admission, in addition to the two benzodiazepine anxiolytics, six drugs including barbituric acid hypnotics, antidepressants and anticholinergic drugs were prescribed. Careful sleep hypnotic guidance and listening to anxiety while gradually reducing barbituric acid hypnotics and benzodiazepines, instead starting trazodon (antidepressant) to increase deep sleep, gradually increasing, benzodiazepines Sodium valproate (an antidepressant) was introduced to deal with the withdrawal symptoms of the drug, and the drug for panic disease could be sorted out.

The sleep progress chart and sleep variables of Case 4 are shown in FIG. Drugs were rearranged while performing sleep evaluation according to the present invention. By visualizing the sleep state and improvement and feeding back the sleep status to the patient, the patient's anxiety was also improved, and it became possible to prescribe an appropriate hypnotic drug.

5. Case 5: Schizophrenia Hypnotic use disorder patients (female, 30s) were evaluated for sleep at the following two time points.
(a) Before change of treatment (December X)
BPRS 39 points, ESS 0 points, PSQI 9 points, BDI 17 points, HA20 16 points Aware sleep: I couldn't sleep, I felt a deep sleep 3, sleepiness 2, sleep onset latency 90 minutes, sleep time 270 minutes, awakening time 60 minutes Dosage: Blonanserin 20 mg, Flunitrazepam 2 mg, Suvorexant 20 mg, Propranolol hydrochloride 30 mg
(b) After treatment change (14 days after (a))
BPRS 21 points, ESS 4 points, PSQI 8 points, BDI 14 points, HA20 19 points Aware sleep: sleep well, deep sleep 8, drowsiness 1, sleep onset latency 60 minutes, sleep time 480 minutes, awakening time 30 minutes Dosage: Blonanserin 8 mg, Blonanserin patch 40 mg, Flunitrazepam 1 mg, Suvorexant 20 mg, Proplanolol hydrochloride 30 mg

Modified electroconvulsive therapy (4 out of 5 as of 12/25) with almost no change in drug prescription improved BPRS, which indicates the severity of psychiatric symptoms, from 39 points to 21 points. bottom. As for subjective sleep evaluation, deep sleep increased from 3 to 8, drowsiness decreased from 2 to 1, and subjective sleep latency, sleep time, and awakening were also improved.

The sleep progress chart and sleep variables of Case 5 are shown in FIG. As the symptoms improved with modified electroconvulsive therapy, sleep onset latency was shortened, light sleep time, deep sleep time, and REM sleep time were extended, and sleep efficiency increased from 76% to 83.3%. The sleep profile (sleep information) obtained by the method of the present invention also correlates with the subjective sleep evaluation and shows that it is useful as auxiliary information for treatment.

6. Case 6: Depression Inpatients with depression (female, 30s) were evaluated for sleep at the following two time points.

(a) Before change of treatment (February X)
BDI 24 points, HA 15 points, PSQI 11 points, ESS 6 points, HAMD 19 points Subjective sleep: sleep onset latency 30 minutes, total sleep time 480 minutes, midway awakening time 10 minutes Dosing: Lexapro (escitalopram) 10 mg, Versomura ( Suvorexant) 20 mg, Bredonin (prednisolone), Pariet (labeprazole), Alphacalcidol, Prograf (tacrolimus), Celcept (mycophenolate mofetil), Slokey (potassium chloride)
(b) After changing treatment (March X)
BDI 41 points, HA 15 points, PSQI 9 points, ESS 5 points, HAMD 18 points Subjective sleep: sleep onset latency 3 minutes, total sleep time 420 minutes, midway awakening time 5 minutes Dosing: Versomura (susvorexant) 20 mg, brothisolam 0.25 mg, sine balta (duloxetine) 40 mg, bledonin (prednisolone), pariet (labeprazole), alphacalcidol, prograf (tacrolimus), cerucept (mycophenolate mofetil), slow cake (potassium chloride)

Although drug therapy was performed, the severity of depression at the first measurement was 19 points (Hamilton Depression Rating Scale), and the severity at the second measurement after changing the main antidepressant drug was 18 points. No change was observed in the image.

The sleep progress chart of Case 6 is shown in FIG. There was no significant change in the sleep progress chart, there was no significant change in total sleep time and sleep efficiency, and the deep sleep time was rather short. This indicates that the sleep information obtained by the method of the present invention changes in conjunction with the pathophysiology of depression and is useful for evaluation of treatment.

7 Case 7: Bipolar disorder Sleep was evaluated at the following three time points in outpatients with bipolar disorder (male, 50s).
(A) At the time of admission for examination (July X)
PSQI 10 points, ESS 10 points, BDI-2 36 points, HA 13 points, HAMD 6 points, YMRS 2 points Subjective sleep time: 300 minutes, Awakening time: 60 minutes, Sleep onset latency 0 minutes Suvorexant 20 mg, Eszopiclone 3 mg , Quetiapine fumarate 25 mg, lamotrigine 400 mg, lithium carbonate 800 mg

(B and c) Home measurement (X + November 2 and 2 days later)
PSQI 13 points, ESS 14 points, BDI-2 25 points, HA 18 points, HAMD 9 points, YMRS 2 points (b) Aware sleep time: 300 minutes, awakening time: 0 minutes, sleep onset latency 30 minutes (c) Subjective sleep time: 270 minutes, awakening time: 90 minutes, sleep onset latency 0 minutes ramelteon 8 mg, eszopiclone 3 mg, lurasidone hydrochloride 60 mg, lamotrigin 400 mg, lithium carbonate 800 mg

The sleep progress chart of Case 7 is shown in FIG. Since there were complaints of light sleep and lack of sleep, sleep was measured at the time of admission to the PSG test. After that, the patient himself measured his sleep at home for two nights because he complained that he did not sleep well and that he did not fall asleep well. In both cases, there was a discrepancy between subjective sleep evaluation and objective sleep evaluation.

Example 4: Distinguishing between healthy and mentally ill patients by the method of the present invention A combination of parameters that can accurately distinguish between inpatients and healthy subjects from the sleep variables of inpatients (92 nights) and healthy subjects (158 nights). In order to find out, the following examination was conducted.

(1) Preliminary study of all 17 sleep variables Since bedtime and wake-up time are fixed in the hospital ward, the total bedtime is considered to be a variable that depends on the hospitalization environment. Therefore, the following seven sleep variables were selected, excluding sleep variables that are considered to be affected by the hospitalization environment such as total bedtime.
・ Ratio: Sleep efficiency, REM sleep appearance rate, light sleep appearance rate, deep sleep appearance rate, arousal response index ・ Latency: REM sleep latency, sleep onset latency

(2) Regression analysis for 7 items of sleep variables Logistic regression was performed by dividing the data of inpatients and healthy subjects excluding repeated measurement examples into 5 groups, 4 groups as a training set, and the remaining 1 group as a test set. Analysis was performed and cross-validation was performed.

Table 3 shows an example of a model in which each model was evaluated using AUC as an index and a high AUC was obtained. Among these models, a three-item formula (model 1: AUC average 0.811) and a four-item formula (model 2: AUC average 0.809) are evaluated as medically valid and highly accurate indexes. Selected as.

For model 1 and model 2, a threshold value was set to separate patients from healthy subjects so that the average of false negative rate and false positive rate was minimized. The sleep variables, coefficients, and constant terms in equation (I) of model 1 and model 2 are shown in Tables 4 and 5, respectively. The sleep variables of Model 1 (sleep efficiency, REM sleep latency, light sleep appearance rate) are indicators that are likely to worsen in psychiatric patients. The sleep variable added in model 2 (sleep onset latency) is a treatment-sensitive indicator.

(3) Verification of Model 1 and Model 2 According to the method of the present invention, sleep variables of 6 inpatients of Example 3 were acquired, evaluated by Model 1, and compared with clinical severity. The results are shown in Table 6.

In all 6 cases, the index at the time of the first measurement exceeded the threshold value and was judged to be a patient. The index was below the threshold in 3 of the 4 cases in which the therapeutic effect was clear after treatment, but was above the threshold in 1 case. On the other hand, it did not fall below the threshold in the cases of recovery tendency even after the treatment and the cases in which the therapeutic effect was unclear. From the above, it was confirmed that both the constructed model 1 and model 2 are valid as formulas for distinguishing between patients and healthy subjects.

(4) Model 7 (special sleep profile)
When the sleep profile and treatment information of the case 2 which could not be evaluated well by the model 1 and the model 2 were confirmed, the hypnotic (susvorexant) which is an orexin receptor blocker was used in this case. Unlike conventional benzodiazepine receptor agonists that increase light sleep, orexin receptor blockers increase deep sleep and REM sleep, so the evaluation formula (model) that replaces the appearance rate of light sleep with the appearance rate of deep sleep. 7) was prepared. Table 8 shows the sleep variables, coefficients, and constant terms in equation (I) of model 7.

When the sleep information of case 2 was input to model 7 and the value of probability (p) was calculated, the first value was 0.707, the second value was 0.505, and the clinical severity and sleep evaluation were closer. became.

From the above results, it was confirmed that by using logistic regression analysis, the symptoms of mentally ill patients and the response of treatment were evaluated from sleep information, and it could be used to support treatment.

Example 5: Distinction between healthy subjects and mentally handicapped patients by the method of the present invention A machine learning model capable of accurately distinguishing between inpatients and healthy subjects from sleep parameters of inpatients (93 nights) and healthy subjects (88 nights) The following studies were conducted to construct it.

(1) Examination of 7 items of sleep variables Since bedtime and wake-up time are fixed in the inpatient ward, the total bedtime is considered to be a parameter that depends on the hospitalization environment. Therefore, we selected the following seven sleep variables, excluding sleep parameters that are considered to be affected by the hospitalization environment such as total bedtime.
・ Ratio: Sleep efficiency, REM sleep appearance rate, light sleep appearance rate, deep sleep appearance rate, arousal response index ・ Latency: REM sleep latency, sleep onset latency

(2) Examination of 19 items of sleep variables In addition to the above sleep variables, the frequency of transitions in the sleep stage was analyzed. The frequency of transitions between sleep stages was calculated by dividing the number of transitions between sleep stages by the total bedtime, and the following 12 types were used.
・ Transition frequency: Awakening → Rem sleep, Awakening → Light sleep, Awakening → Deep sleep, Rem sleep → Awakening, Rem sleep → Light sleep, Rem sleep → Deep sleep, Light sleep → Awakening, Light sleep → Rem sleep, Light sleep → Deep sleep, deep sleep → awakening, deep sleep → rem sleep, deep sleep → light sleep

For each of the above cases, the data of inpatients and healthy subjects were divided into 5 groups, 4 groups were used as a training set, and the remaining 1 group was used as a test set, and cross-validation was performed using XGBoost. AUC (Area Under the Curve) was used as an evaluation index.

As a result of the cross-validation test data, the following results were obtained.
(1) Examination of 7 items of sleep variables: mean AUC: 0.827 (standard deviation: 0.084) (mean AUC of training data: 0.950, standard deviation: 0.010)
(2) Examination of 19 sleep parameters: mean AUC: 0.897 (standard deviation: 0.060) (mean AUC of training data: 0.995, standard deviation: 0.003)
In (2), a significant improvement in accuracy was observed, and it was confirmed that the transition of the sleep stage is an effective index for determining patients with mental disorders.

From the above results, it was confirmed that by using a machine learning model using multiple sleep parameters, the symptoms of mentally ill patients and the response of treatment were evaluated from sleep information, and it could be used to support treatment.

Example 6: Distinction between healthy subjects and mentally handicapped patients in the time series information of sleep stage transitions by the method of the present invention From the time series information of sleep stage transitions of inpatients (109 nights) and healthy subjects (105 nights) In order to build a model that can accurately distinguish between inpatients and healthy subjects, the following studies were conducted.

Cross-validation was performed by dividing the data of inpatients and healthy subjects into 5 groups, learning the neural network with 4 groups as a training set, and evaluating the learned neural network with the remaining 1 group as a test set. AUC (Area Under the Curve) was used as an evaluation index. A Transformer was used as the encoder and a fully connected neural network was used as the classifier.

The average AUC of the cross-validation test data was 0.935 (standard deviation: 0.030) (average AUC of training data: 0.943, standard deviation: 0.018), which exceeded 0.9, which is generally considered to be highly accurate. By treating it as time-series data using a neural network, it was confirmed that it can be used to support treatment by evaluating the symptoms of mentally handicapped patients from sleep information.

Example 7: Comparison of EEG from the forehead and the skin on the temporal bone (posterior left and right ears (mastoid process)) Using ZA-X (Pro Assist), the EEG was applied to the back of the left ear and the right of the forehead (right of the forehead). Fp2), EEG was acquired using the posterior left ear-posterior right ear for EMG, and the waveforms were compared.

As shown in FIG. 15, the waveforms of EEG from the skin (posterior ear (mastoid process)) on the forehead and temporal bone are different in all of arousal (closed eyes), light sleep, deep sleep, and REM sleep. It was confirmed that.

A sleep profiler (Advanced Brain Monitoring Co., Ltd.), which has a relatively large device attached to the forehead, was attached to 22 patients with mental disorders who had given their consent in advance, and sleep was measured. Of the 22 cases, 13 cases (59%) could not be measured, and the causes were non-wearing (4 cases), refusal (1 case), and electrode detachment (8 cases). Of the 9 cases that could be measured, 2 cases could not be evaluated (recording failure (1 case) and sleeplessness due to oppressive feeling (1 case)), and only 7 cases (32%) could be evaluated. there were.

On the other hand, when the same test was conducted on the Z machine (General Sleep) that attaches electrodes to the skin on the temporal bone (posterior left and right ears (mastoid process)), 11 cases / 182 cases (6) could not be measured. %), The cause was electrode dropout in all cases. From this, it was confirmed that the method of acquiring EEG from the skin on the temporal bone (posterior left and right subears (mastoid process)) is more tolerable for patients with mental disorders than the conventional method. rice field.

According to the present invention, the sleep of a mentally handicapped patient can be evaluated easily and with high accuracy. This makes it possible to evaluate the sleep profile of mentally ill patients and the response of treatment, and to provide appropriate treatment.

All publications, patents and patent applications cited in this specification shall be incorporated herein by reference as is.

1 EEG signal processor, 2 display, 3 input operation buttons, 4 electrodes, 4a + electrodes, 4b-electrodes, 4c com electrodes, 5 visual warning sensors, 6 auditory warning sensors,
11 Amplifier, 12 Digitizer, 13 Single Board Computer, 14 LCD Display, 15 Speakers, 16 Main Power Supply, 17 Power Supply for Amplifier
21 processors, 22 storage devices, 23 communication circuits, 24 input devices, 25 displays, 26 buses

Claims

A method of objectively evaluating a subject's sleep, in which the subject's electroencephalogram signal is processed using a device that electrically processes bioelectric activity acquired from electrodes attached to the skin on the temporal bone. An evaluation method including a step of acquiring sleep information after determining a sleep stage and a step of analyzing the sleep information to evaluate the sleep of a subject.
Sleep information includes sleep efficiency, REM sleep time, REM sleep appearance rate, REM sleep latency, light sleep time, deep sleep time, light sleep time appearance rate, deep sleep time appearance rate, sleep onset latency, sleep stage The evaluation method according to claim 1, which is selected from the group consisting of the number or frequency of transitions of the above and the time series of transitions of the sleep stage.
The method according to claim 1 or 2, wherein the subject is a mentally ill patient.
The method according to any one of claims 1 to 3, which comprises a step of comparing sleep information at two or more different time points of the subject.
The method according to claim 4, wherein two or more different time points include before and after treatment, before and after changing treatment, or during remission and recurrence / relapse.
The method according to any one of claims 1 to 5, wherein sleep information is schematized into a sleep progress chart for analysis or comparison.
The process of inputting the subject's sleep information into a machine learning model determined by the sleep information of the healthy person and the mentally handicapped patient acquired in advance and analyzing whether the subject's sleep is closer to the healthy person or the mentally handicapped patient. The method according to any one of claims 1 to 6, which includes.
Claim 7 includes a step of comparing a probability (p) calculated using a machine learning model with a predetermined threshold value, and if the probability (p) is smaller than the threshold value, it is evaluated that the subject's sleep is close to that of a healthy person. The method described in.
The method according to claim 8, wherein the machine learning model is a logistic regression model, and the probability (p) is represented by the following equation.
The method according to claim 8, wherein the machine learning model is a logistic regression model, and the probability (p) is represented by the following equation.
The method according to any one of claims 8 to 10, wherein the threshold value is 0.3 to 0.5, preferably 0.3 to 0.4.
The method according to claim 7, wherein the machine learning model is XGBoost.
The method according to claim 7, wherein the machine learning model is a neural network, preferably deep learning.
A method for supporting the treatment of a psychiatric patient, the step of evaluating the sleep of the patient according to the method according to any one of claims 1 to 13, and the symptom of the patient's psychiatric disorder based on the evaluation result. Or the method comprising the step of evaluating the degree thereof.
A method for supporting the treatment of a mentally ill patient, the step of evaluating the sleep of the patient according to the method according to any one of claims 1 to 13, and the step of selecting a preferable treatment based on the evaluation result. The method described above.
The method of claim 15, wherein an orexin receptor antagonist is presented as a preferred treatment option when sleep onset latency is 30 minutes or more and bedtime or less, or sleep efficiency is less than 75%.
The method according to any one of claims 14 to 16, wherein the step of evaluating the symptom or the degree of mental disorder of the patient or the step of selecting a preferable treatment is carried out using machine learning.
It is a treatment support system for mentally handicapped patients, including an electroencephalogram signal processing device that electrically processes an electroencephalogram signal and an information analysis device.
The electroencephalogram signal processing device is a device that electrically processes an electroencephalogram signal acquired from an electrode attached to the skin on the temporal bone of a patient.
The information analysis device is
A storage unit that stores sleep information of the patient processed by a device that electrically processes an electroencephalogram signal, and a storage unit.
An analysis unit that analyzes and evaluates sleep information stored in the storage unit,
Equipped with an output unit that outputs analysis / evaluation results
The analysis unit evaluates the sleep of the patient according to the method according to any one of claims 1 to 13, or the patient according to the method according to any one of claims 14 to 17. The treatment support system for assessing the symptoms of mental illness or its degree, or selecting a preferred treatment.
An information analysis device to support the treatment of mentally handicapped patients.
A storage unit that stores sleep information of the patient processed by a device that electrically processes an electroencephalogram signal, and a storage unit.
An analysis unit that analyzes and evaluates sleep information stored in the storage unit,
Equipped with an output unit that outputs analysis / evaluation results
The analysis unit evaluates sleep according to the method according to any one of claims 1 to 13, or the mentality of the patient according to the method according to any one of claims 14 to 17. The information analyzer for assessing the symptoms of a disorder or its degree, or selecting a preferred treatment.
A program to support the treatment of mentally ill patients
Processing that acquires sleep information of the patient processed by a device that electrically processes an electroencephalogram signal, stores the sleep information, evaluates / analyzes the stored sleep information, and outputs the evaluation / analysis result. To let the computer do
The evaluation / analysis evaluates sleep according to the method according to any one of claims 1 to 13, or mental disorder of a patient according to the method according to any one of claims 14 to 17. The program for assessing the symptoms or extent of the disease, or selecting the preferred treatment.
A way to help treat patients with mental illness
In the step of processing the brain wave signal of the patient to acquire sleep information using a device that electrically processes the brain wave signal, the step of analyzing the sleep information to evaluate the sleep of the patient, and the evaluation result. Including the step of selecting the preferred treatment based on
The electroencephalogram signal was obtained from an electrode attached to the skin on the patient's temporal bone.
The sleep information includes (i) sleep efficiency, (ii) REM sleep time or rate thereof or REM sleep latency, and (iii) light sleep and / or deep sleep time or rate thereof.
The method described above, wherein an orexin receptor antagonist is presented as the preferred treatment option when sleep onset latency is 30 minutes or more and bedtime or less, or sleep efficiency is less than 75%.