Systems and methods for detecting or predicting relapse in major depressive disorder

Abstract

A system and computer-implemented method for detecting a relapse of depression in a patient is provided. The system includes a wearable device configured to detect patient movement and generate actigraphy data corresponding to the patient's movement, and a computing device for retrieving the actigraphy data from the device. The system and method acquire training data, including training actigraphy data, over a training period and train an anomaly detector using the training data. The system and method then acquire test data from the patient, extract a plurality of features from the test data, and analyze the extracted data using the trained anomaly detector. Self-report testing is used to determine whether anomalies identified by the anomaly detector indicate a high likelihood that the patient will experience a relapse of depression.

Description

[Technical Field]

[0001] (Priority Claim) This application claims priority to U.S. Provisional Patent Application No. 63 / 049,053, filed July 7, 2020, and U.S. Provisional Patent Application No. 63 / 202,871, filed June 28, 2021, the entire contents of which are incorporated herein by reference. [Background technology]

[0002] Major depressive disorder (MDD) is one of the leading causes of disability worldwide (measured as years lived with the condition), with a lifetime prevalence of approximately 15% in the general adult population and associated with significant morbidity and mortality. This condition affects more than 300 million people worldwide. Patients with MDD can experience a wide range of physical, emotional, and cognitive symptoms, including depressed mood, loss of interest or pleasure in all or almost all activities, fatigue and sleep disturbances, and difficulty thinking, concentrating, and making decisions. These symptoms can severely impact a patient's daily life, including how they feel, think, and process daily activities, potentially affecting their health, relationships, employment, education, and overall quality of life. In severe cases of MDD, patients may consider death or suicide. Notably, individuals with MDD are at 20 times higher risk of suicide than the general population. Furthermore, MDD is thought to contribute to an increased risk of developing or worsening other health disorders. For example, MDD may increase the risk of developing conditions such as stroke and type 2 diabetes.

[0003] Various treatment options are available to help patients alleviate MDD symptoms and improve their quality of life. However, even with treatment, MDD is a chronic disorder with recurrent episodes, so patients may experience residual symptoms or relapses or recurrences of depression. In clinical practice, clinicians take a reactive approach by observing patients only during clinical visits and modifying the patient's treatment regimen as needed based on observations made during such visits. MDD is a dynamic disease with episodes of relapse interspersed with periods of remission. Transitions in disease state can occur on a timescale faster than the time between doctor's appointments. Using this reactive approach, clinicians often fail to notice early changes in a patient's symptoms. Relapse or recurrence is often detected only after the patient's depressive symptoms worsen sufficiently to require a clinical visit for evaluation.

[0004] Delaying further treatment after a relapse or recurrence may increase a patient's risk of self-harm or suicide. The percentage of MDD patients who achieve remission also significantly decreases after each treatment failure. Furthermore, prolonged and / or ineffective treatment may prolong a patient's suffering, reduce their expectations, and intensify negative emotions, such as feelings of helplessness. Therefore, early identification and recognition of a relapse or recurrence of depression may enable clinicians to prevent the disease from worsening early, which may be potentially lifesaving and may improve a patient's chances of achieving a meaningful response to treatment, potentially achieving remission. Summary of the Invention [Means for solving the problem]

[0005] One exemplary embodiment of the present invention relates to a computer-implemented method for detecting or predicting a relapse of depression in a patient. The method includes: (i) acquiring, from a wearable device worn by the patient, training data for the patient over a training period. The training data includes training actigraphy data corresponding to the patient's movements over the training period. The training period is a period during which the patient is not experiencing any signs of a relapse of depression. The method also includes (ii) training an anomaly detector using the training data. The anomaly detector is configured to identify deviations from the training data. The method further includes (iii) acquiring, from the wearable device, test data for the patient over a test period after the training period. The test data includes test actigraphy data corresponding to the patient's movements after the training period. The method further includes (iv) extracting a plurality of features from the test data to generate test feature data, the features corresponding to indices of at least one of activity, sleep, circadian rhythm, and multifractal dynamics. The method further includes (v) analyzing the test feature data using the anomaly detector and comparing the test feature data to the training data. The method further includes (vi) administering a self-report test to the patient to obtain a plurality of inputs from the patient if the anomaly detector determines that the test feature data is likely to be anomalous compared to the training actigraphy data, and (vii) analyzing the plurality of inputs from the patient to determine whether the patient is likely to experience symptoms of a return of depression.

[0006] A system for detecting or predicting a relapse of depression in a patient is also provided. The system includes a wearable device including at least one accelerometer configured to detect patient movement. The wearable device is configured to generate actigraphy data corresponding to the patient movement. The system also includes a computing device operably connected to the wearable device to receive the actigraphy data from the wearable device. The computing device includes a user interface for displaying output and receiving input from the patient, a processor, and a non-transitory computer-readable storage medium including a set of instructions executable by the processor. The set of instructions is operable to: acquire, from the wearable device, training actigraphy data corresponding to the patient's movement over a training period, the training period being a period during which the patient is not experiencing signs of a return of depression; train an anomaly detector using training data including the training actigraphy data, the anomaly detector being configured to identify deviations from the training data; acquire, from the wearable device, test actigraphy data corresponding to the patient's movement after the training period; extract a plurality of features from the test actigraphy data to generate test feature data, the features corresponding to indices in at least one of activity, sleep, circadian rhythm, and multifractal dynamics; analyze the test feature data using the anomaly detector and compare the test feature data to the training data; direct a user interface to display a plurality of self-report survey questions to the patient; receive a plurality of inputs from the patient via the user interface in response to the self-report survey questions; and analyze the plurality of inputs from the patient to determine whether the patient is likely to experience signs of a return of depression.

[0007] In another aspect, a computer-implemented method for detecting or predicting a relapse of depression in a patient is provided. The method includes (i) acquiring, from a wearable device worn by the patient, training data for the patient over a training period. The training data includes training actigraphy data corresponding to the patient's movements over the training period, the training period being a period during which the patient is not experiencing symptoms of a relapse of depression. The method also includes (ii) training an anomaly detector using the training data, the anomaly detector being configured to identify deviations from the training data. The method further includes (iii) acquiring, from the wearable device, test data for the patient over a test period, at least a portion of which is after the training period. The test data includes test actigraphy data corresponding to the patient's movements after the training period. The method further includes (iv) extracting a plurality of features from the test data to generate test feature data, the features corresponding to indices related to at least one of monofractal patterns, multifractal dynamics, and sample entropy. The method further includes (v) analyzing the test feature data using an anomaly detector and comparing the test feature data with training data to detect anomalies in the test feature data; and (vi) analyzing the self-report test data to determine whether the patient is likely to experience symptoms of a return of depression when an anomaly is detected in the test feature data, wherein the self-report test data is generated from multiple inputs from the patient in response to a self-report test.

[0008] In a further aspect, a system for detecting or predicting a relapse of depression in a patient is provided. The system includes a wearable device including at least one accelerometer configured to detect patient movement, the wearable device configured to generate actigraphy data corresponding to the patient movement. The system also includes a computing device operatively connected to the wearable device to receive the actigraphy data from the wearable device. The computing device includes a user interface for displaying output and receiving input from the patient, a processor, and a non-transitory computer-readable storage medium including a set of instructions executable by the processor. The set of instructions is operable to: acquire, from the wearable device, training actigraphy data corresponding to the patient's movements over a training period, the training period being a period during which the patient is not experiencing symptoms of a return of depression; train an anomaly detector using training data including the training actigraphy data, the anomaly detector being configured to identify deviations from the training data; acquire, from the wearable device, test actigraphy data corresponding to the patient's movements during a test period, at least a portion of the test period being after the training period; extract a plurality of features from the test actigraphy data to generate test feature data, the features corresponding to at least one activity-related indicator in at least one of monofractal patterns, multifractal dynamics, and sample entropy; analyze the test feature data using the anomaly detector to compare the test feature data to the training data to detect anomalies in the test feature data; and analyze the self-report test data to determine whether the patient is likely to experience symptoms of a return of depression when an anomaly is detected in the test feature data. The self-report test data is generated from a plurality of inputs received from the patient by a user interface in response to a self-report test that includes a plurality of self-report survey questions displayed on the user interface.

[0009] These and other aspects of the present invention will become apparent to those skilled in the art upon reading the following detailed description of the invention, including the drawings and appended claims. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 illustrates an exemplary system for detecting and / or predicting relapse of depression in a patient, according to an exemplary embodiment of the present application. [Figure 2] FIG. 1 illustrates an exemplary method for detecting and / or predicting relapse of depression in a patient, according to an exemplary embodiment of the present application. [Figure 3] FIG. 1 illustrates an exemplary method for training at least one anomaly detector to identify deviations from training data to determine whether an indication of depression relapse is likely based on passive patient data, according to an exemplary embodiment of the present application. [Figure 4] FIG. 1 illustrates an exemplary method for administering at least one self-report test to a patient and analyzing the results of the self-report test(s) to further determine whether the patient is likely to experience symptoms of a relapse of depression, according to an exemplary embodiment of the present application. [Figure 5] FIG. 1 illustrates another exemplary method for administering at least one self-report test to a patient and analyzing the results of the self-report test(s) to further determine whether the patient is likely to experience symptoms of a relapse of depression, according to an exemplary embodiment of the present application. [Figure 6] FIG. 1 is a typical schematic diagram of an LSTM anomaly detector with an encoder and a decoder. [Figure 7] FIG. 3 illustrates a typical timeline that a patient may experience for the exemplary method of FIG. 2 for detecting and / or predicting relapse of depression in a patient. [Figure 8] FIG. 1 illustrates another exemplary method for detecting and / or predicting relapse of depression in a patient, according to an exemplary embodiment of the present application. [Figure 9]FIG. 1 illustrates an exemplary method for identifying anomalies using a dynamic threshold. [Figure 10a] 9 illustrates an example time series of anomaly scores analyzed according to the steps of the exemplary method of FIG. 8. [Figure 10b] 9 illustrates an example time series of anomaly scores analyzed according to the steps of the exemplary method of FIG. 8. [Figure 10c] 9 illustrates an example time series of anomaly scores analyzed according to the steps of the exemplary method of FIG. 8. [Figure 10d] 9 illustrates an example time series of anomaly scores analyzed according to the steps of the exemplary method of FIG. 8. [Figure 10e] 9 illustrates an example time series of anomaly scores analyzed according to the steps of the exemplary method of FIG. 8. [Figure 10f] 9 illustrates an example time series of anomaly scores analyzed according to the steps of the exemplary method of FIG. 8. [Figure 11] FIG. 9 illustrates an example of an implementation of the exemplary method of FIG. 8 over sample time series data. [Figure 12] 9 illustrates an exemplary timeline that a patient may experience for the exemplary method of FIG. 8 for detecting and / or predicting relapse of depression in the patient. [Figure 13] FIG. 1 shows an exemplary timeline for collecting training actigraphy data and analyzing subsequent actigraphy data for a patient experiencing a depressive relapse, according to an exemplary embodiment of Example I. [Figure 14] FIG. 10 shows experimental data corresponding to the proportion of true positive relapse patients detected over a period of time prior to actual depression symptoms, according to an exemplary embodiment of Example II. [Figure 15a] FIG. 10 shows experimental data corresponding to the frequency with which patients are administered self-report tests at various trigger rates, according to an exemplary embodiment of Example III, where patient actigraphy data is used to determine when self-report tests are administered to patients. [Figure 15b]FIG. 10 shows experimental data corresponding to the frequency of patients who were administered self-report testing at various trigger rates when self-report testing was administered weekly according to Example III. [Figure 16a] FIG. 15 shows a subset of the experimental data from FIG. 15a for a time frame in which patients were in remission and not approaching a relapse of depression. [Figure 16b] FIG. 15 shows a subset of the experimental data from FIG. 15b for a time frame in which patients were in remission and not approaching a relapse of depression. [Figure 17a] FIG. 15b shows a subset of the experimental data of FIG. 15a for a time frame in which patients are approaching depression or experiencing a relapse of depression. [Figure 17b] FIG. 15b shows a subset of the experimental data from FIG. 15b for a time frame in which patients are approaching depression or experiencing a relapse of depression. [Figure 18] FIG. 10 shows experimental data of performance indicators of an exemplary method for determining depression relapse using patient actigraphy data and self-report testing, according to an exemplary embodiment of Example III. [Figure 19a] FIG. 10 shows data corresponding to the number of relapsed subjects over an increasing number of clinician visits, showing the distribution of relapse visits analyzed in Example V. [Figure 19b] FIG. 10 shows data corresponding to the number of non-relapse subjects over the course of increasing numbers of clinician visits, showing the distribution of non-relapse visits analyzed in Example V. [Figure 20] FIG. 1 shows an exemplary timeline of three different clinical visits evaluated according to the exemplary embodiment of Example V. DETAILED DESCRIPTION OF THE INVENTION

[0011] As used herein, the term "actigraphy" refers to a method of measuring a patient's movement and / or activity over a period of time, which may correspond to the patient's motor activity, sleep, or circadian rhythms.

[0012] As used herein, the term "relapse" or "returned" refers to having symptoms return after improvement and / or remission of depression within the same depressive episode, or after the return of symptoms as a new depressive episode. The term "relapse" encompasses both recurrence and recurrence of depression.

[0013] As used herein, the term "relapse" or "relapsed" refers to the return of symptoms after improvement and / or remission of depression within the same depressive episode. The same depressive episode can be a recurrence of depressive symptoms within a given period (e.g., within the first six months of starting a treatment regimen). In particular, the returning symptoms can be symptoms that meet the clinical diagnostic criteria for depression, such as those defined in the Statistical Manual of Mental Disorders (DSM-5). There are several different clinical tests, particularly those administered and evaluated by clinicians, that can be used to identify a patient's relapse of depression. In one example, a relapse of MDD can be identified by a clinician examining the patient using the Montgomery-Asberg Depression Rating Scale (MADRS), which is discussed further below.

[0014] The term "recurrent" as used herein refers to symptoms returning as a new depressive episode after improvement and / or remission of depression. The return of depressive symptoms after a predetermined period of time (e.g., after the first 6 months of starting a treatment regimen). Symptoms returning as a new recurrent episode of depression may be symptoms that meet clinical diagnostic criteria for depression, such as the clinical criteria defined in the Statistical Manual of Mental Disorders-5 (DSM-5).

[0015] The term "antidepressant" as used herein refers to any pharmaceutical agent that can be used to treat depression. Suitable examples include, but are not limited to, monoamine oxidase inhibitors, tricyclic serotonin reuptake inhibitors, serotonin noradrenergic reuptake inhibitors, noradrenergic and specific serotonergic agents, or atypical antipsychotics. Other examples include monoamine oxidase inhibitors such as phenelzine, tranylcypromine, and moclobemide; tricyclic antidepressants such as imipramine, amitriptyline, desipramine, nortriptyline, doxepin, protriptyline, trimipramine, clomipramine, and amoxapine; tetracyclic antidepressants such as maprotiline; acyclic antidepressants such as nomifensine; triazolopyridines such as trazodone; serotonin reuptake inhibitors such as fluoxetine, sertraline, paroxetine, citalopram, escitalopram, and fluvoxamine; serotonin receptor antagonists such as nefazadone; venlafaxine, milnacipran, and desbexamine. These include, but are not limited to, serotonin noradrenergic reuptake inhibitors such as lenlafaxine, duloxetine, and levomilnacipran; noradrenergic and specific serotonergic drugs such as mirtazapine; noradrenaline reuptake inhibitors such as reboxetine and edivoxetine; atypical antipsychotics such as bupropion; natural products such as kava and St. John's wort; dietary supplements such as s-adenosylmethionine; neuropeptides such as thyrotropin-releasing hormone; compounds that target neuropeptide receptors such as neurokinin receptor antagonists; and hormones such as triiodothyronine.In some embodiments, the antidepressant is imipramine, amitriptyline, desipramine, nortriptyline, doxepin, protriptyline, trimipramine, maprotiline, amoxapine, trazodone, bupropion, clomipramine, fluoxetine, duloxetine, escitalopram, citalopram, sertraline, paroxetine, fluvoxamine, nefazadone, venlafaxine, milnacipran, reboxetine, mirtazapine, phenelzine, tranylcypromine, moclobemide, kavakava, St. John's wort, s-adenosylmethionine, thyrotropin-releasing hormone, neurokinin receptor antagonist, or triiodothyronine. Preferably, the antidepressant is selected from the group consisting of fluoxetine, imipramine, bupropion, venlafaxine, and sertraline.

[0016] Therapeutically effective amounts / dosage levels and dosing regimens of antidepressants (e.g., monoamine oxidase inhibitors, tricyclic antidepressants, serotonin reuptake inhibitors, serotonin noradrenergic reuptake inhibitors, noradrenergic-specific serotonergic agents, noradrenaline reuptake inhibitors, natural products, dietary supplements, neuropeptides, compounds targeting neuropeptide receptors, hormones, and other pharmaceutical agents described herein) may be readily determined by one of skill in the art. For example, therapeutic dosages and regimens for pharmaceutical agents approved for marketing are generally available and are listed, for example, on package labels, standard dosing guidelines, standard medication references such as the Physician's Desk Reference (Medical Economics Company or online at http: / / www.pdrel.com), or other sources.

[0017] The present application relates to systems and methods for detecting and / or predicting a relapse of depression in a patient using passive patient data from the patient and data corresponding to self-report tests in a computer-implemented method. Passive patient data can include any suitable type of data that can be passively collected during a patient's daily activities. In particular, passive patient data can be collected without the patient's active involvement with sensors and / or devices (e.g., without the patient constantly monitoring sensors and / or devices and manually providing input). For example, passive patient data can include passively collected data corresponding to the patient's physical behavior and / or data corresponding to the patient's use of electronic devices. In one embodiment, the present systems and methods for detecting and / or predicting a relapse of depression can utilize passive patient data including actigraphy data.

[0018] The systems and methods of the present application can be used for patients suffering from MDD, particularly for patients receiving treatment for MDD whose symptoms are in remission. The treatment can include psychotherapy, brain stimulation therapy, and / or administration of antidepressants. Specifically, the patient can be a non-treatment-resistant patient with MDD, such as a patient with MDD who has received and responded to treatment and continues to respond to and receive treatment. The patient can be a patient who has received and responded to treatment and has returned to normal under treatment. In particular, the present application relates to systems and methods for detecting and / or predicting the return of MDD symptoms in patients who have responded to an antidepressant treatment regimen and continue to respond to and receive the antidepressant treatment regimen. Furthermore, the systems and methods of the present application can be used for patients receiving treatment for MDD whose symptoms are in remission but who have a history of previous episodes in which depressive symptoms return. Although the exemplary embodiments described herein refer to depression relapse, it is contemplated that the present application may be used to detect and / or predict any type of depression symptom return, including depression recurrence, which refers to the return of symptoms as a new depressive episode.

[0019] FIG. 1 illustrates an exemplary embodiment of a system 100 for detecting and / or predicting depressive relapse using passive patient data and, optionally, data corresponding to self-reported characteristics of physical behavior (e.g., the patient's self-assessment of activity or sleep adequacy). System 100 comprises a device 200 for passively detecting and generating data corresponding to the patient's physical behavior (e.g., physical activity, sleep, movement, etc.) and a computing device 300 for receiving the data from device 200 and analyzing the data to determine whether the patient is likely to experience symptoms of depressive relapse. In one embodiment, device 200 detects and generates actigraphy data and / or mobility data of the patient. Actigraphy data corresponds to the patient's movements over time. Movement data corresponds to patterns of movement by the patient over time, such as, for example, movement traces. Device 200 is preferably of an appropriate size and shape to be wearable on the patient's body. For example, wearable device 200 may be in the form of a wearable clip that can be attached to the patient for wearing on the patient's body throughout the day. In another embodiment, the device 200 is attached to a wearable band 250 (eg, a watch band) for attaching the device 200 to the patient's wrist when the device 200 is in an operational configuration.

[0020] 1 , device 200 includes a processor 202, a computer-accessible medium 204, at least one sensor 206, and an input / output device 208. Sensor 206 may include actigraphy sensor(s) for detecting patient movement and / or mobility sensor(s) for detecting patient movement patterns. The actigraphy sensor may be any suitable sensor for detecting patient movement. For example, the actigraphy sensor may be an accelerometer for detecting patient movement when device 200 is worn by the patient in an operational configuration. The mobility sensor may be any suitable sensor for detecting patient movement patterns. For example, the mobility sensor may be a Global Positioning System (GPS) device for detecting patient positioning when device 200 is worn by the patient in an operational configuration.

[0021] The sensor(s) 206 are operably connected to the processor 202 to provide data generated by the sensor(s) 206 to the processor 202. The processor 202 receives data from the sensor(s) 206 and generates data corresponding to the patient's physical behavior, such as, for example, actigraphy data and / or mobility data for the patient. The processor 202 may include, for example, one or more microprocessors and use instructions stored on a computer-accessible medium 204 (e.g., a memory storage device). The computer-accessible medium 204 may, for example, be a non-transitory computer-accessible medium containing executable instructions therein. The system 100 may further include a memory storage device 210 separate from the computer-accessible medium 204 for storing actigraphy data and / or mobility data. The input / output device 208 is any suitable device for receiving and / or transmitting data or instructions to and from the actigraphy device 200. In particular, the input / output device 208 may be a transceiver for receiving instructions to the device 200 and / or transmitting data from the device.

[0022] Device 200 is operably connected to computing device 300 to communicate some or all of the data collected by device 200 to computing device 300 or to allow computing device 300 to retrieve some or all of the data from device 200. As shown in FIG. 1 , device 200 can be operably connected to computing device 300 via a communications network 110 (e.g., the Internet, Wi-Fi, a wide area network, a local area network, a cellular network, a personal area network, etc.). In particular, input / output device 208 is operably connected to communications network 110 to receive instructions therefrom or transmit data therethrough. In certain embodiments, communications network 110 is a wireless network, more specifically a short-range wireless network such as a personal area network (e.g., Bluetooth®) having a limited range for connecting devices near the patient. However, it is also contemplated that device 200 can be directly connected to computing device 300 via a wired connection.

[0023] The computing device 300 in this embodiment comprises a processor 302, a computer-accessible medium 304, and an input / output device 306 for receiving and / or transmitting data or instructions to and from the computing device 300. The processor 302 may include, for example, one or more microprocessors and may use instructions stored on the computer-accessible medium 304 (e.g., a memory storage device). The computer-accessible medium 304 may, for example, be a non-transitory computer-accessible medium containing executable instructions therein. The input / output device 306 is operatively connected to the communications network 110 and may receive instructions therefrom or transmit data therethrough. The computing device 300 may also include a user interface 308 (e.g., a touchscreen) for obtaining input from and displaying output to a user. It is contemplated that the user interface 308 may also be two separate components, such as a display and a keyboard, for displaying output to and obtaining input from a patient. The user interface 308 is operatively connected to the processor 302 for providing instructions for generating output on the user interface and providing data corresponding to input obtained from the patient to the processor 302, as described further below. The computing device 300 may further include a memory storage device 310 for storing past actigraphy data, past mobility data, past input from the patient, medical data, pharmacy data, and / or at least one anomaly detector for determining possible signs of depression relapse, wherein the at least one anomaly detector is generated and / or trained by the computing device 300. The computing device 300 may be, for example, a mobile computing device, a smartphone, a computing tablet, a computing device, etc.

[0024] In some embodiments, computing device 300 is also configured to collect additional passive patient data, particularly data corresponding to the patient's electronic device usage. Specifically, computing device 300 is a mobile phone or computing tablet that the patient also uses in their daily activities. For example, the patient may use computing device 300 for activities such as web browsing, social media use, texting, gaming, phone calls, and other activities that may typically be involved with a personal electronic device. Computing device 300 may be configured to track use of computing device 300 during the patient's activities and generate data corresponding to use of computing device 300, such as keyboard usage activity, keystroke dynamics, text context, and the like. In this embodiment, system 100 may collect passive patient data using both device 200 and computing device 300. In an alternative embodiment, system 100 collects and analyzes passive patient data from computing device 300, particularly data corresponding to the patient's electronic device usage, and excludes device 200 from system 100.

[0025] FIG. 2 illustrates an exemplary method 400 for detecting and / or predicting a patient's depression relapse. The exemplary method 400 utilizes both passive patient data and administration of self-report test(s) to determine the likelihood of a patient experiencing a symptom of depression relapse. The passive patient data provides an objective, quantifiable indicator corresponding to the likelihood of a symptom of depression relapse. This portion of the method 400 provides an objective triggering stage that screens patients to determine whether the passive patient data is abnormal compared to passive patient data previously collected from the patient during a period in which the patient was not known to have experienced depressive symptoms. If the method 400 detects an abnormality in the passive patient data, the patient may be at higher risk for a depression relapse. If the passive patient data suggests that the patient may be at risk, further data can be collected in the confirmation stage of the method 400 to more accurately determine whether the patient is likely experiencing a symptom of depression relapse. In particular, the confirmation stage may include self-report test(s) administered along a quantitative scale, thereby providing a further quantifiable indicator corresponding to the likelihood of a symptom of depression relapse.

[0026] The exemplary method 400 utilizes at least one machine learning anomaly detector that trains the anomaly detector based on the patient's own historical data (n=1, where n is the total number of sampled individuals). Thus, the trained anomaly detector is personalized to each individual patient. The exemplary method 400 collects training data from the patient, as described below, and continues to repeatedly acquire and analyze test data if the patient is not known to have relapsed from depression. The exemplary method 400 may be repeated continuously or as frequently as desired, e.g., daily, weekly, biweekly, etc., thereby allowing for regular monitoring and / or earlier detection / prediction of depressive relapse compared to patients who self-report and seek medical attention only after realizing that a relapse has already occurred.

[0027] A patient's relapse into depression may be identified using one or more tests administered by a mental health provider (e.g., a psychiatrist, physician, psychologist, or therapist) or self-administered by the patient. In this exemplary embodiment, computing device 300 may receive medical data corresponding to the patient's medical record, such as an electronic medical record (EMR), and / or pharmacy data corresponding to the patient's medication record, and computing device 300 may analyze the medical data and / or pharmacy data to determine whether the patient has relapsed into depression. If computing device 300 determines from the medical data and / or pharmacy data that the patient has relapsed, exemplary method 400 ends. However, it is contemplated that exemplary method 400 may resume via manual input from a mental health provider into computing device 300 or when computing device 300 detects from the medical data and / or pharmacy data that the patient has returned to remission from depression.

[0028] In one exemplary embodiment, the computing device 300 analyzes the medical and / or pharmacy data and can determine that a depressive relapse has occurred if the medical and / or pharmacy data includes data corresponding to a patient who: (1) has been diagnosed with a depressive relapse by a mental health provider; (2) has experienced severe symptoms of depressive illness (e.g., worsening depression, intentional suicidal ideation, or hospitalization for suicidal behavior), or (3) scores above a predetermined threshold on a quantitative test and subsequent validation, as further described below. In particular, the quantitative test can be an assessment of the MADRS by the patient's mental health provider to determine whether the patient has relapsed into depression. The MADRS measures depression severity and detects changes due to antidepressant treatment. The test consists of 10 items, each scored from 0 (absent or normal) to 6 (severe or continued presence of symptoms), for a total possible score of 60. Higher scores indicate a more severe condition. The MADRS assesses apparent sadness, reported sadness, inner tension, sleep, appetite, concentration, fatigue, interest level, pessimistic thoughts, and thoughts of suicide.

[0029] The computing device 300 determines that a patient has relapsed if the medical and / or pharmacy data includes data corresponding to a patient with a MADRS total score of ≥ 22 and includes data corresponding to a subsequent verification. The data corresponding to the subsequent verification may include data corresponding to (1) a change in treatment regimen (e.g., a change in medication type, medication dose, or medication frequency) within a certain period (e.g., within 14 days) from when the patient was observed to have a MADRS total score of ≥ 22, or (2) a separate test indicating worsening depression. The data corresponding to the separate test may include data corresponding to an increase of at least a predetermined threshold amount on a different quantitative scale, as assessed by a mental health care provider. For example, the separate test may be an assessment by the patient's mental health care provider using the Clinical Global Impression-Severity (CGI-S) scale, which is a scale for assessing the severity of the patient's illness at the time of assessment relative to the mental health care provider's past experience with patients with the same diagnosis and improvement with treatment. Taking into account all clinical experience, patients are rated on the CGI-S scale based on the severity of their mental illness according to the following: 0 = not rated, 1 = normal (not ill at all), 2 = borderline psychotic, 3 = mildly ill, 4 = moderately ill, 5 = significantly ill, 6 = severely ill, 7 = most severely ill. Computing device 300 determines that a subsequent validation has occurred if the medical data includes data corresponding to a follow-up visit with a mental health provider in which the patient was rated as having an increase in CGI-S score of 2 or more from baseline.

[0030] In step 402, passive patient data is collected by device 200 and / or computing device 300 over a predetermined training period to generate training data. The training data may include data corresponding to the patient's physical behavior over the training period and / or data corresponding to the patient's electronic device use over the training period. In one embodiment, the training data includes training actigraphy data and / or training mobility data. Specifically, device 200 may be worn by the patient to detect the patient's movements and generate training actigraphy data over the predetermined training period. The training actigraphy data corresponds to athletic activity within the training period and / or the patient's movements during sleep. Similarly, device 200 may be worn by the patient to detect the patient's movement patterns and generate a set of training mobility data over the predetermined training period.

[0031] The device 200 may be worn by the patient continuously or substantially continuously. For example, the device 200 may be worn substantially continuously such that the device is removed from the patient for short periods of time to allow the patient to engage in activities where the patient may not be suitable or able to wear the device, such as showering, exercising, and / or washing. In other embodiments, the device 200 may be worn by the patient daily. In particular, the device 200 may be worn by the patient daily for a majority of the patient's waking hours (e.g., at least 95%, at least 90%, at least 80%, at least 70%, or at least 60% of the time) and / or while the patient is asleep. The predetermined training period may be any suitable period for collecting training actigraphy datasets for training at least one anomaly detector for determining possible signs of depression relapse. For example, the training period may be one month or more, three months or more, or six months or more. In one embodiment, the training period is three months.

[0032] The training actigraphy data and / or training mobility data may be stored in memory storage device 210 of device 200 until device 200 is operably connected to computing device 300 to transmit all or a portion of the training actigraphy data and / or training mobility data to computing device 300. Specifically, all or a portion of the training actigraphy data and / or training mobility data may be transmitted from device 200 via input / output device 208 and received by computing device 300 via input / output device 306. In an alternative embodiment, the training actigraphy data and / or training mobility data is transmitted continuously from device 200 to computing device 300 as the actigraphy data and / or mobility data is being collected by device 200. More particularly, the actigraphy data and / or mobility data is transmitted wirelessly from device 200 to computing device 300 in real time or substantially real time as the actigraphy data and / or mobility data is being collected by device 200.

[0033] In some embodiments, the training data can further include data corresponding to training self-report data obtained during a predetermined training period. The training self-report data corresponds to self-reported characteristics of physical behavior input by the patient into the computing device 300 over the training period. For example, the processor 302 can direct the user interface 308 to display a plurality of questions prompting responses from the patient and receive a plurality of inputs from the user via the user interface 308 in response to the questions. The plurality of questions can form a self-reported assessment of characteristics of physical behavior (e.g., the patient's activity, sleep adequacy, self-assessment of sleep quality). More specifically, the self-reported assessment includes questions regarding characteristics of physical behavior that cannot be passively measured by the device 200 or the computing device 300. For example, the self-reported assessment can include questions regarding the patient's perception of rest and / or sleep. In one example, the self-reported assessment can include all or a portion of questions from the Medical Outcome Study Sleep (MOS-S) scale assessment. Preferably, the self-report assessment includes a limited number of questions to minimize the patient's burden to actively engage with the user interface 308 (e.g., answer questions). For example, the self-report assessment may include 12 or fewer questions, 10 or fewer questions, 5 or fewer questions, or 3 or fewer questions. In one embodiment, the self-report assessment includes two questions. For example, the self-report assessment includes two questions that ask the patient to provide a quantitative assessment of their sleep perception, such as (1) whether they feel rested and (2) whether they feel they have had enough sleep. The self-report assessment can be repeated at any desired time intervals (e.g., daily, every other day, weekly, etc.) during a predetermined training period, and input obtained during the training period is used to generate the training self-report data.

[0034] The training data obtained in step 402 is used to train at least one anomaly detector for determining likelihood of a symptom of depression relapse. The anomaly detector(s) comprise machine learning anomaly detector(s) configured to identify deviations from the training data, as described further below with respect to step 408. Figure 3 shows an exemplary embodiment of a method 500 for using the training data obtained in step 402 to train at least one anomaly detector for determining likelihood of a symptom of depression relapse. In one exemplary embodiment, the training data may be transmitted to computing device 300 and used by computing device 300 in exemplary method 500 to train at least one anomaly detector for determining likelihood of a symptom of depression relapse.

[0035] At step 502, computing device 300, and in particular processor 302, analyzes and extracts a plurality of features from the training data from step 402 to generate training feature data. In one embodiment, processor 302 analyzes the raw data obtained from device 300 from step 402 and extracts a plurality of features from the raw data. For example, the raw data may be raw accelerometer data obtained from device 200. In one exemplary embodiment, processor 302 analyzes the training data from step 402 and extracts a plurality of features to generate training feature data. The features may correspond to actigraphy, mobility, and / or social activity of the patient. For example, the features may correspond to indicators of at least one of sleep changes, decreased ability to concentrate in an activity, decreased interest or pleasure in an activity, and / or fatigue or loss of energy, such as sleep time, sleep onset, sleep end, sleep disturbances, immobility time, phone unlock time, phone unlock time while at a specific location, talking time, number of locations visited, time spent at a specific location, heart rate, fractal activity patterns, monofractal patterns, or multifractal dynamics of the activity data, entropy of all or a portion of the activity data, among others, as described in Wang et al., “Tracking Depression Dynamics in College Students Using Mobile Phone and Wearable Sensing,” Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies, Vol. 2, No. 1, Article 43 (March 2018), which is incorporated herein by reference in its entirety.

[0036] In particular, processor 302 analyzes and extracts a plurality of actigraphy features from the training actigraphy data from step 402 to generate at least a portion of the training feature data. The extracted actigraphy features may correspond to indicators of at least one of the patient's motor activity, sleep, wakefulness, and circadian rhythm. For example, features can include indices for sleep duration (e.g., total sleep duration per night), sleep pattern (e.g., time of sleep onset, time of sleep end), sleep quality (e.g., average activity counts per minute during rest periods, percentage of sleep fragmentation, percentage of sleep efficiency, behavioral estimates of wakefulness minutes after sleep onset), fractal patterns or dynamics / behavior during activity or sleep (e.g., changes in monofractal patterns or multifractal dynamics during activity or sleep periods, temporal local variations to sleep patterns captured by different scaling characteristics at different activity periods), daytime activity (e.g., average daytime activity counts per minute, peak daytime activity counts per minute), and entropy during activity or sleep (e.g., measures of randomness and chaotic signatures during activity or sleep). More specifically, actigraphy features include indices for detecting sleep disorders, which may include indices for a patient's sleep, wake, and / or circadian rhythms. Sleep disturbances are a common symptom in patients with MDD, and patients are likely to experience poor sleep quality before or during recurrent or recurrent depressive episodes.

[0037] The entropy of actigraphy data provides a quantified measure of data complexity that can be used as one of multiple actigraphy features. In one example, actigraphy data for an activity time series can be represented as x(i), i = 1 to N, with Δt = 1 minute. The sample entropy (SaEn) of an activity time series x(i) involves the reconstruction of a vector from the activity time series, which is an (m-dimensional) state-space representation of the overall dynamics of the system from which the actigraphy data is derived. The vector can be represented as V(i) = {x(i), x(i + δ), ...., x(i + [m-1] δ)}, where δ is the time delay between successive components of the vector. The vector in this example is set to 1 for complexity analysis. SaEn is determined as the logarithmic difference between the probability (density) of vector V(i) occurring within a selected distance r in m dimensions and the probability of vector V(i) occurring within the same selected distance r in m + 1 dimensions. The density of the state space, ρ, m (r) and ρ (m+1) (r) is an index of the fraction of the reconstruction vectors that fall within the selected radius r of the m-th and m+1-th dimensions, respectively. SaEn can be expressed as follows:

[0038]

number

[0039] In one exemplary embodiment, each day of actigraphy data may be evenly divided into four periods: morning (6 AM - 12 PM), afternoon (12 PM - 6 PM), evening (6 PM - 12 AM), and night (12 AM - 6 AM). SaEn may be determined for each day and period. In one example, SaEn is determined as the median SaEn value (in z-score activity counts) over the last seven days (with no identified gaps) of activity counts.

[0040] The features may include one or more of the exemplary actigraphy features listed and defined in Table 1 below.

[0041] [Table 1]

[0042] Furthermore, actigraphy data can contain fractal fluctuations (e.g., temporal, structural, and / or statistical fluctuations over a wide range of time scales), which are believed to be stable within the same individual but may be susceptible to pathological conditions. Data demonstrating fractal regulation are believed to represent the adaptability of physiological systems and reflect the complexity of physiological networks in which regulatory processes function interactively over a wide range of time scales. Thus, measures of fractal fluctuations in actigraphy data can be extracted as actigraphy features in place of, or in addition to, one or more of the features identified above. In one example, actigraphy data exhibit complex temporal fluctuations characterized by scale-invariant (monofractal) patterns that can be utilized as actigraphy features. Monofractal patterns are uniform and have the same scaling properties throughout the signal. Actigraphy data can also exhibit a special class of complex processes called multifractals, which can also be included as actigraphy features. Multifractal behavior is characterized by distinct signatures at different time scales (minutes to hours). Multifractal behavior can involve multiple coexisting dynamical processes that can generate temporally local fluctuations captured by different scaling characteristics at different times.

[0043] Fractal patterns in actigraphy data can be determined using detrended fluctuation analysis (DFA). DFA determines the scaling behavior of fluctuations in actigraphy data over a range of timescales from minutes to hours. DFA examines multiscale correlations of activity fluctuations across multiple timescales. For example, DFA provides the fluctuation amplitude F(n) as a function of timescale n. For long-range correlation data F(n), the power law F(n)~nα According to [theorem], the scaling exponent or variability index (α) quantifies multiscale correlation as follows: if α = 0.5, the variability is uncorrelated ("white noise"); if α > 0.5, the variability is positively correlated (large values ​​are more likely to follow large values, and vice versa); if α < 0.5, the variability is negatively correlated (large values ​​are more likely to follow small values, and vice versa). Many physiological outputs under healthy conditions exhibit variability index (α) values ​​close to 1.0, indicating the most complex underlying control mechanisms. DFA detrends the data using a second-order polynomial function, eliminating the effects of possible linear trends in the data.

[0044] In one embodiment, the multifractal dynamics of actigraphy data is determined using an MFDFA method. In the exemplary MFDFA method of the present application, the variability of the data is generally determined by:

[0045]

number

[0046] In one example, the DFA method may be a particular embodiment of the MFDFA method, where F(n)≈n α The amplitude fluctuations F(n) at different time scales n can be derived as F(n), where the scaling exponent α or fluctuation exponent α indicates the correlation in the fluctuations. Values ​​of α > 0.5 indicate positive correlation (large values ​​are likely to be followed by large values, and vice versa), α < 0.5 indicates negative correlation (large values ​​are likely to be followed by small values, and vice versa), and α = 1 indicates long-range correlation or monofractal expression. Many physiological outputs under healthy conditions exhibit α values ​​close to 1.0, indicating the most complex underlying control mechanisms. The amplitude fluctuations are quantified at different moments. The multifractal spectrum is then calculated as the distribution of different scaling exponents α for different moments. The wider the width (α) of the multifractal spectrum, the 最大 -α 最小) indicates the presence of multifractal dynamics, while shorter widths indicate the absence of multifractal dynamics or the presence of monofractal dynamics. Multifractal spectra exhibit two measurable dimensions: Dq (q-th order singularity / fractal dimension) and hq (q-th order singularity index) (as described in Matic et al., "Objective differentiation of neonatal EEG background grades using detrended fluctuation analysis." Front. Hum. Neurosci., 9:189 (2015) (available at https: / / www.frontiersin.org / articles / 10.3389 / fnhum.2015.00189), incorporated herein by reference). Visual evaluation of the multifractal spectra reveals differences in horizontal and vertical positioning (hq, Dq values), width (width_hq), and the general shape of the multifractal spectra, which reflect the temporal variation of the local Hurst exponent; these features can be used as multiple actigraphy features for training anomaly detectors.Exemplary multifractal detrended fluctuation analysis methods are described in Ihlen et al., "Introduction to multifractal detrended fluctuation analysis in Matlab," Front. Front. Physiol., Vol. 3, Article 141, pp. 1-18 (available at https: / / www.frontiersin.org / articles / 10.3389 / fphys.2012.00141 / full); Ivanov, et al., "Multifractality in human heartbeat dynamics." Nature 399:461-465 (1999) (available at https: / / www.nature.com / articles / 20924); and Franca et al., "On multifractals: a non-linear study of actigraphy data." Physica A: Statistical Mechanics and its Applications. 514:612-619 (2019) (available at https: / / www.sciencedirect.com / science / article / pii / S037843711831255X), and Kantelhardt et al., "Multifractal detrended fluctuation analysis of nonstationary time series," Physica A, 316:87-114 (2002), all of which are incorporated herein by reference.

[0047] Instead of or in addition to the actigraphy features described above, other features can be extracted from the training data. For example, features may be extracted from the training data as described in Barnett et al., "Inferring Mobility Measure from GPS Traces with Missing Data," Biostatistics, pp. 1-33 (2018); Canzian et al., "Trajectory of Depression: Unobstrusive Monitoring of Depressive States by Means of Smartphone Mobility Traces Analysis," UbiComp'15: Proceedings of the 2015 ACM International Joint Conference on Pervasive and Ubiquitous Computing, pp. 1293-1304 (September 2015); and Canzian et al., "From Mobile Phone Monitoring of Depressive States using GPS Traces Analysis to Data-Driven Behavior Change Interventions," Frontiers in Public Health (January 2019), all of which are incorporated herein by reference in their entireties. 2016), can be extracted from patient mobility data, e.g., movement traces, geographical features, total distance covered, maximum distance between two locations, turning radius, standard deviation of displacement, maximum distance from home, number of different places visited, number of different important places visited.In another example, the additional feature may be extracted data corresponding to the patient's electronic device usage, as described in Mastoras et al., "Touchscreen typing pattern analysis for remote detection of the depression tendency," Nature Scientific Reports, 9:13414 (2019) and Zulueta et al., "Predicting Mood Disturbance Severity with Mobile Phone Keystroke Metadata: A BiAffected Digital Phenotyping Study," J. Med. Internet Res., 20(7):e241 (July 2018), which are incorporated by reference in their entireties.

[0048] The processor 302 may extract any suitable number of features from the training data. Increasing the number of features may improve the predictive performance of the systems and methods of the present application, but may be computationally burdensome. Therefore, an appropriate number of features may be selected to balance predictive performance and computational efficiency. In some embodiments, the processor 302 may extract at least 10, at least 20, at least 30, at least 40, or at least 50 different features from the training data. In one particular embodiment, the processor 302 extracts 31 features from the training actigraphy data. More specifically, the processor 302 may extract some or all of the features identified above in Table 1. In another embodiment, the processor 302 may extract features from the training data for detecting sleep disorders, which may include indices of a patient's sleep, wakefulness, and / or circadian rhythm. In further embodiments, the processor 302 may extract features from the training data including monofractal patterns during activity or sleep, multifractal dynamics / behavior during activity or sleep, and / or entropy during activity or sleep.

[0049] In step 504, processor 302 filters the training feature data extracted from the training data such that features and / or time points that are missing a significant portion of the data points (e.g., more than 30%, more than 40%, or more than 50%) are removed from the training feature data to generate filtered training data. For example, the filtered feature data retains only features that are missing less than 40% of the data points for that feature and only time points that are missing less than 40% of the features for that time point.

[0050] In step 506, processor 302 may further modify the filtered training data by imputing missing data points to generate imputed training data. In an exemplary embodiment, missing data points may be estimated by processor 302 by analyzing its nearest neighbors to generate an estimate of the missing data point based on the nearest neighbors. For example, missing data for a feature at a particular time point may be estimated by processor 302 using data from time points adjacent to the time point for which no data is present. Each missing data point may be estimated using its k nearest neighbors, where k may be between 2 and 10, between 3 and 8, or between 4 and 6. In one embodiment, k=5. Processor 302 may estimate the missing data points based on their k nearest neighbors using any suitable method. For example, processor 302 may utilize a multi-parameter Euclidean distance from the training data to separate the k nearest neighbors and use the mean value of the different features as input for the missing feature. In another exemplary embodiment, missing data points may be estimated by processor 302 by analyzing the filtered training data using a bagged tree. Specifically, for each feature, a decision tree module, particularly a bagging tree module, can be created using other remaining features from the filtered training data to impute missing data. Furthermore, the decision tree module can include a surrogate split, such that the decision tree module traces to a left or right child node using the best surrogate predictor to impute missing data points. In step 508, the imputed training data can be centered and scaled using the population mean and standard deviation (data mean (population data) / standard deviation (population data)). A power transformation (e.g., a Yeo-Johnson transformation or a Box-Cox transformation) can also be applied to all features of the imputed training data to generate modified training data with a normalized distribution.

[0051] The modified training data obtained from step 508 is used by processor 302 in step 510 to train at least one anomaly detector for determining possible signs of depression relapse. Once the anomaly detector(s) have been trained by processor 302 according to the exemplary method 500 shown in FIG. 3 , method 400 proceeds to step 404, where test data is acquired from the patient. Similar to step 402, test data is collected by device 200 and / or computing device 300 in step 404 after the training period; therefore, the test data acquired in step 402 is data not previously utilized in training the anomaly detectors in method 500. The test data may be generated in real time or over a desired test period. The test period may range from one day or about one day to two weeks or about two weeks. In one embodiment, the test period is one week or about one week or two weeks or about two weeks. In step 406, processor 302 analyzes and extracts features from the test data in a manner similar to that described above with respect to step 502 to generate test feature data.

[0052] In step 408, processor 302 specifically analyzes the test feature data from step 406 using anomaly detector(s) trained by method 500 in step 510 to determine a likely indication of a depressive relapse. In particular, the anomaly detector(s) compare the test feature data to the training data (or data derived from the training data, such as modified training data) to determine a likely indication of a depressive relapse. In one embodiment, processor 302 may analyze the test feature data using the anomaly detector(s) to determine whether the test feature data is likely to be anomalous compared to the training data. An anomaly is data that has different characteristics from the training data. As previously mentioned, the training data was collected when the patient was not experiencing any signs of a depressive relapse, and therefore, anomalies from the training data correspond to signs of a depressive relapse.

[0053] The anomaly detector(s) comprise machine learning anomaly detector(s) configured to identify deviations from the training data. The anomaly detector(s) can comprise supervised and / or unsupervised learning anomaly detectors. For example, the anomaly detector(s) can use modified training data (e.g., when a patient does not have a relapse of depression) to build a profile of normal cases and identify any further data that deviates from the normal profile as an anomaly (e.g., when a patient experiences symptoms of a relapse of depression). In an alternative, the anomaly detector(s) can identify any further data that deviates from the modified training data as an anomaly by isolating it using multiple binary trees.

[0054] The anomaly detector(s) may include any suitable type of anomaly detector(s) for detecting anomalies from the modified training data. For example, the anomaly detector(s) may utilize a one-class support vector machine (one-class SVM), an isolation forest (IF) module, a one-class neural network (e.g., a long short-term memory (LSTM) network), and other one-class analysis methods. In one embodiment, the anomaly detector(s) include at least one of a one-class SVM and an IF binary tree. Exemplary one-class SVMs that may be applied to the actigraphy data of the present application include those described in Scholkopf et al., "Support Vector Method for Novelty Detection," Advances in Neural Information Processing Systems, 582-588 (2000), Tax & Duin, "Support Vector Data Description," Machine Learning, 54:45-66 (2004), and Manevitz and Yousef, "One-Class SVMs for Document Classification," Journal of Machine Learning Research, 2:139-154 (2001), all of which are incorporated herein by reference in their entireties. In one exemplary embodiment, the anomaly detector(s) comprise an outlier SVM, as described in Manevitz and Yousef, "One-Class SVMs for Document Classification," Journal of Machine Learning Research, 2:139-154 (2001), which is incorporated herein by reference. As another example, the exemplary IF anomaly detector using an IF binary tree may be applied to actigraphy data in the present application.IF anomaly detectors include the iForest methodology described in Liu et al., "Isolation-Based Anomaly Detection," ACM Transactions on Knowledge Discovery from Data, 6(1):1-39 (March 2012), which is also incorporated herein by reference.

[0055] In one embodiment, processor 302 analyzes the test feature data using at least one anomaly detector to generate a binary output (i.e., 0 or 1 indicating whether the test data is likely to correspond to a depression relapse) and / or to generate an anomaly score corresponding to the probability that the test data is likely to correspond to a depression relapse. In particular, the anomaly detector(s) may comprise a one-class SVM anomaly detector configured to generate a binary output, where 0 indicates that the test feature data is likely to correspond to a non-relapse of the patient and 1 indicates that the test feature data is likely to be anomalous compared to the training data. The anomaly detector(s) may further comprise, or alternatively may comprise, a tree-based anomaly detector, specifically an IF anomaly detector, to generate an anomaly score corresponding to the probability that the test feature data is likely to be anomalous. Processor 302 determines that the test feature data is likely to be anomalous compared to the training data if the anomaly score is above a certain threshold. The threshold may be selected to distinguish between signal and noise, i.e., to separate data points likely to correspond to anomalies from data points corresponding to general variability in the patient's test data analyzed by the IF anomaly detector. For example, the processor 302 may determine that if the anomaly score from the IF anomaly detector is ≧0.6, the test feature data is likely to correspond to an anomaly and therefore likely to correspond to a relapse of depression.

[0056] In one embodiment, the anomaly detector can model normal behavior of the actigraphy data and use the model's prediction error to identify anomalies. For example, the anomaly detector can utilize a long-short-term memory (LSTM) neural network to analyze the multifractal dynamics of the test actigraphy data to quantify prediction errors, which are then used to identify anomalies. More specifically, the anomaly detector analyzes the actigraphy data through a stacked LSTM neural network with two components: an encoder that learns a vector representation of the input time series, and a decoder that uses the vector representation to reconstruct the time series. The reconstruction error of the test feature data is used to determine the likelihood of an anomaly.

[0057] FIG. 6 is an exemplary schematic diagram of an LSTM anomaly detector with an encoder and decoder (Enc-Dec AD). Actigraphy feature data 802 extracted from actigraphy data collected from a patient over time is represented as a time series of vectors over time, with t1 indicating the earliest time point in the data series. In FIG. 6, actigraphy feature data 802 is extracted for m features from actigraphy data collected over a period of time having a total length of p. A subset of actigraphy feature data 804, i.e., a time series having length l, may be test feature data. The Enc-Dec AD includes an encoder 604 for training a time series of a subset of actigraphy feature data 806 using results from the encoder 604 and then reconstructing an output sequence 810 using a decoder 808. The Enc-Dec AD determines an error vector for each point in the actigraphy feature data. At time t i The error vector for

[0058]

number

[0059]

number

[0060] Because relapse to depression typically occurs gradually over several weeks, continuous detection of actigraphy markers for potential relapse of depression over a period of time can be useful to improve the specificity and reduce the likelihood of false positives generated by exemplary method 400. Accordingly, processor 302 can analyze multiple results generated by the anomaly detector(s) over a desired period of time, such as at least one week, at least two weeks, or at least one month. The result(s) of analyzing previous test data using the anomaly detector(s) in previous iterations of steps 404 through 408 can be stored in memory storage device 310, as described further below. In step 410, processor 302 analyzes the results from the current iteration of step 408 and any available previous data stored in memory storage device 310 that corresponds to results generated by the anomaly detector(s) using previous test data in previous iterations of step 408. In particular, step 410 analyzes the result(s) from step 408 and previous data stored in memory storage device 310 to determine whether the anomaly detector(s) have consistently identified the test feature data as a possible anomaly over a one or two week period. In one exemplary embodiment, method 400 is repeated weekly, and step 410 analyzes the result(s) generated from step 408 and previous data stored in memory storage device 310 to determine whether the anomaly detector(s) have consistently identified a possible anomaly over two consecutive iterations of method 400.

[0061] If in step 410, processor 302 determines that the patient's test feature data has not been persistently identified by the anomaly detector(s) as a possible anomaly over the desired period of time, the method proceeds to step 412, where the result(s) from step 408 are stored in memory storage device 310 to be used as results generated using previous test data in the next iteration of method 400. If in step 410, processor 302 determines that the patient's test feature data has been persistently identified by the anomaly detector(s) as likely to correspond to an anomaly over the desired period of time, method 400 proceeds to method 600, where at least one self-report test(s) is administered via computing device 300; more specifically, processor 302 automatically proceeds to method 600, where at least one self-report test(s) is administered to the patient via computing device 300, and further determines, based on the self-report test(s), whether the patient is likely to experience symptoms of a depression relapse.

[0062] Generally, the self-report test(s) require the patient to actively engage with the computing device 300 to answer a series of survey questions. The exemplary method 400 imposes a lower burden on the patient because passive patient data is passively collected from the patient (e.g., by wearing the actigraphy device 200) and the self-report test(s) are administered only after the anomaly detector(s) identify an anomaly using the actigraphy data in step 410. Thus, the method 400 administers the self-report test(s) less frequently compared to such tests administered on a regular (e.g., weekly) schedule. The analysis of step 410 enables the processor 302 to use the passively collected actigraphy data from the patient for its sensitivity in detecting depressive relapse. The additional step of administering the self-report test(s) provides additional specificity in predicting and / or detecting depressive relapse, but is administered less frequently, thereby reducing the burden on the patient for active engagement with the computing device 300 (e.g., answering survey questions). Reducing the burden on the patient promotes patient comfort and compliance with the exemplary method 400 .

[0063] 4 illustrates an exemplary method 600 for administering at least one self-report assessment test via computing device 300 and determining whether a patient is likely to experience symptoms of a depressive relapse. In one exemplary embodiment, method 400 is repeated weekly. In step 602, computing device 300 actively engages with the patient by administering the self-report test(s) to the patient. Specifically, processor 302 directs user interface 308 to display a plurality of survey questions prompting responses from the patient and receives a plurality of inputs from the user via user interface 308 in response to the series of survey questions. The self-report test(s) may include any suitable test having a series of survey questions corresponding to symptoms of depression that prompt the patient to enter a series of responses along a quantitative scale (e.g., a numerical scale rating for each symptom). For example, the self-report test(s) may include survey questions to prompt the patient to enter a series of responses along quantitative scales to assess MDD symptoms, anxiety symptoms, sleep disturbances, anhedonia, energy / motivation, antidepressants, adherence, function / disability, health-related quality of life, pain, self-insight regarding first symptoms that may precede relapse, healthcare utilization, and / or stress / resilience.For example, self-report test(s) may include the Pain Frequency, Intensity and Burden Scale (P-FIBS), Health Resource Use Questionnaire (HRUQ), Recent Life Changes Stress Test (RLCST), Perceived Stress Scale (PSS), Snaith Hamilton Pleasure Scale (SHAPS), WHO Disability Assessment Schedule (WHODAS2.0), EuroQol health state in 5 dimensions and 5 levels (EQ-5D-5L), General Anxiety Disorder 7-Item Scale (GAD-7), MOS Sleep-R, and Patient Adherence to Antidepressant Medication. These may include assessments using the Personal Inventory of Depressive Symptoms Questionnaire (PAQ), Quick Inventory of Depressive Symptoms (QIDS-SR 16), Very Quick Inventory of Depressive Symptoms (VQIDS-SR 5), and the Rothschild Scale for Antidepressant Tachyphylaxis (R-SAT).

[0064] In one embodiment, the self-report test(s) include assessment using the QIDS-SR 16 and / or GAD-7. The QIDS-SR 16 is a patient-reported scale designed to assess the severity of depressive symptoms. The QIDS-SR 16 assesses all criteria symptom domains specified by DSM-5 to diagnose a major depressive episode. Patients respond to each of the 16 items on a 4-point scale, with scores ranging from 0 to 3 for each item. The QIDS-SR 16 scoring system converts responses to the 16 separate items into nine DSM-5 symptom criteria domains, including: 1) sad mood; 2) concentration; 3) self-criticism; 4) suicide attempts; 5) interest; 6) energy / fatigue; 7) sleep disturbances (early, middle, and late insomnia or hypersomnia); 8) appetite or weight loss or gain; and 9) psychomotor agitation or slowing. A total score is obtained by adding the scores for each of the nine symptom domains of the DSM-5 MDD criteria. Four items are used to assess sleep disturbances (early, middle, and late insomnia + hypersomnia); two items are used to assess psychomotor agitation and mental retardation; and four items are used to assess appetite (increased or decreased and weight gain or loss). One item is used to assess the remaining six domains (sad mood, interest, energy / fatigue, self-criticism, concentration, and suicidal ideation). Using a scale of none, mild, moderate, severe, and very severe for depression severity, the corresponding QIDS-SR 16 total scores are none (1-5); mild (6-10); moderate (11-15); severe (16-20); and very severe (21-27). The GAD-7 is a seven-item self-report assessment of depressive symptoms. Each item is scored on a 4-point scale (0-3), with a total score range of 0-21. A GAD-7 score of ≥5 correlates with mild depression. A GAD-7 score ≥ 10 correlates with moderate to severe depression.

[0065] In step 603, processor 302 analyzes data corresponding to multiple inputs from the patient in response to multiple survey questions in the self-report test(s) of step 602 to determine whether the patient is likely to experience symptoms of a depressive relapse, and if so, proceeds to step 610. Specifically, processor 302 may analyze a result score generated based on the multiple inputs from the patient in the self-report test and determine that the patient is likely to experience symptoms of a depressive relapse if the result score is equal to or greater than a first threshold. For example, the first threshold for a QIDS-SR 16 score is 11. In another example, the first threshold for a GAD-7 score is 10.

[0066] If the result score(s) of the self-report test(s) does not meet the first threshold, method 600 proceeds to further analyze data corresponding to the plurality of inputs from the patient in response to the plurality of survey questions of the self-report test(s) with previous data stored in memory storage device 310. In step 604, processor 302 uses the previous data stored in memory storage device 310 to determine whether the self-report test(s) have been administered consecutively to collect data on the patient's behavior within a recent predetermined period of time. The predetermined period may be two weeks or three weeks. In an exemplary embodiment, method 400 is repeated weekly, and processor 302 in step 604 analyzes whether the previous data stored in memory storage device 310 indicates that the self-report test(s) have been administered for three consecutive iterations of method 400 (including the test administered in step 602). If so, method 600 proceeds to step 606. If processor 302 determines that self-reported test(s) have been performed over two consecutive iterations of method 400 (including the test performed in step 602), then method 600 proceeds to step 608. More specifically, if processor 302 determines that self-reported test(s) have been performed for the two most recent iterations of method 400 (including the test performed in step 602), then method 600 proceeds to step 608. Otherwise, method 600 proceeds to step 612.

[0067] In step 606, processor 302 further analyzes the data from step 602 and the previous data stored in memory storage device 310 to determine whether the patient is likely to experience symptoms of a depressive relapse, and if so, proceeds to step 610. Specifically, processor 302 may analyze the result score generated based on multiple inputs from the patient in the self-report test and determine that the patient is likely to experience symptoms of a depressive relapse if processor 302 determines that the current result score and the previous data stored in memory storage device 310 indicate that the patient's self-report test score exceeded a second threshold for at least two weeks during the previous predetermined period (e.g., the last three weeks). The second threshold is lower than the first threshold. In this embodiment, processor 302 may determine that the patient is likely to experience symptoms of a depressive relapse if the result score exceeds the second threshold and the previous data in memory storage device 310 indicates that the result score has increased over at least two consecutive iterations of method 400 (step 610). Additionally or alternatively, if processor 302 determines that the outcome score has exceeded the second threshold at least once within the preceding predetermined period of time and that the outcome score has increased during the preceding predetermined period of time (e.g., an increase in the outcome score of at least one point), indicating a worsening of the patient's depressive symptoms, processor 302 may determine that the patient is likely to experience symptoms of a depressive relapse (step 610). For example, the second threshold for the QIDS-SR 16 score is 9. In another example, the second threshold for the GAD-7 score is 6. In a further example, the second threshold for the GAD-7 score is 5.

[0068] Similar to step 606, in step 608, processor 302 analyzes the data from step 602 and previous data stored in memory storage device 300 to determine whether the patient is likely to experience symptoms of a depressive relapse, and if so, proceeds to step 610. In particular, processor 302 in step 608 may analyze the result score generated based on multiple inputs from the patient in the self-report test and determine that the patient is likely to experience symptoms of a depressive relapse if the result score is greater than or equal to a second threshold, the second threshold being less than the first threshold, and processor 302 determines from previous data stored in memory storage device 310 that previous result scores in a previous predetermined period of time (e.g., the last two weeks) were also above the second threshold. In this embodiment, processor 302 may determine that the patient is likely to experience symptoms of a relapse of depression if the result score is above the second threshold and previous data in memory storage device 310 indicates that the previous result score from the immediately preceding consecutive iteration of method 400 is also above the second threshold (step 610). If these criteria are not met, processor 302 proceeds to step 612. In step 612, the result(s) from step 408 and the result score(s) obtained using the self-report test(s) as described above are stored in storage device 310 to be used as the result generated using the previous test data in the next iteration of method 400.

[0069] In an alternative embodiment, if the criteria of steps 602, 606, and 608 are not met, method 600 can proceed to a further step (not shown) of analyzing the result score of step 608. Specifically, the Relative Change Index (CI) is determined as follows:

[0070]

number

[0071] 5 illustrates an exemplary embodiment of a method 700 for administering two self-report tests via computing device 300 to further determine whether a patient is likely to experience symptoms of a depressive relapse. Method 700 is substantially similar to method 600, except as further described below. Method 700 may be used to administer the two self-report tests to provide two independent scales for assessing whether a patient is likely to experience symptoms of a depressive relapse, providing additional specificity in detecting and / or predicting a patient's relapse into depression. Note that exemplary method 700 illustrated in FIG. 5 may be substituted for method 600 in method 400 for detecting and / or predicting symptoms of a depressive relapse based on a patient's actigraphy data illustrated in FIG. 2.

[0072] In step 702, processor 302 may set a trigger variable in memory storage device corresponding to whether to administer a self-report test in the current iteration of method 400 to a value of "ON." In step 704, computing device 300 engages the patient and administers multiple survey questions for QIDS-SR 16 or GAD-7 in a manner substantially similar to that described above in step 602. In one embodiment, step 704 may administer multiple survey questions for both QIDS-SR 16 and GAD-7. Input obtained pursuant to the questions for each of QIDS-SR 16 and GAD-7 may be analyzed separately by processor 302 in accordance with steps 705-718, as further described below.

[0073] Similar to step 603, in step 705, processor 302 analyzes data corresponding to multiple inputs from the patient in response to multiple survey questions for the QIDS-SR 16 or GAD-7 to determine whether the patient is likely to experience symptoms of a depression relapse, and if so, proceeds to step 716. Specifically, processor 302 may analyze data corresponding to multiple inputs from the patient in response to survey questions for the QIDS-SR 16 to generate a QIDS-SR 16 score. Similarly, processor 302 may analyze data corresponding to multiple inputs from the patient in response to survey questions for the GAD-7 to generate a GAD-7 score. If the QIDS-SR 16 score is 11 or greater, or if the GAD-7 score is 10 or greater, method 700 proceeds to step 716.

[0074] If, at step 705, the result score of the QIDS-SR 16 or GAD-7 test does not meet the aforementioned threshold, method 700 proceeds to further analyze the QIDS-SR 16 or GAD-7 score using previous data stored in memory storage device 310. At step 706, processor 302 determines whether the trigger variable was set to "on" within the most recent two or three consecutive iterations of method 400 (including the current trigger variable of step 702). In particular, method 400 is repeated weekly, and processor 302 analyzes whether the previous data stored in memory storage device 310 includes an "on" value (e.g., "1" indicating the trigger is on, "0" indicating the trigger is off) for the trigger variable for three consecutive iterations of method 400 (including the current trigger variable of step 702). If so, method 700 proceeds to step 708. If the processor 302 determines that the trigger variable has been set to an "on" value for the last two iterations of the method 400 (including the current trigger variable in step 702), the method 700 proceeds to step 712. Otherwise, the method 700 proceeds to step 716.

[0075] Similar to step 606, in step 708, processor 302 analyzes the QIDS-SR 16 scores for the most recent three consecutive iterations of method 400. If processor 302 determines from the current QIDS-SR 16 scores and previous data stored in memory storage device 310 that the QIDS-SR 16 scores over the most recent three consecutive iterations of method 400 were ≧9 for at least two iterations, processor 302 determines that the patient is likely to experience symptoms of a depression relapse and proceeds to step 716. Furthermore, if processor 302 determines from the current QIDS-SR 16 score and previous data stored in memory storage device 310 that at least one QIDS-SR 16 score over the most recent three consecutive iterations is ≧9 and the QIDS-SR 16 score has worsened (e.g., indicated by a one-point increase) over the most recent three consecutive iterations, processor 302 determines that the patient is likely to experience signs of a depressive relapse and proceeds to step 716. Similarly, if processor 302 determines from the current GAD-7 score and previous data stored in memory storage device 310 that the GAD-7 score over the most recent three consecutive iterations of method 400 was ≧6 for at least two iterations, processor 302 determines that the patient is likely to experience signs of a depressive relapse and proceeds to step 716. Furthermore, if processor 302 determines from the current GAD-7 scores and previous data stored in memory storage device 310 that at least one GAD-7 score over the last three consecutive iterations is ≧6 and the QIDS-SR 16 score has worsened (e.g., indicated by a 1-point increase) over the last three consecutive iterations, processor 302 determines that the patient is likely to experience symptoms of a depressive relapse and proceeds to step 716. If neither of these criteria is met, processor 302 proceeds to step 718.

[0076] Similar to step 608, in step 712, processor 302 analyzes the QIDS-SR 16 scores for the most recent two consecutive iterations of method 400. If the current QIDS-SR 16 score (based on input provided in step 704) is ≧9 and processor 302 determines from previous data stored in memory storage device 310 that the previous QIDS-SR 16 score in the immediately preceding iteration of method 400 was also ≧9, processor 302 determines that the patient is likely to experience signs of a depressive relapse and proceeds to step 716. Similarly, if the current GAD-7 score (based on input provided in step 704) is ≧6 and processor 302 determines from previous data stored in memory storage device 310 that the previous GAD-7 score in the immediately preceding iteration of method 400 was also ≧6, processor 302 determines that the patient is likely to experience signs of a depressive relapse and proceeds to step 716. If these criteria are not met, processor 302 proceeds to step 718. Similar to step 612, step 718 stores the result(s) from step 408 and the aforementioned QIDS-SR 16 and / or GAD 7 score(s) in storage device 310 for use as results generated using the previous test data in the next iteration of method 400.

[0077] In both steps 610 and 716, processor 302 determines that the patient is likely to experience symptoms of a depression relapse. Following such a determination by processor 302, processor 302 can instruct or output a signal directing adjustments to treatment for depression. Treatment for depression may include psychotherapy, brain stimulation therapy, or administration of an antidepressant. The antidepressant may be an oral antidepressant, an intranasal antidepressant, or a transdermal antidepressant. Adjustments to psychotherapy may include increasing or decreasing the frequency of sessions and / or the length of time of each session. Adjustments to brain stimulation therapy may include increasing or decreasing the frequency and / or intensity of stimulation interventions. Adjustments to antidepressant administration may include changing the antidepressant regimen (e.g., increasing or decreasing the dose and / or frequency of antidepressant administration), changing the type or class of antidepressant, or adding another antidepressant. In some embodiments, adjustments to the antidepressant regimen may be for non-treatment-resistant patients (e.g., intolerant to oral antidepressant therapy). In other embodiments, adjustments to the antidepressant regimen may be for patients who continue to respond to oral antidepressants and are receiving oral antidepressants. In further embodiments, adjustments to the antidepressant regimen may be for patients who are treatment-resistant to oral antidepressants, and the adjustment is the addition of another antidepressant administered intranasally or transdermally. In another embodiment, after determining that the patient is likely to experience symptoms of a depressive relapse, the processor 302 can instruct or output a signal to instruct adjustments to the administration of other medications that may be suitable for controlling depressive symptoms, such as N-methyl-D-aspartate receptor antagonists, ionotropic glutamate receptor antagonists, and esketamine. The administration adjustments may include increasing or decreasing the dose and / or frequency of the medication. Alternatively, adjustments in dosage may involve substituting agents in place of or in addition to the antidepressant.

[0078] At the end of each iteration of method 400, both step 412 and method 600 (if the patient is not identified as having experienced a relapse) proceed to step 414. In step 404, the training actigraphy data is updated to include the most recently collected test actigraphy data as part of the updated training actigraphy data used to retrain at least one anomaly detector in method 500. In an exemplary embodiment, the updated training actigraphy data incorporates the most recently collected test actigraphy data and excludes training actigraphy data obtained more than a predetermined recent time period. For example, if the training actigraphy data is obtained over a three-month period, the most recently collected test actigraphy data is incorporated into the updated training actigraphy data and data older than three months from the updated training actigraphy data.

[0079] FIG. 7 shows an exemplary timeline 900 that a patient may experience during the exemplary method 400 of FIG. 2 for detecting and / or predicting a patient's depression relapse. The patient may begin method 400 at time 904 during an initial visit with a mental health provider 901. The patient may be provided with an actigraphy device 200 and a computing device 300 for remote assessment 902, where data is collected from the patient while they are away from the mental health provider 901 during daytime activities from time 904 to time 912. In an exemplary embodiment, the actigraphy device 200 may be worn by the patient at all times (e.g., 24 hours a day, 7 days a week) to collect the patient's ongoing actigraphy data from time 904 to time 912. The computing device 300 collects self-reported data entered by the patient, such as a self-reported assessment of sleep adequacy. The self-reported assessment may be entered in response to questions regarding the patient's perception of sleep or may include questions from the MOS-S scale assessment of sleep adequacy. In the exemplary timeline 900, self-report assessments are obtained weekly. Although Figure 7 shows that the remote assessment 902 includes the collection of actigraphy data, it is contemplated that the remote assessment 902 may also include the collection of other types of passive patient data.

[0080] Between time 904 and time 906 (which may be, for example, a three-month period), actigraphy device 200 and computing device 300 collect data via remote assessment 902 while the patient does not experience a relapse (as shown in white in FIG. 7 ). At time 906, the patient may visit with mental health provider 901 to confirm that the patient has not experienced a relapse between time 904 and time 906. If the patient has not experienced a relapse, the data collected between time 904 and time 906 may be used to train an anomaly detector. The anomaly detector is used to screen newly obtained remote assessment data from the patient to determine whether the new data is anomalous compared to previously collected remote assessment data.

[0081] Between time 906 and the patient's next visit with the mental health provider 901 at time 910, the remote assessment 902 continues to collect data from the patient and iteratively analyzes the newly collected data using the trained anomaly detector to determine whether the new data is anomalous. With each iteration of method 400 (an example of which is shown as a flag in FIG. 7 ), the anomaly detector is updated using the most recent three months of remote assessment data as training data. In the exemplary timeline 900 shown in FIG. 7 , each white flag 908 reflects an iteration of method 400 in which new remote assessment data is analyzed using the anomaly detector and determined to be not anomalous. The period between time 906 and time 910 may be continuously repeated as long as the new remote assessment data is not determined to be anomalous by the anomaly detector and the mental health provider 901 confirms at time 910 that the patient is not experiencing a relapse.

[0082] As shown in exemplary timeline 900, between time 910 and the patient's next scheduled visit with mental health provider 901, remote assessment 902 continues to collect data from the patient and iteratively analyzes the newly collected data using the trained anomaly detector to determine whether the new data is anomalous. However, the anomaly detector initially determines the new remote assessment data to be non-anomalous (shown as white flag 908), but subsequently detects anomalous new data in subsequent iterations (shown as black flag 911). If an anomaly is detected, further data collection is triggered. For example, if an anomaly is detected, computing device 300 may proceed to method 600 or method 700 to administer at least one self-report test to the patient and further determine whether the patient is likely to experience symptoms of a depression relapse based on the self-report test(s).

[0083] 7 , method 600 or method 700 may determine, based on self-report test(s), that the patient has likely experienced signs of a depressive relapse and may instruct the patient to preemptively visit with mental health provider 901 at time 912 prior to the patient's next scheduled visit with mental health provider 901 (not shown). During the preemptive visit at time 912, mental health provider 901 may confirm that the patient has experienced a relapse (shown in black) and may then provide the patient with early clinical intervention to stabilize their depressive symptoms. Once the patient has stabilized and their depressive symptoms are in remission, as shown at time 914 (shown in white), the patient may resume method 400, returning to time 904 of the exemplary timeline 900.

[0084] FIG. 8 illustrates another exemplary method 1000 for detecting and / or predicting a patient's depression relapse. Exemplary method 1000 is similar to method 400, except as otherwise described below. Exemplary method 1000 utilizes multiple features extracted from actigraphy data and at least one machine-learning anomaly detector that trains an anomaly detector based on the patient's own historical data (n=1, where n is the total number of sampled individuals) to determine possible signs of depression relapse in the patient. A patient's relapse into depression can be identified using one or more tests, as described above with respect to method 400. If computing device 300 determines that the patient has relapsed, exemplary method 1000 ends. Method 1000 can be resumed in a manner similar to method 400 when the patient returns to depression remission. Thus, method 1000 can be continuously used to train one or more anomaly detectors on all non-relapse actigraphy data recorded from the patient, and subsequently continuously analyze and detect anomalies in the recorded data. In one example, the anomaly detector(s) are continuously trained on features extracted from actigraphy data obtained during non-relapse periods through successive iterations of method 1000. The portion of method 1000 for detecting anomalies in actigraphy data recorded from a patient provides an objective assessment of the patient to determine whether the actigraphy data is anomalous by comparing it to actigraphy data previously collected from the patient during periods when the patient is not known to have developed depressive symptom(s) and / or depressive relapse. Once an anomaly case is identified, a self-report symptomatology questionnaire algorithm (SRSQA) is used to confirm the relapse signature. The self-report symptomatology questionnaire or SRSQA may be substantially similar to the self-report test described above with respect to methods 600 or 700 of method 400.

[0085] In step 1002, training data is collected by device 200 and / or computing device 300 over a predetermined training period. The training data includes actigraphy data collected by device 200 over the predetermined training period. Device 200 may be worn by the patient to detect the patient's movements and generate training actigraphy data over the predetermined training period in a manner similar to that described above in step 402. Similar to method 400, method 1000 also ends if computing device 300 determines that the patient has relapsed. Thus, training data is collected during a non-relapse period.

[0086] In step 1004, the training data obtained in step 1002 is used to train at least one anomaly detector for determining possible signs of depression relapse, e.g., in a manner similar to method 500, unless otherwise noted below. Similar to step 502, the training data in step 1004 is analyzed to extract features from the training data to generate training feature data. For example, the training data includes training actigraphy data. Various features are extracted from the training actigraphy data and included in the training feature data. In particular, the features extracted from the training actigraphy data may include monofractal patterns and / or multifractal dynamics of the training actigraphy data and may further include sample entropy (SaEn) over different time frames.

[0087] For example, features extracted from the training actigraphy data include fractal patterns determined using the DFA method described above. In this exemplary method 1000, to ensure reliable estimation of F(n) at time scale n, the DFA method determines fractal patterns in the actigraphy data using actigraphy data from the most recent consecutive activity days (at least two consecutive days) with no gaps of more than 72 minutes (5% of the 1440 minutes of activity counts per day) for each day. In this exemplary method 1000, the variability index (α) of the DFA method is determined at two different time scales, α1 between 10 minutes (i.e., 10 data points with an epoch length of 1 minute) and 90 minutes, and α2 between 120 minutes and 600 minutes, respectively, to capture different activity dynamics regions. Features extracted from the training actigraphy data also include the multifractal dynamics of the actigraphy data using the MFDFA method. In this exemplary method 1300, the multifractal dynamics are determined using the exemplary MFDFA method described above, where the variance of the data is generally expressed as:

[0088]

number

[0089] Additionally, features extracted from the training actigraphy data included the sample entropy (SaEn) of the training actigraphy data for each day and for each of the four periods: morning (6 AM - 12 PM), afternoon (12 PM - 6 PM), evening (6 PM - 12 AM), and night (12 AM - 6 AM), determined as described above.

[0090] The extracted training feature data may be filtered in the same manner as in step 504, and processor 302 may then modify the filtered training feature data by imputing missing data points to the generated imputed training data in a manner similar to step 506. In one example, missing data is imputed with a value of 0. The imputed training data may be centered and scaled using the population mean and standard deviation (data mean(population data) / standard deviation(population data)) in the same manner as in step 508. A power transformation (e.g., a Yeo-Johnson transformation or a Box-Cox transformation) may also be applied to all features of the imputed training data to generate modified training data with a normalized distribution.

[0091] The modified training data is used to train at least one machine learning anomaly detector to determine possible signs of depression relapse. In one example, the anomaly detector models normal behavior based on the training feature data and uses the model's prediction error to identify anomalies. Specifically, the anomaly detector is the Enc-Dec AD described above. The Enc-Dec AD's encoder is trained using the training feature data to learn a vector representation of the actigraphy data time series. The modified training data is used to determine the mean (μ) and standard deviation (Σ) of the normal distribution N(μ,Σ) of the time series using maximum likelihood estimation. μ and Σ are then used by the Enc-Dec AD to determine an anomaly score to assess whether an anomaly has been detected.

[0092] In step 1006, device 200 acquires test data from the patient, similar to step 404 described above. The test data includes test actigraphy data from the patient. The test data may be generated in real time or over a desired test period, e.g., w days, such as 14 days. In some embodiments, the entire w-day period may be generated in step 1006. In other embodiments, the test data may include a portion of previously generated data and new data obtained in step 1006 over a desired data collection period (e.g., y days). The portion of previously collected data may be for a period of wy days. For example, the test data may include 13 days of previously generated data and 1 day of new data. Processor 302 analyzes and extracts features from the test data in a manner similar to that described above with respect to step 1004 to generate test feature data.

[0093] During step 1006, computing device 300 may also administer self-report assessment test(s) via computing device 300 to collect self-report test data from the patient. For example, self-report tests may be administered weekly during a desired testing period. Thus, self-report tests need not be performed with each iteration of method 1000. The self-report test(s) may include survey questions prompting the patient to enter a series of responses along a quantitative scale (e.g., a numerical scale rating for each symptom), as described above with respect to methods 400 and 600. For example, the self-report test(s) may include survey questions prompting the patient to enter a series of responses along a quantitative scale for assessing depressive symptoms, mood-related cognitions, energy / motivation, anhedonia, pain, healthcare utilization, stress / resilience, function / disability, health-related quality of life, anxiety, and / or sleep disturbances. In one embodiment, the self-report test(s) administered during step 1006 include assessments using the VQIDS-SR 5 and / or GAD-7. The VQIDS-SR 5 is a patient-reported measure designed to assess the severity of depressive symptoms. It assesses core depression domains extracted from the QIDS-SR 16 to rapidly identify major depressive episodes: sad mood, self-perception, engagement, fatigue, and psychomotor retardation. The total VQIDS-SR 5 score is obtained by adding the scores for each of the five depression domains. A VQIDS-SR 5 score of ≥ 5 correlates with mild depression. A VQIDS-SR 5 score of ≥ 6 correlates with moderate to severe depression. The GAD-7 is as described above.

[0094] In step 1008, processor 302 analyzes the test feature data from step 1006 using the anomaly detector(s) trained in step 1002. In particular, the anomaly detector(s) compare the test feature data to the training feature data to generate an anomaly score that quantifies the likelihood that the training feature data is anomalous during the test period. In one particular embodiment, the anomaly detector is an Enc-Dec AD trained in step 1004. The decoder of the Enc-Dec AD generates predicted outputs for time frames of the test feature data based on vector representations learned from the training feature data. The Enc-Dec AD generates an anomaly score a (i) a (i) =(e (i) -μ) T Σ -1 (e (i) −μ), where μ and Σ are as determined in step 1004 using the training feature data;

[0095]

number

[0096]

number

[0097] The anomaly scores generated in step 1008 can be analyzed using a dynamic, data-driven threshold to identify instances of anomalies. FIG. 9 illustrates an exemplary method 1050 for identifying anomalies using a dynamic threshold. Method 1050 is an unsupervised anomaly scoring method. In one example, given a time series of anomaly scores generated based on test feature data, method 1050 identifies anomalies based on a w-day window. The w-day window is a w ={a(p+1:p+w)}∀p=0~(Nw)+1 in steps of one day, where N is the total number of days of actigraphy data recorded from the patient. In some examples, N may be the sum of the number of days of training and test data collected from the patient. The window may span any suitable number of days, for example, w may be 1 to 30 days. In particular, w is the same value as the desired test period mentioned above in step 1006. In a preferred embodiment, w=14 days. For each iteration over the w-day window, the past anomaly scores, i.e., a, are used to determine a data-driven threshold for anomaly detection. 全て ={a(1:T c )}∀T c = step of one day from w to N (a w The historical anomaly score (a 全て ) may include or consist of anomaly scores generated based on the training feature data and the test feature data.

[0098] In step 1052, a 全て The non-anomalous section of

[0099]

number

[0100]

number

[0101] In step 1054, a w Potential anomalies within a w A second anomaly threshold above which an anomaly score may indicate a potential anomaly.

[0102]

number

[0103]

number

[0104]

number

[0105]

number

[0106]

number

[0107]

number

[0108]

number

[0109]

number

[0110]

number

[0111]

number

[0112]

number

[0113]

number

[0114] In step 1056, the potential anomalous cases identified in step 1054 are pruned to identify the cases that are most likely to be anomalous. w Potential abnormal cases in a w This is the non-anomalous section of

[0115]

number

[0116]

number

[0117]

number

[0118] 10a-10f show example timelines of anomaly scores for actigraphy data collected during an exemplary time period from a patient, where the most recently acquired 14-day period of actigraphy data is analyzed to identify anomaly cases according to steps 1052-1056. FIG. 10a shows an example timeline of actigraphy data collected from days 338 through 367 after study initiation. In this exemplary embodiment, the anomaly scores for days 338 through 354 are generated from training feature data based on actigraphy data collected during the time period from days 338 through 354 (excluding day 354). The anomaly scores for days 354 through 367 are generated from test feature data based on actigraphy data collected during the time period from days 354 through 367. The shaded area shown in FIG. 10a represents the anomaly scores in this exemplary embodiment. 全て In this exemplary embodiment, w is 14 days. So in this example, this set of anomaly scores has a 14 As can be seen from Figure 10a, a 14 over the period from day 354 to day 367 (shown by cross-hatching), a 全て This example time series ε 全て is shown as a horizontal dotted line across Figure 10a. As can be seen, the anomaly score on day 367 is 全て Therefore, the anomaly score on day 367 is

[0119]

number

[0120]

number

[0121]

number

[0122]

number

[0123] In step 1010, computing device 30 analyzes the self-report test data corresponding to and / or within a desired period of time following any anomaly identified by method 1050 to determine whether the anomaly indicates that the patient is likely to experience symptoms of a relapse of depression. For each anomaly, the corresponding self-report test data or self-report data within a desired period of time following the anomaly is analyzed to determine whether the anomaly indicates that the patient is likely to experience symptoms of a relapse of depression. In particular, step 1008 determines whether the anomaly was identified by method 1050 within the first w days or w / 2 days of the test characteristic data.

[0124] In one example, w is a 14-day period, and the anomalies further analyzed in step 1008 are identified based on the presence of anomaly cases within the most recent seven days of the 14-day period. Method 1000 considers whether to collect self-reported test data in step 1006 in this example on a weekly (i.e., every seven days) basis. For each week in which an anomaly is identified by method 1050, self-reported test data for the identified week is collected in step 1006. Additionally, for each week in which an anomaly was identified in the previous week by method 1000 (i.e., one week after the identified anomaly week), self-reported test data is also collected in step 1006. The self-reported test data from both of these weeks is analyzed to determine whether the anomaly indicates that the patient is likely to experience symptoms of a depressive relapse. Thus, the self-reported test data for the week most recent in which the anomaly was identified and the week following the week in which the anomaly was identified is analyzed to determine whether the anomaly indicates that the patient is likely to experience symptoms of a depressive relapse. The self-report test data includes responses to assessments using the VQIDS-SR 5 and / or GAD-7. Specifically, the self-report test data includes responses to assessments using the VQIDS-SR 5 and GAD-7, which combined utilize a total of 12 questions. The self-report test is administered to the patient if an abnormality is detected in the identified week and may also be administered to the patient in weeks following the week in which the abnormality was identified. In one example, if an abnormality is detected in the identified week, the self-report test is administered to the patient, and if the self-report data includes assessments (e.g., VQIDS-SR 5 and / or GAD-7 scores) above a set of high predetermined threshold(s) (e.g., VQIDS-SR 5≧6 and / or GAD-7≧10), method 1000 determines that the patient is likely to experience symptoms of a depressive relapse. However, if the self-report test data includes ratings above a set of lower predetermined threshold(s) (e.g., VQIDS-SR 5≧5 and / or GAD-7≧5), method 1000 continues to repeat, obtaining additional self-report test data one week after the week in which the abnormality was identified. If the self-report test data includes ratings above the set of lower predetermined threshold(s) (e.g., VQIDS-SR 5≧5 and / or GAD-7≧5) for two consecutive weeks, method 1000 determines that the patient is likely to experience symptoms of a relapse of depression.Method 1000 does not require the administration of self-report testing during other weeks, thus reducing the burden on the patient being monitored by repeated repetitions of method 1000. Furthermore, this example method 1000 identifies anomalous cases that indicate a patient is likely to experience symptoms of a depressive relapse after confirmation using self-report test data (specifically, two weeks of self-report test data), thus reducing the instances of false positives compared to detecting anomalies in test data (including test actigraphy data) alone or identifying relapse based on weekly collection of self-report test data alone.

[0125] A patient may be determined to be likely to experience symptoms of depressive relapse if the VQIDS-SR 5 score is 6 or greater either in the week in which the abnormality was identified or the week following the week in which the abnormality was identified. Alternatively, a patient may be determined to be likely to experience symptoms of depressive relapse if the GAD-7 score is 10 or greater either in the week in which the abnormality was identified or the week following the week in which the abnormality was identified. In another example, a patient may be determined to be likely to experience symptoms of depressive relapse if the VQIDS-SR 5 score is ≧5 and / or the GAD-7 score is ≧5 for the week in which the abnormality was identified and the week following the week in which the abnormality was identified. Following processor 302's determination that the patient is likely to experience symptoms of depressive relapse, processor 302 may instruct or output a signal directing adjustments to treatment for depression. Treatment for depression may include psychotherapy, brain stimulation therapy, or administration of an antidepressant, as described above.

[0126] At the end of each iteration of method 1000, in step 1012, the training actigraphy data is updated to include the most recently collected test actigraphy data as part of the updated training actigraphy data used to retrain at least one anomaly detector in step 1004. In an exemplary embodiment, the updated training actigraphy data incorporates the most recently collected test actigraphy data and excludes training actigraphy data obtained more recently than a predetermined period of time. Similar to method 400, exemplary method 1000 may be repeated continuously or at a desired frequency, e.g., daily, weekly, biweekly, etc. In a particular embodiment, exemplary method 1000 is repeated daily. In one example, exemplary method 1000 is repeated every y days (y is less than or equal to the length of w days in method 1050). Notably, y is the same value as the desired data collection period described above in step 1006. In one example, y is 1 day and w is 14 days.

[0127] Although method 1000 for detecting and / or predicting relapse of depression in a patient has been described above, it is contemplated that method 1000 can be modified to detect and / or predict relapse of other neurological disorders, particularly those in which changes in activity patterns (recorded by actigraphy data) indicate relapse of such neurological disorders, such as schizophrenia and bipolar disorder. In particular, the self-report test data collected during step 1006 and analyzed in step 1010 may be modified. The self-report test data may be collected from self-report test(s) for quantitatively assessing symptoms of other neurological disorders. For example, a self-report test for schizophrenia may be the Symptoms of Schizophrenia (SOS) Inventory. As another example, a self-report test for bipolar disorder may be the Hypomanic Scale, the Mood Disorder Questionnaire, the Menfi, Pisa, Paris, and San Diego Temperament Assessment-Automated Questionnaire Version, the Bipolar Spectrum Diagnostic Scale, the General Behavioral Inventory, and the Hypomania Checklist.

[0128] FIG. 11 illustrates an example implementation of method 1000 over a sample time series of data collected from days 57 through 381. In particular, FIG. 11 includes a sample time series of MADRS scores 1102, corresponding abnormality scores 1104 generated from actigraphy collected from patients analyzed according to Enc-Dec AD described in step 1008, and corresponding VQIDS-SR 5 scores 1110 and GAD-7 scores 1112 based on self-reports collected from the patients. The time series labeled 1106 represents weekly assessments of the abnormality scores 1104, with short bars indicating that the abnormality score for that week does not include an abnormality according to method 1050 and tall bars indicating that the abnormality score for that week includes a potential abnormal instance identified by method 1050. The time series labeled 1108 indicates as tall bars the weeks in which self-reported test data is analyzed according to step 1008. The timeline shown as 1114 illustrates a timeline determination of whether the patient is likely to experience symptoms of a depressive relapse. Tall bars indicate that the patient is likely to experience symptoms of a depressive relapse. Short bars indicate that self-reported test data has been analyzed, and as can be seen from timeline 1114, step 1008 identifies that by day 381, depression is likely to relapse and the patient's MADRS score will rise above 22, indicating a depressive relapse.

[0129] FIG. 12 shows an exemplary timeline 1200 that a patient of the exemplary method 1000 of FIG. 10 may experience for detecting and / or predicting a patient's depression relapse. The patient may begin method 1000 at time 1204 during an initial visit to a mental health provider 1201 equipped with an actigraphy device 200, from time 1204 to time 1212, and actigraphy data is collected from the patient during daily activities between visits with the mental health provider 1201. In one exemplary embodiment, the actigraphy device 200 may be worn by the patient at all times (e.g., 24 hours a day, 7 days a week) to collect the patient's ongoing actigraphy data from time 1204 to time 1212. While FIG. 12 shows that the remote assessment 1202 includes the collection of actigraphy data, it is contemplated that the remote assessment 1202 may also include the collection of other types of passive patient data. Between time 1204 and time 1206 (which may be, for example, a two-month period), device 200 collects data via remote assessment 1202 during which the patient does not experience a relapse as training data (as shown in white in FIG. 12 ). At time 1206, the patient may visit mental health provider 1201 to confirm that the patient has not experienced a relapse between time 1204 and time 1206. If the patient has not experienced a relapse, the data collected between time 1204 and time 1206 may be used to train an anomaly detector (e.g., the Enc-Dec AD described above). The anomaly detector is used to analyze newly obtained remote assessment data from the patient to determine whether the new data indicates that the patient is likely to experience symptoms of a depression relapse.

[0130] Between time 1206 and the patient's next visit with the mental health provider 1201 at time 1210, the remote assessment 1202 continues to collect data from the patient. The remote assessment 1202 also includes a computing device 300 that collects self-report data entered by the patient, as described above in step 1006. In the exemplary timeline 1200, self-report assessments are obtained weekly. The computing device 300 iteratively analyzes newly collected actigraphy and self-report data according to method 1000 to determine whether the new data indicates that the patient is likely to experience symptoms of a depressive relapse. With each iteration of method 1000 (an example of which is shown as a flag in FIG. 12 ), the anomaly detector is updated using data including the most recent remote assessment data as training data. In the exemplary timeline shown in FIG. 12 , each white flag 1208 reflects an iteration of method 1000 in which new remote assessment data is analyzed using method 1000 and determined not to indicate that the patient is likely to experience symptoms of a depressive relapse. The period between time 1206 and time 1210 may be repeated continuously as long as new remote assessment data is not determined to indicate that the patient is likely to experience symptoms of a relapse of depression and mental health provider 1201 confirms at time 1210 that the patient is not experiencing a relapse.

[0131] As shown in exemplary timeline 1200, between time 1210 and the patient's next scheduled visit with mental health provider 1201, remote assessment 1202 continues to collect data from the patient and iteratively analyzes the newly collected data using method 1000 to determine whether the patient is likely to experience symptoms of a depressive relapse. However, as shown in the example of FIG. 12, the anomaly detector determines new remote assessment data (shown as white flag 1208) that indicates the patient is not likely to experience symptoms of a depressive relapse, but in a subsequent iteration subsequently detects that the patient is likely to experience symptoms of a depressive relapse (shown as black flag 1211).

[0132] When it is determined that the patient has likely experienced signs of a depressive relapse, the computing device 300 instructs the patient to preemptively visit with the mental health provider 1201 at time 1212 prior to the patient's next scheduled visit (not shown) with the mental health provider 1201. During the preemptive visit at time 1212, the mental health provider 1201 may confirm that the patient has experienced a relapse (shown in black) and may then provide the patient with early clinical intervention to stabilize their depressive symptoms. Once the patient has stabilized and their depressive symptoms are in remission, as shown at time 1214 (shown in white), the patient may resume method 1000 and return to time 1204 of the exemplary timeline 1200.

[0133] Patients at risk of relapse can be continuously monitored according to the above-described method 1000. Actigraphy data is collected continuously, and the patient's relapse status is determined by a mental health provider during regularly scheduled (e.g., bimonthly) visits. At each scheduled visit, if the mental health provider determines that the patient has not relapsed, all of the actigraphy data collected up to the visit is used to train an anomaly detector, specifically, the Enc-Dec AD. The trained model is used to detect anomalous cases in subsequent clinical visit activity data. These anomalous cases are pruned, and the remaining cases are then confirmed for relapse risk by self-reported symptom assessments. This process continues with each non-relapse visit to the mental health provider, and the Enc-Dec AD is retrained. The retrained Enc-Dec AD is used to identify anomalies in subsequently acquired actigraphy data. If method 100 determines that the patient is likely to experience symptoms of a depressive relapse, the computing device 300 can instruct the patient to contact a mental health provider. Alternatively, the computing device 300 can send an alert to the patient's mental health care provider, who can then follow up with the patient upon receiving the alert, for example, by scheduling an appointment or calling the mental health care provider from their office to check the patient's symptoms. Based on the mental health care provider's assessment, early intervention can be performed as needed, ultimately preventing impending relapse and resulting in better patient outcomes.

[0134] Those skilled in the art will appreciate that the exemplary embodiments described herein may be implemented in any number of ways, such as as separate software modules or as a combination of hardware and software. For example, the exemplary methods may be embodied in one or more programs stored on a non-transitory storage medium and including lines of code that, when compiled, may be executed by one or more processor cores or separate processors. A system according to one embodiment comprises multiple processor cores and a set of instructions that execute on the multiple processor cores to perform the exemplary methods described above. The processor cores or separate processors may be incorporated into or communicate with any suitable electronic device, such as an on-board processing device within the device or a processing device external to the device, such as a mobile computing device, smartphone, computing tablet, computing device, or the like, that can communicate with at least a portion of the device. [Example]

[0135] Example I In Example I, a patient is provided with an exemplary actigraphy device 200 to collect actigraphy data and determine the patient's depression relapse according to the exemplary methods 400 and 700 described above. Additionally, the patient is asked to provide a daily quantitative assessment of their sleep perception in response to two self-report assessment questions: (1) whether they feel rested; and (2) whether they feel they have had enough sleep. Input from these self-report assessment questions is used to generate training self-report data included in the training data used to train the anomaly detector used in Example I. The patient in Example I met DSM-5 diagnostic criteria for non-psychotic recurrent MDD within the previous 24 months and was taking oral antidepressants, but did not meet criteria for a major depressive episode at the start of actigraphy monitoring with actigraphy device 200. As shown in FIG. 13 , training actigraphy data was collected over a three-month period and updated as method 400 was repeated weekly. The MADRS test was administered to patients by a mental health provider approximately every 8 weeks in addition to collecting daily actigraphy data and repeating Method 400 weekly to provide a test separate from the patient's actigraphy to identify whether the patient experienced a relapse of depression. This separate test was determined to have relapsed based on an initial MADRS score of ≥ 22 and, at a subsequent follow-up visit with the mental health provider, the mental health provider determining that the patient's CGI-S score increased by 2 from the baseline obtained before starting Method 400 (i.e., on or before Day 1), or a change in medication type or medication dose within 14 days of the patient's first MADRS total score of ≥ 22. For this test, patients were also considered to have relapsed regardless of MADRS score if they were hospitalized for worsening depression, had an intentional suicide attempt or suicidal behavior, or were otherwise indicated by the mental health provider to have relapsed.This separate test also determines that a patient is non-relapsed after an initial MADRS score of ≥ 22, and at a subsequent follow-up visit with a mental health provider, the MADRS score decreases to less than 22, the CGI-S score does not change by more than 2 from baseline, and the patient's medication also remains unchanged within 14 days from the time the patient first experienced a MADRS total score of ≥ 22. As shown in Figure 13 and listed in Table 2 below, the patient of Example I was determined by this separate test to be non-relapsed from days 297 to 332 and to be relapsed from days 339 to 381. Note that because the MADRS test is administered at 8-week intervals, the patient's relapse is detected using the MADRS scale on day 381, but it is possible that the patient may have relapsed shortly after the previous MADRS test was administered from days 339 to 381, as shown in Figure 13 and Table 3.

[0136] Example I utilizes processor 302 of computing device 300 to analyze a patient's actigraphy data according to the exemplary method 400 for detecting signs of depressive relapse based on the patient's actigraphy data, described above, repeated weekly. Example I utilizes two separate anomaly detectors, a one-class SVM anomaly detector and an IF anomaly detector, to determine the likelihood of relapse using the actigraphy data. As shown in Table 2, processor 302 using one SVM anomaly detector reports a value of 0 if it determines that the test data is consistent with the training data (e.g., unlikely to correspond to relapse) and reports a value of 1 if it determines that the test data is anomalous (e.g., likely to correspond to relapse). Computing device 300 using the IF anomaly detector reports an anomaly score corresponding to the likelihood that the test data corresponds to an anomaly. Additionally, Example I utilizes method 700 to administer two self-report tests via computing device 300 to further determine whether the patient is likely to have experienced signs of depressive relapse. Specifically, Example I administers the QIDS-SR 16 and / or GAD-7 tests and analyzes current and previous results from the last three weeks to determine whether a patient is at risk for relapse. In this example, data from the last three weeks can be analyzed as shown in Table 2 below.

[0137] [Table 2]

[0138] Note that in Table 2, the number of weeks of available test data refers to the number of most recent consecutive weeks for which QIDS-SR 16 and / or GAD-7 scores are available within the most recent 3 weeks of data. The QIDS-SR 16 and GAD-7 scores for Example I are reported in Table 3 below.

[0139] [Table 3]

[0140] As shown in Table 3 above, Example I demonstrated that actigraphy data collected from patients and analyzed by a computing device identified patients as at risk for relapse by day 381 based on the patient's actigraphy data and scores obtained using the QIDS-SR 16 and GAD-7 tests. This identification was within seven days of when MADRS analysis, as described above, would indicate the patient had experienced a depressive relapse. This data demonstrates that the use of patient actigraphy data in Example I identifies depressive relapse within seven days of its occurrence, thereby providing early identification of relapse and enabling an earlier ability to direct changes to the patient's treatment in response to the relapse.

[0141] Example II In Example II, exemplary actigraphy devices 200 were provided to 41 patients who subsequently relapsed. Data for Example II was collected in a manner similar to that described above for Example I. The actigraphy devices 200 collected actigraphy data, and the computing device 300 collected the patients' quantitative assessments of their sleep perceptions in response to the two self-report assessment questions described above in Example I. The data was analyzed by the processor 302 of the computing device 300, and an SVM anomaly detector was used to detect and / or predict signs of depression relapse in the patients, following the same methodology described above in Example I. Performance metrics for Example II are shown below in Table 4.

[0142] [Table 4]

[0143] Figure 14 shows data corresponding to the proportion of a group of 30 true positive patients who were determined to have likely relapsed over a period of time prior to the actual onset of relapse, as determined by the MADRS method described in Example I. The data in Figure 14 show that using actigraphy data 7 days or more earlier than the actual onset of relapse, 83% of the 30 true positive patients were identified as likely to have relapsed.

[0144] Example III In Example III, the exemplary actigraphy device 200 was provided to approximately 330 subjects, and approximately 88 subjects relapsed. As described above in Example I, the actigraphy device 300 collected actigraphy data and self-reported training data, and the data was analyzed by the computing device 400 to determine the patient's relapse to depression according to the same method as described above in Example I.

[0145] Figure 15a shows data corresponding to the frequency of patients receiving self-reported testing across various trigger rates, with the self-reported surveys administered to patients when determined by actigraphy data, as described above in Example I. The data shown in Figure 9a have a two-sample Kolmogorov-Smirnov goodness-of-fit hypothesis test P-value of ≦0.001, indicating a distributional difference in trigger rate distributions. Figure 9b shows data corresponding to the frequency of patients receiving self-reported testing at various trigger rates, with the self-reported testing administered weekly. As seen in Figure 9a, a greater frequency of patients receiving self-reported testing as directed by actigraphy exhibited lower trigger rates (e.g., 0.2) compared to Figure 9b, where the greatest frequency was a trigger rate of 1.0. The data shown in Figures 9a and 9b demonstrate that self-reported testing is less frequently administered when such testing is triggered by the patient's actigraphy data, thus reducing patient burden and promoting patient compliance. Figure 10a shows a subset of the data shown in Figure 9a for a time frame in which the patient was in remission and not approaching relapse. Similarly, Figure 10b shows a subset of the data shown in Figure 9b for a time frame in which the patient was in remission and not approaching relapse. Figure 11a shows a subset of the data shown in Figure 9a for a time frame in which the patient was approaching relapse. Similarly, Figure 11b shows a subset of the data shown in Figure 9b for a time frame in which the patient was approaching relapse. Figure 12 shows performance metrics for the method of assessing depression relapse using actigraphy data in Example III. In Figure 12, ACC stands for accuracy, SEN stands for sensitivity, SPEC stands for specificity, BAC stands for balanced accuracy, PPV stands for positive predictive value, and NPV stands for negative predictive value.

[0146] Example IV In Example IV, exemplary actigraphy devices 200 were provided to 41 patients who subsequently relapsed. The data for Example IV was collected in a manner similar to that described above for Example I. The actigraphy device 200 collected actigraphy data, and the computing device 300 collected the patient's quantitative assessment of sleep perception in response to the two self-report assessment questions described above for Example I. The data was analyzed by the processor 302 of the computing device 300 to detect and / or predict signs of depressive relapse in the patient. Example IV utilizes a method similar to that described above for Example I, except as noted below, using an SVM anomaly detector, repeated weekly. Example IV utilizes method 600 to administer the same two self-report tests as Example I via the computing device 300 and further determines whether the patient is likely to have experienced signs of depressive relapse by analyzing current and previous results from the most recent three weeks to determine whether the patient is at risk for relapse, as shown in Table 5 below.

[0147] [Table 5]

[0148] Note that in Table 5, the number of weeks of available test data refers to the number of most recent consecutive weeks for which QIDS-SR 16 and / or GAD-7 scores are available within the most recent three weeks of data. The performance indicators for Example IV are shown in Table 6 below.

[0149] [Table 6]

[0150] Figure 8 shows data corresponding to the proportion of the group of 30 true positive patients who were determined to have likely relapsed over a period prior to actual signs of relapse, as determined by the MADRS method described in Example I. The data in Figure 8 show that 83% of the 30 true positive patients were identified as likely to have relapsed using actigraphy data 7 days or more earlier than actual signs of relapse.

[0151] Example V In Example V, the method 1000 of FIG. 8 , which uses Enc-Dec AD to identify abnormalities based on a 14-day (w=14 days) window, was evaluated using data collected from 211 MDD subjects. The data from the 211 MDD subjects included longitudinal assessments of self-reported symptoms (measured weekly, biweekly, etc.) and continuously collected actigraphy data for over a year or until the first episode of depressive relapse. The subjects also attended regular bimonthly clinical visits. Each subject completed self-report assessments using a smartphone during and between clinical visits at regular intervals (e.g., weekly to bimonthly). Actigraphy data was continuously collected from each subject using a device worn on the non-dominant wrist, such as the Philips Actiwatch by Philips Respironics, which measures acceleration in a direction parallel to the surface of the device with continuous sampling at 32 Hz. The raw acceleration data recorded by the device were aggregated into 15-second epochs of counts (actigraphy data) reflecting movement amplitude. To minimize epochs of zero activity counts, a fine-resolution activity count data stream was created by summing 15-second epochs within each minute to obtain 1440 activity count data points per day.

[0152] Each clinical visit was labeled as a recurrence if the patient met any one of the criteria listed below. (1) The MADRS total score at the time of the study visit was 22 or higher, and a worsening of symptoms was confirmed over an interval of approximately 1 to 2 weeks. (2) If a subject receives 22 or more MADRS assessments at a study visit (scheduled or unscheduled), a further visit (i.e., recurrence verification visit) will be scheduled within 1 to 2 weeks. Subjects with a MADRS score of 22 or greater at the recurrence verification visit are considered to have relapsed. A CGI-S change from baseline ≥ 2 at the relapse validation visit or a medication change occurring within ± 14 days of the study visit will be considered a relapse. (3) Hospitalization due to worsening depression (4) Intentional suicide attempt or suicidal behavior (5) Decisions by mental health professionals

[0153] If a clinical visit did not meet the above criteria, it was labeled as non-relapse. Labeling was assumed to persist until the day after the previous clinical visit. Among 211 MDD subjects, a total of 1190 visits (1140 non-relapse visits and 50 relapse visits) corresponding to 50 relapse subjects and 161 non-relapse subjects were evaluated. The first clinical visit after starting actigraphy data collection for all subjects was used to train Enc-Dec AD for each subject and therefore was not included in the 1190 visits for evaluation. Furthermore, the last visit evaluated for a relapse subject was the relapse visit. Therefore, the evaluation in Example V was up to the first relapse, and multiple relapses were not considered in this example.

[0154] The distribution of the 211 MDD subjects according to each of the relapse criteria identified above is provided in Table 7.

[0155] [Table 7]

[0156] The characteristics of the 211 MDD subjects are summarized in Table 8 below.

[0157] [Table 8]

[0158] Additionally, Figures 19a and 19b show all available visits for relapse and non-relapse subjects, respectively.

[0159] Actigraphy features were analyzed to extract DFA, MFDFA, and sample entropy features. These features are believed to be less affected by uncontrolled daily schedules and environmental conditions, which may provide an objective assessment of the complexity of circadian supraperiodic rhythms and activity patterns. In Example V, DFA features were determined using a second-order polynomial function to trend the actigraphy data and eliminate the influence of possible linear trends in the data. To ensure reliable estimation of F(n) at time scale n, the most recent consecutive activity days (at least two consecutive days) with no gaps greater than 72 minutes (5% of the 1440 minutes of activity counts per day) were used for each day. The variability index (α) was calculated at two different time scales to capture different regions of activity dynamics: α1 between 10 minutes (i.e., 10 data points with a 1-minute epoch length) and α2 between 120 minutes and 600 minutes, respectively. MFDFA features are extracted using values ​​of q varying from -5 to 5 in increments of 0.1, and s varying from 10 to 600 minutes. Sample entropy features are extracted for each day and for each of four periods: morning (6 AM to 12 PM), afternoon (12 PM to 6 PM), evening (6 PM to 12 AM), and night (12 AM to 6 AM). More specifically, SaEn is determined as the median SaEn value (in z-scored activity counts) across the last seven activity days (with no identified gaps) of the activity count.

[0160] All actigraphy features were calculated for each day on the last 7 consecutive days of activity counts, with a minimum requirement of at least 2 consecutive days of activity counts. Activity counts were further subjected to signal quality checks to detect the following: (i) isolated large spikes with amplitudes 10 standard deviations away from the global mean level, and (ii) sequences of zeros with durations exceeding 60 minutes. Identified data points or segments were labeled as gaps and handled appropriately in feature calculations.

[0161] The Enc-Dec AD is trained using features extracted from actigraphy data from an initial period before the first bimonthly clinical visit. The Enc-Dec AD of Example V is identified according to the parameters listed in Table 9.

[0162] [Table 9]

[0163] If the patient was determined not to have relapsed to depression at the first bimonthly clinical visit, subsequently collected actigraphy data was analyzed according to method 1000 using the trained Enc-Dec AD to identify abnormalities. This process continued with each non-relapse visit to the clinic, with the Enc-Dec AD being retrained at each visit to indicate that the patient had not relapsed to depression. The retrained Enc-Dec AD was used to identify abnormalities in data collected for subsequent visits. Any identified abnormalities were further evaluated according to step 1008, evaluating the self-report symptom questionnaires, VQIDS-SR 5 and GAD-7, collected simultaneously with the identified abnormality, as a further confirmatory step to determine whether the abnormality indicates that the patient is likely to experience symptoms of a depressive relapse. In Example V, after analyzing data from the self-report symptom questionnaires during the one week in which the abnormality was detected and the one week following the abnormality, the abnormality is determined to indicate that the patient is likely to experience symptoms of a depressive relapse. If data from the self-report symptom questionnaire is unavailable during the week in which the abnormality is detected or during the week following the abnormality, the abnormality is not determined to be indicative of a likely symptom of a relapse of depression.

[0164] 20 shows exemplary timelines of three different clinical visits illustrating that the Enc-Dec AD of Example V is successively trained with actigraphy data at each non-relapse visit, and abnormalities are predicted in the actigraphy data of subsequent visits. The top timeline 2002 shows patients with at least one relapse-free visit, with the previous actigraphy data used as training data for analyzing the test data. As shown in the middle timeline 2004, the test data from the previous visit (shown in timeline 2002) becomes the training data for the visit shown in timeline 2002. Similarly, in the bottom timeline 2006, the test data from the previous visits (shown in timelines 2002 and 2004) becomes the training data for the visit shown in timeline 2006. For each relapse and non-relapse subject, the process continues until the first relapse, which is the last clinical visit for the relapsed patient in Example V.

[0165] The performance indicators for determining non-recurrence versus recurrence in Example V are shown in Table 10 below.

[0166] [Table 10]

[0167] As shown in Table 10 above, Example V achieved a sensitivity of 0.66, a specificity of 0.82, and a balanced accuracy of 0.74 in predicting recurrence. The observed prevalence of recurrence was 4.2%, and Example V achieved a positive predictive value of 0.14 and a negative predictive value of 0.98. Example V also achieved an overall false alarm rate (FAR) of 0.18 among recurring and non-recurring subjects (FAR of 0.28 for recurring subjects and FAR of 0.16 for non-recurring subjects).

[0168] As shown in Table 10 above, Example V has a false alarm rate of 0.18 across relapse and non-relapse subjects. The false alarm rate in relapse subjects was 0.28 and the false alarm rate in non-relapse subjects was 0.16, demonstrating the ability of Example V to detect and / or predict depressive relapse more frequently in relapse subjects, and therefore demonstrating that a determination by Example V of a high likelihood of experiencing depressive relapse may ultimately result in a relapse event.

[0169] As shown in Table 10, the performance metrics for determining non-relapse versus relapse in Example V are compared to performance metrics for detecting abnormalities in test actigraphy data alone or identifying relapse based on weekly collection of self-report test data alone.

[0170] [Table 11]

[0171] As seen in Table 11, Example V provides comparable sensitivity while significantly increasing specificity compared to anomaly detection alone or self-reported test data alone. In particular, Example V provides a significant reduction in FAR compared to weekly collection of only anomaly detection or self-reported test data. Example V provides a reduction in FAR that could not be achieved with either of these components alone. As shown in Table 11, Example V, which utilizes a specific time-aligned combination of anomaly detection and self-reported test data, provides an unexpectedly greater (or synergistic) reduction in FAR than the combination of two separate analyses of anomaly detection and self-reported test data. This data suggests that for any subject at risk for relapse based on baseline assessment, a positive prediction by the framework will most likely result in eventual relapse unless early intervention is initiated. The method of Example V can identify patients likely to experience symptoms of depressive relapse a median of 21 days before the onset of depressive relapse, providing a window of opportunity for adjusting depression treatment before the onset of depressive relapse.

[0172] Considering the reduction in FAR, the method of Example V also significantly reduces provider burden compared to anomaly detection alone or self-reported test data alone. As seen in Table 12 below:

[0173] [Table 12] * Weekly Rating Total=8237; ** Total bimonthly visits = 1190

[0174] Table 12 shows provider burden, represented by the percentage of scheduled bimonthly preemptive visits corresponding to a determination that a patient is likely to experience symptoms of depression relapse based on anomaly detection alone, self-reported test data alone, or the methods of Example V. In particular, Example V provides a significant reduction in provider burden compared to weekly collection of anomaly detection or self-reported test data alone. Example V provides a reduction in provider burden that could not be achieved by either of these components alone. As shown in Table 12, Example V, which utilizes a specific time-aligned combination of anomaly detection and self-reported test data, unexpectedly reduces provider burden to a greater extent (or synergistically) than the combination of two separate analyses of anomaly detection and self-reported test data.

[0175] Table 12 also shows patient burden, represented by the percentage of total weekly scheduled self-report assessments that correspond to a determination that the self-report data will be analyzed (i.e., used in the analysis) in determining whether the patient is likely to experience symptoms of relapse. For weekly self-report assessments scheduled after a determination that the patient is likely to experience symptoms of relapse, these assessments are not part of the percentage representing patient burden. This data is collected across a study population of 211 subjects who were followed for a year or more. As shown in Table 12, selectively administering the self-report test according to Example V also significantly reduces patient burden, thereby providing an effective method of monitoring patients that is sufficiently sensitive, has a low FAR, and is low burden on both the patient and the provider.

[0176] The invention described and claimed herein is not limited in scope by the specific embodiments disclosed herein, which are intended to illustrate certain aspects of the invention. Any equivalent embodiments are intended to be within the scope of the invention. Indeed, various modifications of the invention in addition to those shown and described herein will become apparent to those skilled in the art from the foregoing description. Such modifications are also intended to fall within the scope of the appended claims. All publications cited herein are incorporated by reference in their entirety.

Claims

1. 1. A computer-implemented method for detecting or predicting relapse of depression in a patient, comprising: (i) acquiring training data of the patient over a training period from a wearable device worn by the patient, the training data including training actigraphy data corresponding to the patient's movements over the training period, the training period being a period during which the patient is not experiencing signs of a return of depression; (ii) training an anomaly detector using the training data, the anomaly detector configured to perform supervised and / or unsupervised learning and identify deviations from the training data by building a profile of normal cases using modified training data and identifying data that deviate from the normal profile as anomalies; (iii) acquiring test data for the patient from the wearable device during a test period after the training period, the test data including test actigraphy data corresponding to movements of the patient after the training period; (iv) extracting a plurality of features from the test data to generate test feature data, the features corresponding to indices related to at least one of activity, sleep, circadian rhythm, and multifractal dynamics; (v) analyzing the test feature data using the anomaly detector and comparing the test feature data with the training data; (vi) if the anomaly detector determines that the test feature data is likely abnormal compared to the training actigraphy data, administering a self-report test to the patient to obtain a plurality of inputs from the patient; (vii) analyzing the plurality of inputs from the patient to determine whether the patient is likely to experience symptoms of a return of depression; 20. A computer-implemented method comprising:

2. 10. The method of claim 1, further comprising: (viii) updating the training data to include the test data and repeating steps (ii) through (vii) until the patient is determined to have reverted to depression.

3. 3. The method of claim 2, wherein steps (ii) to (vii) are repeated weekly.

4. 4. The method of claim 1, wherein the training data further comprises data corresponding to self-reported characteristics of the patient's physical behavior over the training period, and the testing data further comprises data corresponding to self-reported characteristics of the patient's physical behavior during the testing period.

5. Step (vi) presenting a plurality of self-report survey questions to the patient via a user interface; receiving the plurality of inputs from the patient via the user interface in response to the self-report survey questions; The method according to any one of claims 1 to 4, comprising:

6. Step (vii) is analyzing the plurality of inputs to generate a result score for the self-reported test; comparing the result score to at least one threshold to determine whether the patient is likely to experience symptoms of a return of depression; The method according to any one of claims 1 to 5, comprising:

7. The method of any one of claims 1 to 6, wherein the anomaly detector is a one-class support vector machine module.

8. The method of any one of claims 1 to 6, wherein the anomaly detector is an isolation forest module.

9. The method of any one of claims 1 to 8, wherein the training period is at least 3 months.

10. 6. The method of claim 5, wherein the plurality of self-report survey questions correspond to symptoms of depression and the plurality of inputs from the patient correspond to a rating on a numerical scale for each symptom.

11. 11. The method of any one of claims 1 to 10, further comprising adjusting the dose of antidepressant administered to the patient when it is determined that the patient is likely to experience symptoms of a return of depression.

12. 11. The method of any one of claims 1 to 10, further comprising the step of increasing the dose of antidepressant administered to the patient when it is determined that the patient is likely to experience symptoms of a return of depression.

13. 1. A system for detecting or predicting a return of depression in a patient, comprising: a wearable device comprising at least one accelerometer configured to detect movement of the patient, the wearable device configured to generate actigraphy data corresponding to the patient's movement; a computing device operably connected to the wearable device to receive actigraphy data from the wearable device, the computing device comprising: a user interface for displaying output and receiving input from the patient; a processor, and a non-transitory computer-readable storage medium including a set of instructions executable by the processor, the set of instructions comprising: acquiring, from the wearable device, training actigraphy data corresponding to the patient's movements over a training period, the training period being a period during which the patient is not experiencing signs of a return of depression; training an anomaly detector using training data including the training actigraphy data, the anomaly detector configured to perform supervised and / or unsupervised learning and identify deviations from the training data by constructing a profile of normal cases using modified training data and identifying data that deviate from the normal profile as anomalies; acquiring test actigraphy data from the wearable device corresponding to the patient's movements after the training period; extracting a plurality of features from the test actigraphy data to generate test feature data, the features corresponding to indices of at least one of activity, sleep, circadian rhythm, and multifractal dynamics; analyzing the test feature data using the anomaly detector and comparing the test feature data with the training data; directing the user interface to display a plurality of self-report survey questions to the patient; receiving a plurality of inputs from the patient via the user interface in response to the self-report survey questions; and analyzing the plurality of inputs from the patient to determine whether the patient is likely to experience symptoms of a return of depression.

14. The system of claim 13 , wherein the wearable device is configured, in an operational configuration, to be worn around a wrist of the patient.

15. 15. The system of claim 13 or 14, wherein the user interface is a touch screen.

16. The system of any one of claims 13 to 15, wherein the computing device is selected from the group consisting of a mobile computing device, a smartphone, and a computing tablet.

17. The system of any one of claims 13 to 16, wherein the anomaly detector is a one-class support vector machine module.

18. The system of any one of claims 13 to 16, wherein the anomaly detector is an isolation forest module.

19. 19. The system of any one of claims 13-18, wherein the plurality of self-report survey questions correspond to symptoms of depression, and the plurality of inputs from the patient correspond to a rating of each corresponding symptom on a numerical scale.

20. 20. The system of any one of claims 13-19, wherein the set of instructions further comprises instructions operable to direct an output indicating an adjustment of a dose of antidepressant administered to the patient when the computing device determines that the patient is likely to experience symptoms of a return of depression.

21. 1. A computer-implemented method for detecting or predicting relapse of depression in a patient, comprising: (i) acquiring training data of the patient over a training period from a wearable device worn by the patient, the training data including training actigraphy data corresponding to the patient's movements over the training period, the training period being a period during which the patient is not experiencing signs of a return of depression; (ii) training an anomaly detector using the training data, the anomaly detector configured to perform supervised and / or unsupervised learning and identify deviations from the training data by building a profile of normal cases using modified training data and identifying data that deviate from the normal profile as anomalies; (iii) acquiring test data for the patient from the wearable device during a test period, at least a portion of the test period being after the training period, the test data including test actigraphy data corresponding to movements of the patient after the training period; (iv) extracting a plurality of features from the test data to generate test feature data, the features corresponding to indices related to at least one of monofractal patterns, multifractal dynamics, and sample entropy; (v) analyzing the test feature data using the anomaly detector to compare the test feature data with the training data and detect anomalies in the test feature data; (vi) analyzing self-report test data to determine whether the patient is likely to experience symptoms of a return of depression when an abnormality is detected in the test characteristic data, the self-report test data being generated from a plurality of inputs from the patient in response to a self-report test; 20. A computer-implemented method comprising:

22. 22. The method of claim 21, wherein the self-reported test is collected from a time contemporaneous with the detected abnormality.

23. 23. The method of claim 21 or 22, wherein the self-reported test is collected from the patient after an abnormality is detected.

24. 24. The method of claim 21, further comprising the step of: (vii) updating the training data to include the test data and repeating steps (ii) to (vi) until the patient is determined to have reverted to depression.

25. 25. The method of claim 24, wherein steps (ii) through (vii) are continuously repeated until the patient is determined to have reverted to depression.

26. Step (vi) analyzing the self-reported test data to generate a result score for the self-reported test; comparing said result score to at least one threshold to determine whether the patient is likely to experience symptoms of a return of depression; The method of any one of claims 21 to 25, comprising:

27. 27. The method of any one of claims 21 to 26, wherein the anomaly detector utilizes a long short term memory (LSTM) neural network, the anomaly detector comprising an encoder and a decoder.

28. Step (v) identifying non-anomalous sections of the test feature data using a first anomaly threshold; determining potential anomalous instances in the test feature data using a second anomaly threshold, the second anomaly threshold being determined based on the non-anomalous sections; pruning the potential anomaly cases based on a percent reduction for each potential anomaly case to identify the anomalies in the test feature data; 28. The method of claim 27, comprising:

29. The method of any one of claims 21 to 28, wherein the training period is at least 14 days.

30. 30. The method of any one of claims 21-29, further comprising adjusting the dose of antidepressant administered to the patient when it is determined that the patient is likely to experience symptoms of a return of depression.

31. 31. The method of any one of claims 21 to 30, further comprising increasing the dose of antidepressant administered to the patient when it is determined that the patient is likely to experience symptoms of a return of depression.

32. 1. A system for detecting or predicting a return of depression in a patient, comprising: a wearable device comprising at least one accelerometer configured to detect movement of the patient, the wearable device configured to generate actigraphy data corresponding to the patient's movement; a computing device operably connected to the wearable device and configured to receive actigraphy data from the wearable device, the computing device comprising: a user interface for displaying output and receiving input from the patient; a processor, and a non-transitory computer-readable storage medium including a set of instructions executable by the processor, the set of instructions comprising: acquiring, from the wearable device, training actigraphy data corresponding to the patient's movements over a training period, the training period being a period during which the patient is not experiencing signs of a return of depression; training an anomaly detector using training data including the training actigraphy data, the anomaly detector configured to perform supervised and / or unsupervised learning and identify deviations from the training data by constructing a profile of normal cases using modified training data and identifying data that deviate from the normal profile as anomalies; acquiring, from the wearable device, test actigraphy data corresponding to the patient's movements during a test period, at least a portion of the test period being after the training period; and extracting a plurality of features from the test actigraphy data to generate test feature data, the features corresponding to at least one activity-related measure of at least one of monofractal patterns, multifractal dynamics, and sample entropy; analyzing the test feature data using the anomaly detector to compare the test feature data with the training data to detect anomalies in the test feature data; and analyzing self-report test data to determine whether the patient is likely to experience symptoms of a return of depression when an abnormality is detected in the test characteristic data; The system, wherein the self-report test data is generated from a plurality of inputs received from the patient by the user interface in response to a self-report test comprising a plurality of self-report survey questions displayed on the user interface.

33. 33. The system of claim 32, wherein the wearable device, in an operational configuration, is configured to be worn around a wrist of the patient.

34. 34. The system of claim 32 or 33, wherein the user interface is a touch screen.

35. The system of any one of claims 32 to 34, wherein the computing device is selected from the group consisting of a mobile computing device, a smartphone, and a computing tablet.

36. 36. The system of any one of claims 32 to 35, wherein the anomaly detector utilizes a long short term memory (LSTM) neural network, and wherein the anomaly detector comprises an encoder and a decoder.

37. 37. The system of any one of claims 32-36, wherein the plurality of self-report survey questions correspond to symptoms of depression, and the plurality of inputs from the patient correspond to a rating of each corresponding symptom on a numerical scale.

Images