WO2024251704A1

WO2024251704A1 - Therapy for the treatment of nervous system disorders using machine learning and a report to patients

Info

Publication number: WO2024251704A1
Application number: PCT/EP2024/065270
Authority: WO
Inventors: Sebastian Schulze-Luckow; Andrew B. Kibler; Koeun LIM
Original assignee: Biotronik SE and Co KG
Current assignee: Biotronik SE and Co KG
Priority date: 2023-06-07
Filing date: 2024-06-04
Publication date: 2024-12-12
Anticipated expiration: 2025-12-07

Abstract

The present invention relates to a medical device (11) comprising: a storage medium (10); and processing circuitry (13) operably coupled to the storage medium (P) and configured to: sense electroencephalogram (EEG) data of a patient (P); perform feature-based delineation of the sensed EEG data to obtain tremor features present in the EEG data and indicative of an episode of tremor in the patient; determine, based on the feature-based delineation, that the tremor features satisfy threshold criteria for application of a machine learning model for verifying that the episode of tremor has occurred in the patient (P); in response to determining that the tremor features satisfy the threshold criteria for application of the machine learning model, apply the machine learning model, trained using EEG data for a plurality of patients (P), to the sensed EEG data to verify, based on the machine learning model, that the episode of tremor has occurred in the patient (P); and in response to verifying, by the machine learning model, that the episode of tremor has occurred in the patient (P): generate a report comprising an indication that the episode of tremor has occurred in the patient (P) and one or more of the tremor features that coincide with the episode of tremor; and output, for display, the report comprising the indication that the episode of tremor has occurred in the patient (P) and the one or more of the tremor features that coincide with the episode of tremor.

Description

THERAPY FOR THE TREATMENT OF NERVOUS SYSTEM DISORDERS USING MACHINE LEARNING AND A REPORT TO PATIENTS

The present invention relates to a medical device, a computing system and a method, particularly for detecting tremors in patients using machine learning and feature delineation.

Typical symptoms of diseases like Parkinson’s are tremors. A tremor is an involuntary, often rhythmic, muscle contraction and relaxation involving back and forth movements of one or more body parts. It is one of the most common involuntary movements, and frequently occurs in the hands of a patient, but it can also affect other body parts such as arms, eyes, face, head, vocal folds, trunk, and legs.

Known solutions for diagnosing/treating tremors require visiting a physician for clinical presentation, magnetic resonance therapy of the brain after medical diagnosis, and measurement of muscle activity by means of electrodes in the clinic.

However, all of the above measures typically involve visiting a physician. The typical population of patients is regularly older than 65 years, the repeated visit of the doctor is necessary for the diagnosis of this population. Thus, particularly, only a temporary measurement (time window) is provided, not a permanent monitoring, so that an insufficient diagnosis is a possible outcome. Furthermore, the diagnosis depends on the experience of the physician, wherein such diagnosis usually confirms that a tremor is either present or not, and may describe the quality of the tremor as severe or mild - but will regularly not be given as objectively measured value. Therefore, existing solutions are often partially unnecessary, expensive, time consuming, i.e. they tie up personnel, not comparable to the advance of the state of knowledge, and possibly prone to error as the physician evaluates. Therefore, based on the above, the problem to be solved by the present invention is to provide a medical device, a computing system and a method to improve detection of involuntary body movements such as tremors.

This problem is solved by a medical device having the features of claim 1, by a computing system having the features of claim 13 as well as by a method having the features of claim 14. Preferred embodiments of these aspects of the present invention are stated in the corresponding dependent claims and are described below.

According to claim 1, a medical device is disclosed, comprising: a storage medium; and processing circuitry operably coupled to the storage medium and configured to: sense electroencephalogram (EEG) data of a patient (e.g. in form of one or more EEG waveforms); perform feature-based delineation of the sensed EEG data to obtain tremor features present in the EEG data and indicative of an episode of tremor in the patient; determine, based on the feature-based delineation, that the tremor features satisfy threshold criteria for application of a machine learning model for verifying that the episode of tremor has occurred in the patient; in response to determining that the tremor features satisfy the threshold criteria for application of the machine learning model, apply the machine learning model, trained using EEG data for a plurality of patients, to the sensed EEG data to verify, based on the machine learning model, that the episode of tremor has occurred in the patient; and in response to verifying, by the machine learning model, that the episode of tremor has occurred in the patient: generate a report comprising an indication that the episode of tremor has occurred in the patient and one or more of the tremor features that coincide with the episode of tremor; and output, for display, the report comprising the indication that the episode of tremor has occurred in the patient and the one or more of the tremor features that coincide with the episode of tremor. According to an embodiment of the present invention, features of an EEG can be extracted and used to detect tremor by a number of means. For instance, a 2-step signal processing can be performed, whereby features are extracted from the EEG, comprising at least one of: the frequency band power, Fourier transform spectrum and power/energy, convolutional filters to locate signal patterns. At least one of those features are evaluated using a simple mathematical trend assessment such as for example: crossing an absolute threshold or thresholds, significant changes from baseline, or dynamic over units of time. At least one of the features listed above may also be fed into a regression feature analytics tool which assesses multiple factors simultaneously and is trained on exemplary data with and without tremor.

In a further aspect of the present invention, instead of EEG data, spinal cord compound action potentials or photo plethysmography signals (PPG) can be sensed. PPG can detect oxygenation as well as correlates of blood pressure, and so additional diagnostics are available. Instead of tremors, the data can be evaluated in view of one of: seizures, migraine, pain episodes. This alternative use of spinal cord compound action potentials or photo plethysmography signals (PPG) also applies to the other aspects of the present invention, i.e., to the computing system and the method as described further down below.

According to a preferred embodiment of the medical device according to the present invention, the medical device is an implantable medical device, particularly a brain pacemaker comprising electrodes for sensing EEG signals and particularly for applying electrical stimulation to the brain of a patient.

According to a preferred embodiment of the medical device according to the present invention, the episode of tremor in the patient is an episode of tremor of a first classification in the patient, wherein to apply the machine learning model to the sensed EEG data to verify that the episode of tremor has occurred in the patient, the processing circuitry is configured to apply the machine learning model to the sensed EEG data to verify that the episode of tremor of the first classification has occurred in the patient, wherein the processing circuitry is further configured to apply the machine learning model to the sensed EEG data to determine that an episode of tremor of a second classification has occurred in the patient in response to determining, based on the feature- based delineation, that the episode of tremor of the first classification has occurred in the patient, and wherein to generate the report comprising the indication that the episode of tremor has occurred in the patient and one or more of the tremor features that coincide with the episode of tremor, the processing circuitry is configured to generate a report comprising an indication that the episode of tremor of the first classification has occurred in the patient, an indication that the episode of tremor of the second classification has occurred in the patient, and the one or more of the tremor features that coincide with the episode of tremor of the first classification.

Furthermore, according to a preferred embodiment of the medical device, in order to perform feature-based delineation of the EEG data to obtain the tremor features present in the EEG data, the processing circuitry is configured to perform at least one of refractory processing, noise processing, delineation of the EEG data to obtain tremor features present in the EEG data.

According to an embodiment, refractory processing is understood as analyzing the physiological response immediately following a period of stimulation. This includes assessment of action potentials generated by stimulation within milliseconds, to assessment of slow EEG response changes which can take place over time periods from tens of milliseconds to minutes following stimulation. The same features described above would be assessed from EEG for this purpose.

Further, according to a preferred embodiment of the medical device, to apply the machine learning model to verify that the episode of tremor has occurred in the patient, the processing circuitry is configured to apply the machine learning model to verify that an episode of tremor has occurred in the patient.

For instance, typical tremor types are resting tremor, postural tremor, kinetic tremor, taskspecific tremor, and intention tremor. Resting tremor occurs when a body part is at complete rest against gravity. Tremor amplitude decreases with voluntary activity. Examples of resting tremor are / occur as: Drug-induced parkinsonism (neuroleptics, metaclopromide, and phenothiazines), long-standing essential tremor, Parkinson's disease, Parkinson's plus syndromes (rest tremor is less common), or Wilson's disease.

Postural tremor occurs during maintenance of a position against gravity and increases with action. Action or kinetic tremor occurs during voluntary movement. Examples of postural and action tremors are /occur due to: Essential tremor (primarily postural), metabolic disorders (thyrotoxicosis, pheochromocytoma, hypoglycemia), drug-induced parkinsonism (lithium, amiodarone, P-adrenergic agonists), toxins (alcohol withdrawal, heavy metals), neuropathic tremor (neuropathy). Task-specific tremor emerges during a specific activity. An example of this type is primary writing tremor. Intention (or terminal) tremor manifests as a marked increase in tremor amplitude during a terminal portion of targeted movement. Examples of intention tremor include cerebellar tremor and multiple sclerosis tremor.

Furthermore, according to a preferred embodiment of the medical device, the machine learning model trained using EEG data for the plurality of patients comprises a machine learning model trained using a plurality of electroencephalogram (EEG) waveforms, each EEG waveform labeled with one or more episodes of tremor of one or more classifications in a patient of the plurality of patients.

According to yet another embodiment of the medical device, to determine that the tremor features satisfy the threshold criteria, the processing circuitry is configured to determine that at least one of a physiological parameter of the patient or a parameter of the medical device satisfies the threshold criteria.

Furthermore, in a preferred embodiment of the medical device, to apply the machine learning model to the sensed EEG data, the processing circuitry is configured to apply the machine learning model to the sensed EEG data in response to determining that the tremor features satisfy the threshold criteria and determining that a noise of at least one of the tremor features is less than a predetermined threshold.

According to a further embodiment of the medical device, to apply the machine learning model to the sensed EEG data, the processing circuitry is configured to apply the machine learning model to the sensed EEG data in response to determining that the tremor features satisfy the threshold criteria and determining that the patient is in a first posture state of a plurality of posture states.

As an example, posture states may include at least one of standing, sitting, laying down (reclining), bending over; and may include an additional determination of behavior states (walking, standing still, light activity). Additional inputs to the machine learning model to determine posture states and behavior states include accelerometer, gyroscope, or inputs from wearable sensors. The posture state information can be used to derive information regarding which postures the patient tends to be in while experiencing tremor. The posture information can affect further diagnostics and therapy decisions.

Furthermore, in a preferred embodiment of the medical device, the EEG data of the patient comprises an electroencephalogram (EEG) of the patient, and wherein to generate the report comprising the indication that the episode of tremor has occurred in the patient and the one or more of the tremor features that coincide with the episode of tremor, the processing circuitry is configured to: identify a subsection of the EEG of the patient, wherein the subsection comprises EEG data for a first time period prior to the episode of tremor, a second time period during the episode of tremor, and a third time period after the episode of tremor, and wherein a length of time of the EEG of the patient is greater than the first, second, and third time periods; identify one or more of the tremor features that coincide with the first, second, and third time periods; and include, in the report, the subsection of the EEG and the one or more of the tremor features that coincide with the first, second, and third time periods.

According to an embodiment, the first time period of EEG is established to be a baseline time period or pre-tremor signaling period with signals that indicate a tremor is soon to occur. This period lasts minutes to seconds. The third time period is the post-tremor period which may contain a refractory period following the tremor, an interim period of EEG where features do not match tremor features but also have not returned to baseline characteristics, where this third time period may last tens of seconds to an hour. Further, according to a preferred embodiment of the medical device, to determine, based on the feature-based delineation, that the tremor features satisfy threshold criteria for application of the machine learning model for verifying that the episode of tremor has occurred in the patient, the processing circuitry is configured to determine, based on the feature-based delineation, that the tremor features are indicative that an episode of tremor of a first classification has occurred in the patient, and wherein to apply the machine learning model to the sensed EEG data to verify that the episode of tremor has occurred in the patient, the processing circuitry is configured to apply the machine learning model to the sensed EEG data to verify the determination based on the feature-based delineation that the tremor features are indicative of the episode of tremor of the first classification.

According to yet another preferred embodiment of the medical device, the processing circuitry is further configured to process the sensed EEG data to generate filtered EEG data, wherein to apply the machine learning model to the sensed EEG data to verify that the episode of tremor has occurred in the patient, the processing circuitry is further configured to apply the machine learning model to the filtered EEG data to verify that the episode of tremor has occurred in the patient.

Further, according to a preferred embodiment of the medical device, applying the machine learning model to the sensed EEG data to verify that the episode of tremor has occurred in the patient comprises at least one of a first determination, based on the machine learning model, that the episode of tremor has not occurred in the patient or a second determination, based on the machine learning model, that an episode of tremor of a different type has occurred in the patient, the method further comprising: in response to the at least one of the first determination and the second determination, updating, by the medical device, a counter of incorrectly detected episodes of tremor in the patient; and in response to determining that a value of the counter is greater than a predetermined threshold, switching from performing feature-based delineation of the sensed EEG data to obtain tremor features present in the EEG data and indicative of an episode of tremor in the patient to applying a second machine learning model, trained using EEG data for a plurality of patients, to the sensed EEG data to obtain, based on the machine learning model, tremor features present in the EEG data and indicative of an episode of tremor in the patient

According to yet another aspect of the present invention, a computing system is disclosed, the computing system comprising: means for sensing EEG data of a patient; means for performing feature-based delineation of the sensed EEG data to obtain tremor features present in the EEG data and indicative of an episode of tremor in the patient; means for determining, based on the feature-based delineation, that the tremor features satisfy threshold criteria for application of a machine learning model for verifying that the episode of tremor has occurred in the patient; means for applying, in response to determining that the tremor features satisfy the threshold criteria for application of the machine learning model, the machine learning model, trained using EEG data for a plurality of patients, to the sensed EEG data to verify, based on the machine learning model, that the episode of tremor has occurred in the patient; and means for, in response to verifying, by the machine learning model, that the episode of tremor has occurred in the patient, generating a report comprising an indication that the episode of tremor has occurred in the patient and one or more of the tremor features that coincide with the episode of tremor; and means for outputting, for display, the report comprising the indication that the episode of tremor has occurred in the patient and the one or more of the tremor features that coincide with the episode of tremor.

Furthermore, according to yet another aspect of the present invention, a method is disclosed, the method comprising: sensing, by a medical device comprising processing circuitry and a storage medium, electroencephalogram (EEG) data of a patient; performing, by the medical device, feature-based delineation of the sensed EEG data to obtain tremor features present in the EEG data and indicative of an episode of tremor in the patient; determining, by the medical device and based on the feature-based delineation, that the tremor features satisfy threshold criteria for application of a machine learning model for verifying that the episode of tremor has occurred in the patient; in response to determining that the tremor features satisfy the threshold criteria for application of the machine learning model, applying, by the medical device, the machine learning model, trained using EEG data for a plurality of patients, to the sensed EEG data to verify, based on the machine learning model, that the episode of tremor has occurred in the patient; and in response to verifying, by the machine learning model, that the episode of tremor has occurred in the patient: generating, by the medical device, a report comprising an indication that the episode of tremor has occurred in the patient and one or more of the tremor features that coincide with the episode of tremor; and outputting, by the medical device and for display, the report comprising the indication that the episode of tremor has occurred in the patient and the one or more of the tremor features that coincide with the episode of tremor.

Particularly, the method according to the present invention preferably uses a medical device or computing system according to the present invention. Furthermore, the method can be further characterized by the individual features and embodiments of the medical device or computer system as described herein, particularly by the respective subject-matter of the claims 2 to 12 that can be formulated as method steps in this regard.

In the following, embodiments of the aspects of the present invention as well as further features and advantages of the present invention are described with reference to the Figures, wherein

Fig. 1 an embodiment of a computing system and method according to the present invention.

Fig. 1 shows an embodiment of a medical device 11 / computer system 1 and method according to the present invention allowing to provide energy-efficient tremor detection based on feature delineation and machine learning, and to report occurrences of tremor in a patient.

Particularly, feature delineation algorithms may use EEG data sensed from a patient to perform, e.g., tremor detection. Such feature delineation algorithms may be optimized for real-time, embedded, and low-power applications, such as for use by an implantable medical device as depicted in Fig. 1. Furthermore, particularly, machine learning methods for tremor detection, such as deep-learning and artificial intelligence (Al), can detect tremor or exclude episodes where tremor is absent with a high degree of accuracy without the complexity of algorithms such as feature delineation. However, machine learning models are computationally involved and thus may affect battery longevity of a medical device. In this regard, the present invention allows to use machine learning when threshold criteria are satisfied so that tremor detection benefits from machine learning while allowing at the same time to save energy in the context of an implantable medical device.

As indicated in Fig. 1, a medical device 11 according to an embodiment of the present invention comprises a storage medium 10 and processing circuitry 13 operably coupled to the storage medium 10 and configured to: sense electroencephalogram (EEG) data (e.g. EEG waveforms) of a patient P, and perform feature-based delineation of the sensed EEG data to obtain tremor features present in the EEG data and indicative of an episode of tremor in the patient, and to determine, based on the feature-based delineation, that the tremor features satisfy threshold criteria for application of a machine learning model for verifying that the episode of tremor has occurred in the patient. Furthermore, in response to determining that the tremor features satisfy the threshold criteria for application of the machine learning model, the medical device 11 is configured to apply the machine learning model, trained using EEG data for a plurality of patients, to the sensed EEG data to verify, based on the machine learning model, that the episode of tremor has occurred in the patient. Furthermore, in response to verifying, by the machine learning model, that the episode of tremor has occurred in the patient P, the medical device 11 is configured to generate a report comprising an indication that the episode of tremor has occurred in the patient P and one or more of the tremor features that coincide with the episode of tremor; and to output (e.g. to a user), for display, the report comprising the indication that the episode of tremor has occurred in the patient P and the one or more of the tremor features that coincide with the episode of tremor.

Preferably, the medical device 11 is an implantable medical device 11 such as a brain pacemaker and comprises at least one electrode 110 to sense EEG data / waveforms to be analyzed with respect to the occurrence of a tremor of patient P. Alternatively, sensing of EEG data and/or detection of tremor as described herein may be performed by an external device 17 in addition to, or instead of implantable medical device 11. Such an external medical 17 device may be positioned externally to patient 4 (e.g., positioned on the skin of patient 4) and may carry out any or all of the functions described herein with respect to the implantable medical device 11.

Particularly, the external device 17 can be a computing device configured for use in settings such as a home, clinic, or hospital, and can be configured to communicate with the implantable medical device 11 via wireless radio communication 14. Particularly, the external device 12 can be or comprise one of: a programmer, an external monitor, a mobile device, such as a mobile phone, a laptop, a tablet computer, a personal digital assistant (PDA) etc. Furthermore, external device 12 can also be a wearable device, particularly one of: a smart watch, a smart necklace, a smart anklet, smart glasses. Particularly, each of these smart devices is characterized in that it comprises a user interface for displaying information to a user (e.g. patient) and receiving input from the user, and a processor configured to perform functions of the smart device.

Particularly, a user, e.g., a physician, technician, surgeon, electro-physiologist, or other clinician, can operate external device 17 to e.g. retrieve physiological or diagnostic information from the implantable medical device 11. In this regard, external device 17 can also be configured to program the implantable medical device 11 and can further serve as an access point to facilitate communication with the implantable medical device 11.

Particularly, the external device 17 can comprise a processor 130 configured to analyze EEG data and/or other sensed signals transmitted from the implantable medical device 11 to the external device 17. External device 17 can further comprise a storage medium 100 for storing EEG data and the like.

Furthermore, particularly, processing circuitry 13 of medical device 11 can be configured to transmit patient data, including EEG data, for patient P to another (e.g. remote) device, particularly to external device 17. Particularly, processing circuitry 13 of medical device 11 is configured to transmit a determination that patient P is undergoing an episode of tremor. Particularly, further, external device 17 can be configured to interrogate the implantable medical device 11 to retrieve data, including device operational data as well as physiological data accumulated in the storage medium 10 of implantable medical device 11. Such interrogation can be conducted automatically according to a schedule and/or can be conducted in response to a remote or local user command. Particularly, wireless communication 14 between the implantable medical device 11 and the external device 17 can be based on radio frequency (RF) communication, which may be an RF link established via Bluetooth, WiFi, or medical implant communication service (MICS). In some examples, external device 17 can comprise a user interface configured to allow patient P, a clinician, or another user to remotely interact with the implantable medical device 11.

Preferably, medical device 11 is configured to perform tremor detection, verification, and reporting. Particularly, implantable medical device 11 implements a machine learning model, such as neural network, a deep learning system, or other type of predictive analytics system.

Particularly, implantable medical device 11 or external device 17, is configured to conduct feature delineation of EEG data such as an EEG waveform to make a preliminary detection of tremor in patient P. Particularly, the implantable medical device 11 can be configured to apply a machine learning model to EEG data of patient P to verify that feature delineation of the EEG data has correctly detected an episode of tremor. Furthermore, particularly, the medical device 11 can be configured to apply a machine learning model to EEG data of patient P to verify that feature delineation of the EEG data has correctly classified an episode of tremor as a particular type of tremor. In the following, tremor detection and processing (e.g. issuing reports) is described as being performed by the implantable medical device 11. However, these functions can also be performed by other devices, such as external device 17, or by a plurality of devices (e.g., implantable medical device 11 and external device 17) operating in conjunction with one another.

Particularly, the implantable medical device 11 is configured to sense EEG data of patient P and to perform feature-based delineation of the EEG data to obtain tremor features indicative of an episode of tremor in patient P. particularly, the implantable medical device 11 is configured to determine whether the tremor features satisfy threshold criteria for application of a machine learning model for verifying the feature-based delineation of the EEG data. Furthermore, particularly, the implantable medical device 11 is further configured to determined that a noise of at least one of the tremor features is less than a predetermined threshold. Particularly, implantable medical device 11 is configured to determine that the patient is in a first posture state of a plurality of posture states or a first activity state of a plurality of activity states. In response to determining that the tremor features satisfy the threshold criteria, implantable medical device 11 is configured to apply the machine learning model to the sensed EEG data to, e.g., verify that the episode of tremor has occurred in patient P or to detect one or more additional types of tremor that have occurred in patient P. Particularly, the implantable medical device 11 can be configured to classify episodes of tremor by comparing tremor features coincident with the episode of tremor or with tremor features of a tremor database maintained by the implantable medical device 11. Furthermore, particularly, the implantable medical device 11 can be configured to compare first tremor features of the EEG data to tremor features defined by an entry of the tremor database. Particularly, in response to determining that the first tremor features of the EEG data are similar to tremor features defined by an entry of the tremor database, the implantable medical device 11 determines that the first tremor features indicate that an episode of tremor has occurred in patient P that is a classification defined by the matching entry within the tremor database.

Furthermore, particularly, in response to determining that the first tremor features of the EEG data are not similar to the tremor features defined by any entries of the tremor database, the implantable medical device 11 is adapted to apply a machine learning model to determine a classification of an episode of tremor as evidenced by the first tremor features. Further, preferably, the implantable medical device 11 is configured to store the determined tremor classification and tremor features as a new entry in the tremor database to update the tremor database. In case the implantable medical device 11 is subsequently detecting, via feature delineation, second tremor features that are similar to tremor features of an entry of the tremor database, implantable medical device 11 is configured to determine that the second tremor features are indicative of an episode of tremor of the same classification as the episode of tremor defined in the entry of the tremor database and including tremor features that match the second tremor features.

The present invention provides an improved tremor detection and classification by an implantable medical device. Particularly, analysis of EEG data is based on machine learning models only in case the EEG signals that have been identified by feature delineation are likely presenting an episode of tremor in the patient P. Furthermore, by using low-power feature delineation to limit the use of computationally-complex, power-intensive machine learning models to only the most relevant EEG data, power usage is advantageously decreased.

Furthermore, storage medium 10 of the implantable medical device 11 can store computer- readable instructions that, when executed by processing circuitry 13, cause the implantable medical device 11 and processing circuitry 13 to perform functions of the implantable medical device 11. Storage medium 10 can include any volatile, non-volatile, magnetic, optical, or electrical media, such as a random-access memory (RAM), read-only memory (ROM), nonvolatile RAM (NVRAM), electrically-erasable programmable ROM (EEPROM), flash memory, or any other digital media. Furthermore, particularly, processing circuitry 13 can be formed as or comprise fixed function circuitry and/or programmable processing circuitry. Processing circuitry 13 can further be formed as or comprise one or more of: a microprocessor, a controller, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA). It can further include equivalent discrete or analog logic circuitry. Furthermore, particularly, the processing circuitry 13 can comprise multiple components, such as any combination of one or more microprocessors, one or more controllers, one or more DSPs, one or more ASICs, or one or more FPGAs, as well as other discrete or integrated logic circuitry. The functions attributed to processing circuitry 13 herein may be embodied as software, firmware, hardware or any combination thereof.

Furthermore, particularly, processing circuitry 13 can be configured to transmit sensed EEG data of patient P to external device 17 as shown in Fig. 1 via radio communication 14. Particularly, implantable medical device 11 can be configured to transmit EEG data to external device 17 for data processing or review by a clinician/physician. Particularly, implantable medical device 11 can be configured to transmit one or more segments of the EEG data in response to detecting, via feature delineation, an episode of tremor. Furthermore, particularly, implantable medical device 11 can transmit one or more segments of the EEG data in response to instructions from external device 17. To this end, patient P can operate external device 17 when experiencing one or more tremors to trigger implantable medical device 11 to transmit the EEG data to the external device 17 for analysis e.g. by a monitoring center or clinician/physician.

Furthermore, particularly, a user can retrieve data from the implantable medical device 11 by using external device 17, or by using another local or networked computing device configured to communicate with processing circuitry 13 via communication link 14. The user (e.g. clinician or physician) may also program parameters of implantable medical device 11 using external device 17 or another local or networked computing device. Particularly, the user can select one or more parameters defining how the implantable medical device 11 senses EEG data from patient P.

Furthermore, particularly, processing circuitry 13 can perform feature delineation to identify one or more features of EEG waveform of patient P to detect an episode of tremor in patient P. Particularly, processing circuitry 13 can be configured to identify one or more relative changes in the one or more identified features that are indicative of an episode of tremor in patient P. Furthermore, particularly, processing circuitry 13 can identify one or more interactions between multiple identified features that are indicative of an episode of tremor in patient P. Further, processing circuitry 13 can be configured to analyze patient data, particularly EEG data, that represents one or more values that are averaged over a short-term period of time to detect the episode of tremor.

Furthermore, particularly, processing circuitry 13 can be configured to apply feature delineation to classify the detected episode of tremor as an episode of tremor of a certain type. Further, particularly, processing circuitry 13 is configured to perform feature delineation of a reduced complexity so as to conserve power in the implantable medical device 11. This may enable the implantable medical device 11 to perform initial or preliminary detection of a tremor.

Additionally, processing circuitry 13, can be configured to apply a machine learning model to the EEG data to verify or classify the detection of episodes of tremor based on feature delineation. While processing circuit 13 can perform a more comprehensive and detailed analysis of the EEG data based on the machine learning model so as to more accurately detect tremor compared to feature delineation, machine learning can require more computational resources and power compared to feature delineation. However, by employing the machine learning model to verify or classify the detection of episodes of tremor by means of feature delineation, implantable medical device 11 can utilize the high accuracy offered by machine learning systems, while minimizing the power consumption or battery longevity of the implantable medical device 11. Particularly, processing circuitry 13 can transmit, via communication link 14, one or more of the EEG data, the one or more tremor features present in the EEG data, an indication of an episode of tremor verified based on the machine learning model, or an indication of a classification of the detected episode of tremor as determined based on the machine learning model, to external device 17.

Furthermore, preferably, the machine learning model implemented by the implantable medical device 11 is trained with training data that comprises EEG data for a plurality of patients labeled with descriptive metadata. Particularly, during a training phase, processing circuitry 13 processes a plurality of EEG waveforms. Typically, the plurality of EEG waveforms originates from a plurality of different patients. Each EEG waveform is labeled with one or more episodes of tremor of one or more types. For example, a training EEG waveform can include a plurality of segments, each segment labeled with a descriptor that specifies an absence of tremor or a presence of a tremor of a particular classification. Particularly, a clinician/physician can label the presence of tremor in each EEG waveform by hand. Particularly, the presence of tremor in each EEG waveform is labeled according to a classification by a feature delineation algorithm. Furthermore, the implantable medical device 11 can operate to convert the training data into vectors and tensors (e.g., multidimensional arrays) upon which the processing circuitry 13 can apply mathematical operations, such as linear algebraic, nonlinear, or alternative computation operations. Furthermore, implantable medical device 11 can use the training data to teach the machine learning model to weigh different features depicted in the EEG data. Particularly, implantable medical device 11 can use the EEG data to teach the machine learning model to apply different coefficients that represent one or more features in an EEG as having more or less importance with respect to an occurrence of a tremor of a particular classification. By processing numerous such EEG waveforms labeled with episodes of tremor, implantable medical device 11 can build and train a machine learning model to receive EEG data from a patient P, that processing circuitry 13 has not previously analyzed, and process such EEG data to detect the presence or absence of tremor of different classifications in the patient with a high degree of accuracy. Typically, the greater the amount of EEG data on which the machine learning model is trained, the higher the accuracy of the machine learning model in detecting or classifying tremor in new EEG data.

Furthermore, after the machine learning model has been trained, the implantable medical device 11 may receive patient data, such as EEG data, for a particular patient P. Furthermore, particularly, the processing circuitry 13 applies the trained machine learning model to the patient data to detect an episode of tremor in patient P. Further, implantable medical device 11 is configured to apply the trained machine learning model to the patient data to classify the episode of tremor in patient P as indicative of a particular type of tremor. Furthermore, particularly, implantable medical device 11 can output a preliminary determination that the episode of tremor is indicative of a particular type of tremor, as well as an estimate of certainty in the determination. The information can be output via display 170 of external device 17. In response to determining that the estimate of certainty in the determination is greater than a predetermined threshold (e.g., 50%, 75%, 90%, 95%, 99%), processing circuitry 13 can classify that the episode of tremor as the particular type of tremor. Particularly, processing circuitry 13 is configured to use the machine learning model to verify that feature delineation implemented in the implantable medical device 11 has correctly detected an episode of tremor or that said feature delineation has correctly classified an episode of tremor as being of a particular type.

Furthermore, implantable medical device 11 can process one or more tremor features of EEG data instead of the raw EEG data itself. The one or more tremor features may be obtained via feature delineation performed by implantable medical device 11, as described above. Here, implantable medical device 11 can train the machine learning model via a plurality of training tremor features labeled with episodes of tremor, instead of the plurality of EEG waveforms labeled with episodes of tremor as described above.

Furthermore, particularly, implantable medical device 11 can be configured to apply the machine learning model also to other types of data to determine that an episode of tremor has occurred in patient P. Particularly, implantable medical device 11 can apply the machine learning model to one or more characteristics of EEG data that are correlated to tremor in the patient, an activity level of implantable medical device 11, an input impedance of implanted medical device 11, or a battery level of implantable medical device 11.

Furthermore, processing circuitry 13 can generate, from the EEG data, an intermediate representation of the EEG data. Particularly, processing circuitry 13 can be configured to apply one or more signal processing, signal decomposition, wavelet decomposition, filtering, or noise reduction operations to the EEG data to generate the intermediate representation of the EEG data. Particularly, implantable medical device 11 can be configured to process such an intermediate representation of the EEG data to detect and classify an episode of tremor in patient P. Furthermore, implantable medical device 11 can be configured to train the machine learning model via a plurality of training intermediate representations labeled with episodes of tremor, instead of the plurality of raw EEG waveforms labeled with episodes of tremor as described above. The use of such intermediate representations of the EEG data may allow for the training and development of a lighter-weight, less computationally complex machine learning model by processing circuitry 13. Further, the use of such intermediate representations of the EEG data may require less iterations and fewer training data to build an accurate machine learning model, as opposed to the use of raw EEG data to train the machine learning model.

Particularly, the functions described herein can be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored as one or more instructions or code on a computer-readable medium and executed by a hardware-based processing circuitry. Computer-readable media can include non-transitory computer-readable media, which corresponds to a tangible medium such as data storage media (e.g., RAM, ROM, EEPROM, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer).

Furthermore, instructions may be executed by one or more processors, such as one or more digital signal processors (DSPs), general purpose microprocessors, application specific integrated circuits (ASICs), field programmable logic arrays (FPGAs), or other equivalent integrated or discrete logic circuitry. Accordingly, the notion processor or processing circuitry as used herein may refer to any of the foregoing structure or any other physical structure suitable for implementation of the described techniques. Also, the techniques could be fully implemented in one or more circuits or logic elements.

Claims

1. A medical device (11) comprising: a storage medium (10); and processing circuitry (13) operably coupled to the storage medium (P) and configured to: sense electroencephalogram (EEG) data of a patient (P); perform feature-based delineation of the sensed EEG data to obtain tremor features present in the EEG data and indicative of an episode of tremor in the patient; determine, based on the feature-based delineation, that the tremor features satisfy threshold criteria for application of a machine learning model for verifying that the episode of tremor has occurred in the patient (P); in response to determining that the tremor features satisfy the threshold criteria for application of the machine learning model, apply the machine learning model, trained using EEG data for a plurality of patients (P), to the sensed EEG data to verify, based on the machine learning model, that the episode of tremor has occurred in the patient (P); and in response to verifying, by the machine learning model, that the episode of tremor has occurred in the patient (P): generate a report comprising an indication that the episode of tremor has occurred in the patient (P) and one or more of the tremor features that coincide with the episode of tremor; and output, for display, the report comprising the indication that the episode of tremor has occurred in the patient (P) and the one or more of the tremor features that coincide with the episode of tremor.

2. The medical device of claim 1, wherein the episode of tremor in the patient (P) is an episode of tremor of a first classification in the patient (P), wherein to apply the machine learning model to the sensed EEG data to verify that the episode of tremor has occurred in the patient, the processing circuitry (13) is configured to apply the machine learning model to the sensed EEG data to verify that the episode of tremor of the first classification has occurred in the patient (P), wherein the processing circuitry (13) is further configured to apply the machine learning model to the sensed EEG data to determine that an episode of tremor of a second classification has occurred in the patient in response to determining, based on the feature- based delineation, that the episode of tremor of the first classification has occurred in the patient, and wherein to generate the report comprising the indication that the episode of tremor has occurred in the patient (P) and one or more of the tremor features that coincide with the episode of tremor, the processing circuitry (13) is configured to generate a report comprising an indication that the episode of tremor of the first classification has occurred in the patient (P), an indication that the episode of tremor of the second classification has occurred in the patient (P), and the one or more of the tremor features that coincide with the episode of tremor of the first classification.

3. The medical device according to one of the preceding claims, wherein to perform feature-based delineation of the EEG data to obtain the tremor features present in the EEG data, the processing circuitry is configured to perform at least one of refractory processing, noise processing, delineation of the EEG data to obtain tremor features present in the EEG data.

4. The medical device according to one of the preceding claims, wherein to apply the machine learning model to verify that the episode of tremor has occurred in the patient (P), the processing circuitry (13) is configured to apply the machine learning model to verify that an episode of tremor has occurred in the patient.

5. The medical device according to one of the preceding claims, wherein the machine learning model trained using EEG data for the plurality of patients (P) comprises a machine learning model trained using a plurality of electroencephalogram (EEG) waveforms, each EEG waveform labeled with one or more episodes of tremor of one or more classifications in a patient (P) of the plurality of patients (P).

6. The medical device according to one of the preceding claims, wherein to determine that the tremor features satisfy the threshold criteria, the processing circuitry (13) is configured to determine that at least one of a physiological parameter of the patient or a parameter of the medical device (11) satisfies the threshold criteria.

7. The medical device according to one of the preceding claims, wherein to apply the machine learning model to the sensed EEG data, the processing circuitry (13) is configured to apply the machine learning model to the sensed EEG data in response to determining that the tremor features satisfy the threshold criteria and determining that a noise of at least one of the tremor features is less than a predetermined threshold.

8. The medical device according to one of the preceding claims, wherein to apply the machine learning model to the sensed EEG data, the processing circuitry (13) is configured to apply the machine learning model to the sensed EEG data in response to determining that the tremor features satisfy the threshold criteria and determining that the patient (P) is in a first posture state of a plurality of posture states.

9. The medical device according to one of the preceding claims, wherein the EEG data of the patient (P) comprises an electroencephalogram (EEG) of the patient (P), and wherein to generate the report comprising the indication that the episode of tremor has occurred in the patient (P) and the one or more of the tremor features that coincide with the episode of tremor, the processing circuitry (13) is configured to: identify a subsection of the EEG of the patient (P), wherein the subsection comprises EEG data for a first time period prior to the episode of tremor, a second time period during the episode of tremor, and a third time period after the episode of tremor, and wherein a length of time of the EEG of the patient (P) is greater than the first, second, and third time periods; identify one or more of the tremor features that coincide with the first, second, and third time periods; and include, in the report, the subsection of the EEG and the one or more of the tremor features that coincide with the first, second, and third time periods.

10. The medical device according to one of the preceding claims, wherein to determine, based on the feature-based delineation, that the tremor features satisfy threshold criteria for application of the machine learning model for verifying that the episode of tremor has occurred in the patient (P), the processing circuitry (13) is configured to determine, based on the feature-based delineation, that the tremor features are indicative that an episode of tremor of a first classification has occurred in the patient (P), and wherein to apply the machine learning model to the sensed EEG data to verify that the episode of tremor has occurred in the patient (P), the processing circuitry (13) is configured to apply the machine learning model to the sensed EEG data to verify the determination based on the feature-based delineation that the tremor features are indicative of the episode of tremor of the first classification.

11. The medical device according to one of the preceding claims, wherein the processing circuitry (13) is further configured to process the sensed EEG data to generate filtered EEG data, wherein to apply the machine learning model to the sensed EEG data to verify that the episode of tremor has occurred in the patient (P), the processing circuitry (13) is further configured to apply the machine learning model to the filtered EEG data to verify that the episode of tremor has occurred in the patient (P).

12. The method of any of claims 1 through 11, wherein applying the machine learning model to the sensed EEG data to verify that the episode of tremor has occurred in the patient (P) comprises at least one of a first determination, based on the machine learning model, that the episode of tremor has not occurred in the patient (P) or a second determination, based on the machine learning model, that an episode of tremor of a different type has occurred in the patient (P), the method further comprising: in response to the at least one of the first determination and the second determination, updating, by the medical device (11), a counter of incorrectly detected episodes of tremor in the patient (P); and in response to determining that a value of the counter is greater than a predetermined threshold, switching from performing feature-based delineation of the sensed EEG data to obtain tremor features present in the EEG data and indicative of an episode of tremor in the patient (P) to applying a second machine learning model, trained using EEG data for a plurality of patients (P), to the sensed EEG data to obtain, based on the machine learning model, tremor features present in the EEG data and indicative of an episode of tremor in the patient (P).

13. A computing system (1) comprising: means for sensing EEG data of a patient (P); means for performing feature-based delineation of the sensed EEG data to obtain tremor features present in the EEG data and indicative of an episode of tremor in the patient (P); means for determining, based on the feature-based delineation, that the tremor features satisfy threshold criteria for application of a machine learning model for verifying that the episode of tremor has occurred in the patient (P); means for applying, in response to determining that the tremor features satisfy the threshold criteria for application of the machine learning model, the machine learning model, trained using EEG data for a plurality of patients (P), to the sensed EEG data to verify, based on the machine learning model, that the episode of tremor has occurred in the patient (P); and means for, in response to verifying, by the machine learning model, that the episode of tremor has occurred in the patient (P), generating a report comprising an indication that the episode of tremor has occurred in the patient (P) and one or more of the tremor features that coincide with the episode of tremor; and means for outputting, for display, the report comprising the indication that the episode of tremor has occurred in the patient (P) and the one or more of the tremor features that coincide with the episode of tremor.

14. A method comprising: sensing, by a medical device (11) comprising processing circuitry (13) and a storage medium (10), electroencephalogram (EEG) data of a patient (P); performing, by the medical device (11), feature-based delineation of the sensed EEG data to obtain tremor features present in the EEG data and indicative of an episode of tremor in the patient (P); determining, by the medical device (11) and based on the feature-based delineation, that the tremor features satisfy threshold criteria for application of a machine learning model for verifying that the episode of tremor has occurred in the patient (P); in response to determining that the tremor features satisfy the threshold criteria for application of the machine learning model, applying, by the medical device (11), the machine learning model, trained using EEG data for a plurality of patients (P), to the sensed EEG data to verify, based on the machine learning model, that the episode of tremor has occurred in the patient (P); and in response to verifying, by the machine learning model, that the episode of tremor has occurred in the patient (P): generating, by the medical device (1), a report comprising an indication that the episode of tremor has occurred in the patient (P) and one or more of the tremor features that coincide with the episode of tremor; and outputting, by the medical device (11) and for display, the report comprising the indication that the episode of tremor has occurred in the patient (P) and the one or more of the tremor features that coincide with the episode of tremor.