WO2022008694A1

WO2022008694A1 - Method for determining probability of a mood disorder

Info

Publication number: WO2022008694A1
Application number: PCT/EP2021/069079
Authority: WO
Inventors: Olga Matveeva; Evgeny Orlov
Original assignee: Individual
Current assignee: Individual
Priority date: 2020-07-08
Filing date: 2021-07-08
Publication date: 2022-01-13
Anticipated expiration: 2023-01-08

Abstract

The present invention relates to methods for determining a score indicative of the probability of a subject to develop a mood disorder the method comprising the steps of obtaining at least one parameter indicative of the state of mood of the subject, wherein for the parameter(s) at least two data points are obtained at differing points in time; weighting at least one data point of the parameter(s) obtained in the previous step upon comparing data point(s) with a predefined pattern to provide a weighted set of data points of the parameter(s); comparing the weighted set of data points of the parameter(s) of the previous step with a reference signature obtained from at least one reference subject, wherein the at least one reference subject has not previously been diagnosed with a mood disorder; determining a score indicative of the probability of the subject to develop a mood disorder based on the comparison obtained in the previous step. The invention furthermore relates to a computer implemented program e.g. using deep learning and may be implemented in a mobile device, such as a wearable. Further, self-reported data and data from tests can be combined with the parameters in some embodiments of the method.

Description

Method for determining probability of a mood disorder

Consequences of mood disorders are associated with high health burdens and financial burdens for individuals and for society (Simon, Gregory E. "Social and economic burden of mood disorders." Biological psychiatry 54.3 (2003): 208-215.). Reliable and accurate scoring of mood disorders is of key importance for prediction, diagnosis, prevention and treatment of mood disorders. For many mood disorders prevention or early interventions are more effective than late interventions (Chanen, Andrew M., Michael Berk, and Katherine Thompson. "Integrating early intervention for borderline personality disorder and mood disorders." Flarvard review of psychiatry 24.5 (2016): 330-341.; Conus, Ph. "First episode of mood disorders: an opportunity for early intervention in bipolar disorders." L'Encephale 36 (2010): S71-6.). Scoring mood disorders is a field of research that has been trying to reduce biases and involved time intense procedures since a long time (Gavazzi, Stephen M., Mary A. Fristad, and Julie C. Law. "The understanding mood disorders questionnaire." Psychological reports 81.1 (1997): 172-174.; Choi, Bernard C K, and Anita W P Pak. “A catalog of biases in questionnaires.” Preventing chronic disease vol.2,1 (2005): A13.). Mood disorders are commonly scored using questionnaires (American Psychiatric Association (2013). Diagnostic and Statianual (5th ed.). Arlington, VA: American Psychiatric Association). These questionnaires capture data from a time period from the memory of a subject. The data is prone to be distorted by the state of mood during the questionnaire, the subjective memory access, and the subjective memory interpretation of the subject (Choi, Bernard C K, and Anita W P Pak. “A catalog of biases in questionnaires.” Preventing chronic disease vol.2,1 (2005): A13.). Further, subject background data, such as sociodemographic data and/or BMI, correlate with the probability to develop certain mood disorders (Choi MR, Eun HJ, Yoo TP, et al. The effects of sociodemographic factors on psychiatric diagnosis. Psychiatry Investig. 2012;9(3):199-208; Simon GE, Von Korff M, Saunders K, et al. Association between obesity and psychiatric disorders in the US adult population. Arch Gen Psychiatry.2006;63(7):824-830). Cognitive function, such as prefrontal cortex function has been reported to be altered during mood disorders (Koenigs, Michael, and Jordan Grafman. “The functional neuroanatomy of depression: distinct roles for ventromedial and dorsolateral prefrontal cortex.” Behavioural brain research vol.201,2 (2009): 239-43.; George, M.S., Ketter, T.A. and Post, R.M. (1994), Prefrontal cortex dysfunction in clinical depression. Depression, 2: 59-72.). More recent methods integrate the analysis of autonomic nervous system biomarkers to detect aspects of mood disorders. WO/2019/209986 A1 describes a wearable/portable sensor and computing technology, to continuously monitor heartbeat data to automatically detect high-stress episodes. US 2020/0054262 A1 discloses a stress analysis method using bio-signals and determination of usefulness using a deep neural network algorithm to determine a stress level value. US 2019/9239791 A1 relates to a system and method for monitoring and predicting the mental health of a person.

Current methods are not suitable to detect chronic patterns, obtain and/or process data for efficiently, specifically and accurately determining a score indicative of the probability of a subject to develop a mood disorder.

There is thus a need for reliable and efficient means and methods for obtaining and processing data for determining a score indicative of the probability of a subject to develop a mood disorder, such as burnout, generalized anxiety disorder and/or depression.

The above technical problem is solved by the embodiments provided herein and as characterized in the claims.

Accordingly, the invention relates to, inter alia, the following embodiments:

1. A method for determining a score indicative of the probability of a subject to develop a mood disorder, the method comprising the steps of:

(a) obtaining at least one parameter indicative of the state of mood of the subject, wherein for the parameter(s) at least two data points are obtained at differing points in time;

(b) weighting at least one data point of the parameter(s) obtained in a) upon comparing data point(s) with a predefined pattern to provide a weighted set of data points of the parameter(s);

(c) comparing the weighted set of data points of the parameter(s) of b) with a reference signature obtained from at least two reference subjects, wherein at least one of the reference subject(s) has not previously been diagnosed with a mood disorder;

(d) determining a score indicative of the probability of the subject to develop a mood disorder based on the comparison obtained in (c).

2. The method of embodiment 1 , wherein at least two data points in step (a) differ at least two weeks in time of obtainment. The method of embodiment 1 or 2, wherein the method comprises a step of distinguishing between short-term events and long-term events. The method of embodiment 3, wherein the step of distinguishing between shortterm events and long-term events comprises thresholding. The method of embodiment 1 to 4, wherein the parameters comprise autonomic nervous system biomarkers obtained by at least one heartbeat sensor, optical sensor, accelerometer, gyroscope, skin conductance sensor, temperature sensor, sonic sensor and/or microphone. The method of embodiment 5, wherein the parameters additionally comprise self-reported subject background data. The method of embodiment 6, wherein the self-reported subject background data comprises sociodemographic data and/or BMI. The method of embodiment 5 to 7, wherein the parameters comprise at least one interpreted sensor dataset obtained by combination of autonomic nervous system biomarkers and/or by weighting of at least one data point of autonomic nervous system biomarkers. The method of embodiment 8, wherein at least one of the interpreted sensor datasets is indicative of heart rate, heart rate variability, skin conductance, finger temperature, respiratory rate, sleep quality, insomnia, circadian sleep rhythm, phone usage, ambient sounds, bedtime and wake time, sleep duration, sleep quality, cognitive skill and/or facial expression . The method of embodiment one of the previous embodiments, wherein the parameters comprise at least one test that determines a score indicative of cognitive skill. The method of embodiment 10, wherein at least one test is indicative of working memory, executive function, attention, and/or cognitive flexibility. The method of embodiments 11, wherein cognitive flexibility is indicative of prefrontal brain network dysfunction. The method of embodiment any one of the previous embodiments, wherein the parameters comprise at least one standardized test. The method of embodiment any one of the previous embodiments, wherein the parameters comprise the Wisconsin card sorting test and/or the Stroop test. The method of embodiment any one of the previous embodiments, wherein at least one test is adapted for usage on at least one mobile device. The method of any one of the previous embodiments, wherein at least one parameter is obtained from at least one sensor comprised in at least one wearable device. The method of embodiment 16, wherein the wearable device is a smartwatch and/or a fitness tracker. The method of any one of the previous embodiments, wherein at least two data points in at least one parameter differ in time of obtainment at least: i) 1 week for cognitive skill ii) 12 hours for heart rate variability, iii) and/or 24 hours for sleep. The method of any one of the previous embodiments, wherein at least one step comprises the use of a computer implemented method. The method of embodiment 19, wherein step (a) comprises the use of an application programming interface (API) to obtain parameters. The method of embodiment 19 to 20, comprising oversampling techniques in particular the oversampling techniques SMOTE, ADASYN. The method of embodiment 19 to 21, wherein the computer implemented method comprises the use of machine-learning. The method of embodiment 19 to 22, wherein the generation of the predefined pattern comprises using a machine learning technique. The method of embodiment 22 or 23, wherein the machine learning technique comprises the use of for Logistic regression, CART, Bagging, Random Forest, Gradient Boosting, Linear Discriminant Analysis, Gaussian Process Classifier and/or Gaussian NB for classification. The method of embodiment 22, 23 or 24, wherein the machine learning technique, wherein the machine learning technique comprises the use of Linear, Lasso, Ridge, ElasticNet, KNN, DecisionTree, SVR, AdaBoost, GradientBoost, Random Forest, and/or ExtraTrees for regression. The method of embodiment 22, 23, 24 or 25, wherein the machine learning technique comprises a deep learning technique, in particular simple multilayer perceptron (MLP) and/or convolutional neural networks for classification and/or regression. The method of embodiment 22 to 26, wherein the machine learning technique used for generation of the predefined pattern comprises the use of a score indicative of the probability of the subject to develop a mood disorder. The method of any one of the previous embodiments, wherein the reference signature obtained from at least one reference subject is based on a standardized mood disorder questionnaire score. The method of embodiment 28, wherein the standardized mood disorder questionnaire score comprises the composite Maslach Burnout Inventory score, and/or the GAD-7 score, and/or the PHQ-9 score. The method of any one of the previous embodiments, wherein the method further comprises administering to the subject having a score indicative for a mood disorder a pharmaceutically effective amount of a pharmaceutical product for the treatment of a mood disorder. A pharmaceutical product for use in the treatment of a mood disorder in a subject, wherein the subject has a score indicative for a mood disorder determined according to any one of the embodiments 1 to 29. A method for monitoring a mood disorder the method comprising the steps of:

1 ) determining a score indicative of the probability of a subject to develop a mood disorder using the method of any one of the embodiments 1 to 29 at a first timepoint; 2) determining a score indicative of the probability of a subject to develop a mood disorder using the method of any one of the embodiments 1 to 29 at a second timepoint;

3) comparing the score of step (1 ) with the score of step (2); and

4) monitoring the mood disorder in the subject based on the comparison of step (3). A library comprising a score indicative of the probability of the subject to develop a mood disorder determined according to any one of the embodiments 1 to 29. A storage device comprising computer-readable program instructions to execute the method according to any one of the embodiments 1 to 29 or 32, preferably additionally comprising the library according to embodiment 33. A system for determining a score indicative for a mood disorder comprising at least one sensor for obtainment of at least one parameter indicative for a mood disorder and the storage device of embodiment 34. A server comprising the storage device of embodiment 34, at least one processing device, and a network connection for receiving at least one parameter indicative of the state of mood of the subject and/or at least one weighted set of data points of the parameter(s). The method of any one of the embodiments 1 to 30 or 32, the pharmaceutical product for use according to embodiment 31 , the storage device of embodiment 34, the system of embodiment 35, the server of embodiment 36 or the library according to embodiment 33, wherein the mood disorder is at least one disorder selected from the group of burnout, generalized anxiety disorder, and depression. The method of embodiment 37, the storage device of embodiment 37, the pharmaceutical product for use according to embodiment 37, the system of embodiment 37, the server of embodiment 37 or the library according to embodiment 37, wherein the mood disorder is burnout. Accordingly, the invention relates to a method for determining a score indicative of the probability of a subject to develop a mood disorder, the method comprising the steps of:

(c) comparing the weighted set of data points of the parameter(s) of b) with a reference signature obtained from at least two reference subjects, wherein at least one of the reference subjects has not previously been diagnosed with a mood disorder;

The step of obtaining (step (a)) a parameter does not refer to obtaining from the human body, but rather comprises obtaining the parameter from a sensor and/or retrieving it from a data source.

The weighting (step (b)) and comparing (step (c)) can be achieved by any method known in the art (see e.g. Example). In some embodiments, the weighting (step (b)) described herein comprises machine-learning based weighting. In some embodiments, the comparing (step (c)) described herein is based on a machinelearning technique and/or a thresholding technique. That is, the present invention is based, at least in part, on the surprising discovery that specific weighting and/or excluding of data points according to a reference pattern can improve accuracy of a method for determining a score indicative of the probability of a subject to develop a mood disorder. The weighted or excluded data points may be data points with altered relevance as an indicator for mood disorders and/or can be at least one data point that is misleading as an indicator for mood disorders. The inventors found that reference data from healthy reference subjects can be used e.g. to determine thresholds and exclude healthy background parameter variations (see example). Therefore, the determined score can be particularly accurate, when the reference subjects comprise healthy subjects, preferably healthy subjects and non-healthy subjects (e.g. healthy subjects and subjects having a mood disorder). As demonstrated herein, the method of the invention allows a subject and/or healthcare professional to more accurately and objectively evaluate a subject's mental health and/or diagnose a mood disorder such as burnout, generalized anxiety disorder or depression. The method includes and excludes and/or weights data points in an objective, unbiased, manner to unveil patterns that would otherwise be unperceivable to a subject and/or healthcare practitioner.

In certain embodiments, for example wherein the score is indicative for an exhaustion- related mood disorder, such as burnout, changes induced by stressful events assumed to be unrelated to the mood disorder, such as traffic-induced stress, can be excluded or given a low weight in the parameter(s). The method of the invention provides accurate results by specifically reducing the assigned weight of a data point upon comparing data point(s) with a predefined pattern to provide a weighted set of data points of the parameter(s).

Weighting at least one data point of the parameter(s) obtained in the method of the invention upon comparing data point(s) with a predefined pattern to provide a weighted set of data points of the parameter(s) allows to factor in that some of the data points of the parameter(s) are misleading, of less relevance, and/or of no relevance for determining a score indicative of the probability of the subject to develop a mood disorder. Subsequently, the signal to noise ratio of the method can be improved by this step and resources can be allocated to processing data promising a certain relevance for determining a score indicative of the probability of the subject to develop a mood disorder.

Therefore, the method of the invention in certain embodiments weights at least one data point of the parameter(s) obtained in the methods of the invention upon comparing data point(s) with a predefined pattern to provide a weighted set of data points of the parameter(s) in an unbiased and standardized manner. This results in a weighted set of data points of the parameter(s) with high relevance for determining a score indicative of the probability of a subject to develop a mood disorder and allows the invention to have particular high accuracy, reproducibility, reliability and/or specificity. In certain embodiments, the invention relates to the method of the invention, wherein at least two data points in step (a) differ at least two weeks in time of obtainment.

The inventors found, that data obtained from a subject over a period of at least 14, 15, 16, 17, 18, 19, 20, 21 days, 4 weeks, 5 weeks, 6 weeks, 7 weeks or 8 weeks provide signals that enable to differentiate between patterns relevant for the accurate prediction of a mood disorder such as resilience/exhaustion.

The data points can be obtained with regular intervals or with interruptions. However, it is relevant for the invention that at least two data points have a difference of at least 14, 15, 16, 17, 18, 19, 20, 21 days, 4 weeks, 5 weeks, 6 weeks, 7 weeks or 8 weeks in time of obtainment, preferably a substantially complete dataset of at least 14, 15, 16, 17, 18, 19, 20, 21 days, 4 weeks, 5 weeks, 6 weeks, 7 weeks or 8 weeks is obtained. The data points that differ at least 14 days are preferably obtained from the same source (e.g. the same sensor).

In certain embodiments, the invention relates to the method of the invention, wherein the method comprises, wherein an at least 14, 15, 16, 17, 18, 19, 20, 21 days, 4 weeks, 5 weeks, 6 weeks, 7 weeks or 8 weeks in time window is analyzed, preferably wherein the maximal difference of time of obtainment between data points is longer than the time window and wherein the time window is a sliding time window.

Accordingly, the invention is at least in part based on the surprising finding, that data obtained over a certain period of time is particularly useful for the accuracy for the determination of a score indicative for a mood disorder.

In certain embodiments, the invention relates to the method of the invention, wherein the method comprises a step of distinguishing between short-term and long-term events.

The term “short-term event”, as used herein, refers to data points of a parameter and/or at least one weighted set of data points that are indicative of an event (e.g. a stress reaction) that is resolved in a short time period.

The term “long-term event”, as used herein, refers to data points of a parameter and/or at least one weighted set of data points that are indicative of an event (e.g. a stress reaction) that is not resolved in a short time period. In some embodiments, the step of distinguishing between short-term and long-term events is part of the weighting step of the method of the invention. In some embodiments, the step of distinguishing between short-term and long-term events is part of the comparison step of the method of the invention. In some embodiments, the step of distinguishing between short-term and long-term events is part of the weighting step and of the comparison step of the method of the invention.

In certain embodiments of the invention, the weight of at least two data points of the parameter in (b) is the lower, the higher the value of deviation from the predefined pattern divided by the difference in time of obtainment in (a).

In certain embodiments, the invention relates to the method of the invention, wherein the step of distinguishing between short-term events and long-term events comprises thresholding.

The long-term pattern is not necessarily a consecutive time period, but can also refer to several episodes within a consecutive time period. In some embodiments, the event is resolved in a short time period as described herein, if less than a certain proportion of the data points of a parameter and/or the at least one weighted set of data points during a certain time period (e.g. at least 14, 15, 16, 17, 18, 19, 20, 21 days, 4 weeks,

5 weeks, 6 weeks, 7 weeks or 8 weeks) are below a threshold. In some embodiments, the event is resolved in a short time period as described herein, if less than 70% of the data points of a parameter and/or the at least one weighted set of data points during a certain time period (e.g. at least 14, 15, 16, 17, 18, 19, 20, 21 days, 4 weeks, 5 weeks,

6 weeks, 7 weeks or 8 weeks) are below a threshold. The person skilled in the art is aware that “below a threshold” means “above a threshold” depending on whether the thresholding is applied to in the context of correlation or inverse correlation with a mood disorder.

The inventors have noticed that standard metrics such as standard deviation or mean for the observational period are not enough to distinguish accurately between healthy (Fig. 2), healthy but chronically stressed (Fig. 3) and people who have a mood disorder or are at the risk of such (Fig. 4). Thus, the method of the invention considers a meaningful biological pattern in the values of the data point of the parameter(s) and/or weighted sets of data points of the parameter(s) (Example 1 ). Reactions to stressful events, detected e.g. by deviations of data points from a predefined pattern, are not necessarily indicative for the probability of a subject to develop a mood disorder. In subjects with a mood disorder or with a high probability to develop a mood disorder reactions to stressful events are often not only characterized by an acute stress response, such as short-term deviations from a predefined pattern, but also by exhaustion symptoms, indicated by prolonged deviations from a predefined pattern. Thus, for a score indicative of the probability of a subject to develop a mood disorder, short-term deviations, in particular large short-term deviations, of data points from a predefined pattern can be of lower relevance compared to prolonged steady deviations of data points from a predefined pattern. Accordingly, specific weighting of data points depending on the amount of deviation and duration of deviation allows the invention to be surprisingly high in accuracy, reproducibility, reliability and/or specificity for determining a score indicative for the probability of the subject to develop a mood disorder. Furthermore, certain mood disorders are diagnosed by symptomatic patterns that persist over at least two weeks. The inventors found that patterns in the parameters described herein or derived from the parameters described herein obtained over at least 14 days can be analyzed to determine a score indicative of the probability of the subject to develop a mood disorder that corresponds surprisingly accurate to diagnosis according to the DSM-5.

In certain embodiments of the invention, recurrent deviation from the predefined pattern divided by the difference in time of obtainment in (a) increases the weight of the data points.

This aspect of the weighting step allows to factor in that recurrent deviations from a predefined pattern increase relevance of these data points for determining a score indicative of the probability of the subject to develop a mood disorder. In an example, repeated short-term deviations from a predefined pattern represents repeated exposure to stressful events, which can result in exhaustion instead of resilience of the subject and therefore increase the probability of the subject to develop a mood disorder and/or increase the relevance for determining a score indicative for the probability of the subject to develop a mood disorder.

Accordingly, specific weighting of data points depending to amount of deviation and duration of deviation and deviation patterns, such as the recurrence of deviation allows the invention to have particular high accuracy, reproducibility, reliability and/or specificity for determining a score indicative for the probability of the subject to develop a mood disorder.

In certain embodiments of the invention, the parameters of the invention comprise autonomic nervous system biomarkers obtained by at least one heart-rate sensor, optical sensor, accelerometer, gyroscope, skin conductance sensor, temperature sensor, respiratory rate sensor, sonic sensor and/or microphone.

These sensors are particularly useful in obtaining the parameter(s) in the method of the invention in that these sensors are particularly simple, non-invasive, user-friendly, energy efficient, sensitive, accurate, stable and/or light.

Specific combinations of these sensors are particularly useful in that the information obtained by the sensors of the combination synergistically enables the method to be simpler, non-invasive, user-friendlier, more energy efficient, more sensitive, more accurate, more stable and/or lighter.

Accordingly, the method of the invention can be surprisingly useful for determining a score indicative of the probability of a subject to develop a mood disorder by the use of specific sensors and/or a specific combination of sensors for obtaining the parameters.

In certain embodiments of the method of the invention, the parameters of the method additionally comprise self-reported subject background data, in particular sociodemographic data and/or BMI.

Accordingly, the method of the invention can be surprisingly useful for determining a score indicative of the probability of a subject to develop a mood disorder by the addition of self-reported subject background data, in particular the addition of sociodemographic data and/or BMI to the parameters.

In certain embodiments of the invention, the parameters of the method comprise at least one interpreted sensor dataset obtained by combination of autonomic nervous system biomarkers and/or by exclusion of at least one data point of autonomic nervous system biomarkers.

Interpreted sensor datasets can be obtained by interpreting, such as combining autonomic nervous system biomarkers, weighting and/or excluding at least one data point of autonomic nervous system biomarkers. These interpreted sensor datasets can be indicative heart rate, heart rate variability, skin conductance, finger temperature, respiratory rate, sleep quality, insomnia, circadian sleep rhythm, phone usage, ambient sounds, bedtime and wake time, sleep duration, sleep quality, cognitive skill and/or facial expression. The interpreted sensor dataset(s) may be preprocessed using a computer implemented program for HRV measurement (Welltory, HRV4Training or any other software measuring HRV) or some models of the fitness trackers and their respective software (such as Apple watch) or chest straps for heart rate measurement and their respective software (such as Polar chest strap) . The preprocessed information embodied by the interpreted sensor datasets is surprisingly more useful for determination of a score indicative for the probability of a mood disorder, than the non- preprocessed information of the underlying autonomic nervous system biomarkers.

Accordingly, the method of the invention can be surprisingly useful for determining a score indicative of the probability of a subject to develop a mood disorder by using at least one parameter comprising at least one interpreted sensor dataset obtained by combination autonomic nervous system biomarkers and/or by exclusion of at least one data point of autonomic nervous system biomarkers, in particular if the interpreted sensor datasets are indicative of heart rate, heart rate variability, skin conductance, finger temperature, respiratory rate, sleep quality, insomnia, circadian sleep rhythm, phone usage, ambient sounds, bedtime and wake time, sleep duration, sleep quality, cognitive skill and/or facial expression.

In certain embodiments, the parameters comprise at least one test that determines a score indicative for cognitive skill, such as working memory, executive function, attention, and/or cognitive flexibility, in particular prefrontal brain network dysfunction.

Therefore, parameters that comprise at least one score of at least one test that embodies additional information, such as the cognitive function and/or alterations of cognitive function of the subject, to improve the method in determining a score indicative of the probability of a subject to develop a mood disorder.

Accordingly, the method of the invention can be surprisingly useful for determining a score indicative of the probability of a subject to develop a mood disorder by combining at least one parameter with at least one test that determines a score indicative for working memory, executive function, attention, cognitive flexibility, and/or prefrontal brain network dysfunction.

In certain embodiments, at least one parameter of the method comprises at least one standardized test. Standardized tests allow particular reliable data obtainment.

Accordingly, the method of the invention can be surprisingly useful for determining a score indicative of the probability of a subject to develop a mood disorder by combining at least one parameter with at least one standardized test.

In certain embodiments of the method, the Wisconsin card sorting test and/or the Stroop test are particularly useful tests, since their score synergistically improves the accuracy and/or consistency of the method for determining a score indicative of the probability of a subject to develop a mood disorder.

Generally, scores from the Wisconsin card sorting test and the Stroop test can be indicative for cognitive skill. The Wisconsin card sorting test is particularly useful to detect a score indicative for cognitive flexibility and/or working memory. The Stroop test is particularly useful to detect a score indicative for attention and/or cognitive control. These cognitive skills are indicative for function of the prefrontal area, which may be altered in subjects prone to develop mood disorders, such as burnout, generalized anxiety disorder and/or depression.

Accordingly, the method of the invention can be surprisingly useful for determining a score indicative of the probability of a subject to develop a mood disorder by combining at least one parameter with the Wisconsin card sorting test and/or the Stroop test.

In certain embodiments of the invention, at least one test is adapted for usage on at least one mobile device. This enables the invention to be surprisingly simple, non- invasive, user-friendly, energy efficient, sensitive, accurate, stable, light, transportable and/or accessible.

The adaptation of the invention or tests of the invention for use on a mobile device, such as a fitness tracker or a smartwatch, allows the obtainment of data in an automated and continuous way, reducing the required time and attention of the subject. Accordingly, the method of the invention can be surprisingly useful for determining a score indicative of the probability of a subject to develop a mood disorder, if the method or a test of the method is adapted for usage on at least one mobile device.

In certain embodiments of the invention, at least two data points in at least one parameter differ in time of obtainment at least 1 week for cognitive skill, 12 hours for heart rate variability (HRV), and/or 24 hours for sleep. The obtainment of a parameter over a prolonged period of time allows the invention to identify long-term patterns, such as patterns of chronic stress and/or exhaustion. These long-term patterns allow identifying predicting trends and accurate weighting of data points, i.e. weighting of data points according to relevance,

Accordingly, the use of the invention for specific time periods of obtainment can be surprisingly useful for determining a score indicative for the probability of the subject to develop a mood disorder.

In certain embodiments of the method, the method comprises the use of an application programming interface (API) to obtain parameters. Such APIs are provided by hardware manufacturer, e.g. by a manufacturer of wearable activity tracker such as MiBand or Fitbit Versa.

Accordingly, the use of an API allows for versatile and flexible application of various hardware compositions. Further, the compatibility of the method of the invention with more than one wearable model allows direct comparison of usefulness of at least one parameter from different wearable models for determining a score indicative of the probability of a subject to develop a mood disorder with the method of the invention.

In certain embodiments of the method, at least one step of the method comprises the use of a computer implemented method. The use of a computer implemented method allows fast, efficient processing in the method. The computer implemented method may comprise oversampling techniques in particular the oversampling techniques SMOTE, ADASYN to allow management of imbalanced data sets and reduce bias in obtained data.

Further, the computer implemented method may use a machine-learning technique to recognize patterns in obtained data, in particular the data underlying the predefined pattern, in certain embodiments of the method. It will be understood that each block of the flowchart in Figure 1 and combinations of blocks in the flowchart in Figure 1 , can be computer implemented. The use of a machine-learning technique allows automatic pattern detection through experience. For determining a score indicative of the probability of a subject to develop a mood disorder a machine-learning technique is surprisingly useful, because the automated nature of this technique allows unbiased, fast and efficient processing of obtained data, which is of particular relevance for this task. Further, the obtained data in the method of the invention includes large data sets and data from several sources, which could not be efficiently processed otherwise, e.g. by explicit programming.

Further, a machine learning technique can be used to generate the predefined pattern and/or an updated version of the predefined pattern, by the use of the score indicative of the probability of a subject to develop a mood disorder as training data. Therefore, the predefined pattern can be used to generate an updated version of the predefined pattern by training using the score generated by the method of the invention earlier. The updated version of the predefined pattern can then be used in the method of the invention to increase accuracy for determining a score indicative of the probability of a subject to develop a mood disorder.

Accordingly, the use of a machine-learning technique in the method of the invention is surprisingly useful for determining a score indicative of the probability of a subject to develop a mood disorder. The use or the combined use of specific machine learning techniques may be particularly surprisingly useful.

In certain embodiments of the method, the machine learning technique comprises the use of Logistic regression, CART, Bagging, Random Forest, Gradient Boosting, Linear Discriminant Analysis, Gaussian Process Classifier and/or Gaussian NB for classification.

In certain embodiments of the method, the machine learning technique, wherein the machine learning technique comprises the use of Linear, Lasso, Ridge, ElasticNet, KNN, DecisionTree, SVR, AdaBoost, GradientBoost, RandomForest, and/or ExtraTrees for regression.

In certain embodiments of the method, wherein the machine learning technique comprises a deep learning technique, in particular simple multilayer perceptron (MLP) and/or convolutional neural networks for classification and/or regression. In certain embodiments of the method, the method is used for both, classification and regression.

In certain embodiments of the method, one or more machine learning techniques and/or one or more steps comprising a machine learning technique are used in the method.

Scores of traditional mood disorder scoring methods, such as standardized mood disorder questionnaire scores are well established, researched and validated (Maslach, C.; Jackson, S.E.; Leiter, M.P. (1996-2016). Maslach Burnout Inventory Manual (Fourth Edition). Menlo Park, CA: Mind Garden, Inc.; R. L. Spitzer, K. Kroenke, J. W. Williams, B. Lowe: A brief measure for assessing generalized anxiety disorder: the GAD-7. In: Arch Intern Med. 166, 2006, S. 1092-1097; Kroenke, Kurt, and Robert L. Spitzer. "The PHQ-9: a new depression diagnostic and severity measure." Psychiatric annals 32.9 (2002): 509-515.).

In certain embodiments of the method, the determined score of the method of the invention results in a surprising accuracy, when compared to a score of a traditional mood disorder scoring method. In certain embodiments of the method the determined score results in an RMSE <= 5, in particular an RMSE <= 4, in particular an RMSE <= 3, in particular an RMSE <= 2, in particular an RMSE <= 1.5 for regression and/or an accuracy of at least 75%, in particular of at least 80%, in particular of at least 85%, in particular of at least 90%, in particular of at least 95% for classification, when compared to a standardized mood disorder questionnaire score of the subject, and/or when compared to a score of a part of a standardized mood disorder questionnaire score of the subject.

A high accuracy of the determined score when compared to scores of traditional mood disorder scoring methods, accordingly allows the determined score to be surprisingly useful and valid for prediction, prevention, diagnosis, monitoring, intervention and/or management of mood disorders in certain embodiments of the method.

In certain embodiments of the method, reference signature obtained from at least one reference subjects comprises a standardized mood disorder questionnaire score, in particular the composite Maslach Burnout Inventory score, and/or the GAD-7 score, and/or the PHQ-9 score from reference subjects. Data from these standardized mood disorder questionnaires is useful as data for the reference signature and/or training data for the invention, because their scores have been proven to be useful in diagnosing mood disorders. The well-documented use of these questionnaires for diagnosis of mood disorders is beneficial, because it facilitates the determined score of the invention to be validated and integrated in current treatment procedures. The use of traditional mood disorder scoring methods comprised in the reference signature obtained from at least one reference subjects allows for improvement of accuracy of classification and/or regression of the determined score, when compared to a standardized mood disorder questionnaire score of the subject, and/or when compared to a score of a part of a standardized mood disorder questionnaire score of the subject. This improvement of accuracy is unexpected, since the parameters obtained of a subject may not necessarily be obtained from the same composition of data sources as the parameters obtained of a reference subject. In an example the parameters of a subject are obtained by sensors, while the parameters of the reference subject are obtained by standardized mood disorder questionnaire of the reference subjects.

The Maslach Burnout Inventory, the GAD-7, and/or the PHQ-9 questionnaire are useful scores, since these questionnaires are shorter and therefore simpler and more effective to implement compared to full-length questionnaires and other questionnaires used in the clinics to score or diagnose mood disorders, such as burnout, generalized anxiety disorder and/or depression. Without being bound to a specific mechanism or theory, mood disorders, such as burnout, generalized anxiety disorder and/or depression may have common underlying cognitive skill alterations and/or common alterations in autonomic nervous system biomarkers. Therefore, one embodiment of the method of the invention may not only be useful for determining a score indicative of the probability of the subject to develop burnout, but surprisingly the same embodiment may also be useful for determining a score indicative of the probability of the subject to develop depression, generalized anxiety disorder and/or a combination of burnout, generalized anxiety disorder and/or depression. Accordingly, the method of the invention is surprisingly useful for determining a score indicative of the probability of the subject to develop a mood disorder, by the use of the Maslach Burnout Inventory score, and/or the GAD-7 score, and/or the PHQ-9 score from reference subjects. In certain embodiments of the method, the determined score of a subject further results in an RMSE <= 5, in particular an RMSE <= 4, in particular an RMSE <= 3, in particular an RMSE <= 2, in particular an RMSE <= 1 .5 for regression and/or an accuracy of at least 75%, in particular of at least 80%, in particular of at least 85%, in particular of at least 90%, in particular of at least 95% for classification of emotional exhaustion, in an RMSE <= 5, in particular an RMSE <= 4, in particular an RMSE <= 3, in particular an RMSE <= 2, in particular an RMSE <= 1 .5 for regression and/or an accuracy of at least 75%, in particular of at least 80%, in particular of at least 85%, in particular of at least 90%, in particular of at least 95% for classification of depersonalization, and/or in an RMSE <= 5, in particular an RMSE <= 4, in particular an RMSE <= 3, in particular an RMSE <= 2, in particular an RMSE <= 1 .5 for regression and/or an accuracy of at least 75%, in particular of at least 80%, in particular of at least 85%, in particular of at least 90%, in particular of at least 95% for classification of personal accomplishment, when compared to the dimension(s) score of the composite Maslach Burnout Inventory score of a subject.

Accordingly, the unexpected accuracy allows the method to be particularly useful for prediction, prevention, diagnosis, monitoring, intervention and/or management of mood disorders.

In certain embodiments of the method, the machine learning technique used for generation of the predefined pattern comprises the use of a score indicative of the probability of the subject to develop a mood disorder.

By using a score indicative of the probability of the subject to develop a mood disorder i.e. the score determined by the method of the invention, the predefined pattern can be updated in order to improve the accuracy of the method of the invention thereinafter. The method of the invention can update the predefined pattern either by having a training period for the generation of a pretrained model or the predefined pattern may be continuously updated e.g. by continual learning.

Accordingly, the method may improve the accuracy of the determined score by repeatedly using a score indicative of the probability of the subject to develop a mood disorder to update the predefined pattern.

In certain embodiments of the method, the reference signature obtained from at least one reference subject is based on a standardized mood disorder questionnaire score. The use of a standardized mood disorder questionnaire score of a certain mood disorder for the reference signature, may allow the method of the invention to be indicative for this specific mood disorder.

Accordingly, the method of the invention is particularly useful if the signature obtained from at least one reference subject is based on a standardized mood disorder questionnaire score.

In certain embodiments of the method, the standardized mood disorder questionnaire score comprises the composite Maslach Burnout Inventory score, and/or the GAD-7 score, and/or the PHQ-9 score.

The use of a the composite Maslach Burnout Inventory score, and/or the GAD-7 score, and/or the PHQ-9 score for the reference signature, may allow the method of the invention to be indicative for this specific burnout, generalized anxiety disorder, and/or depression.

Accordingly, the method of the invention is particularly useful if the signature obtained from at least one reference subject is based on the composite Maslach Burnout Inventory score, and/or the GAD-7 score, and/or the PHQ-9 score.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, suitable methods and materials are described below. In case of conflict, the present specification, including definitions, will control. In addition, the materials, methods, and examples are illustrative only and not intended to be limiting.

The general methods and techniques described herein may be performed according to conventional methods well known in the art and as described in various general and more specific references that are cited and discussed throughout the present specification unless otherwise indicated.

While aspects of the invention are illustrated and described in detail in the figures and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive. It will be understood that changes and modifications may be made by those of ordinary skill within the scope and spirit of the following claims. In particular, the present invention covers further embodiments with any combination of features from different embodiments described above and below.

In certain embodiments, the invention relates to the method of the invention, wherein the method further comprises administering to the subject having a score indicative for a mood disorder a pharmaceutically effective amount of a pharmaceutical product for the treatment of a mood disorder.

An "effective amount" of an agent, e.g., a pharmaceutical product, refers to an amount effective, at dosages and for periods of time necessary, to achieve the desired therapeutic or prophylactic result.

The term "treatment" (and grammatical variations thereof such as "treat" or "treating"), as used herein, refers to clinical intervention in an attempt to alter the natural course of the individual being treated, and can be performed either for prophylaxis or during the course of clinical pathology. Desirable effects of treatment include, but are not limited to, preventing occurrence or recurrence of disease, alleviation of symptoms, diminishment of any direct or indirect pathological consequences of the disease, preventing metastasis, decreasing the rate of disease progression, amelioration or palliation of the disease state, and remission or improved prognosis.

Therefore, the score indicative of the probability of a subject to develop a mood disorder described herein, can be used in the context of a method of treatment, wherein the score is used, e.g., to identify susceptible subjects and/or influence treatment parameters such as which pharmaceutical product is used in the treatment and/or dose and/or administration patterns thereof.

In certain embodiments, the invention relates to a pharmaceutical product for use in the treatment of a mood disorder in a subject, wherein the subject has a score indicative for a mood disorder determined according to the invention.

In some embodiments, the pharmaceutical product described herein comprises an antidepressant agent and/or an anxiolytic agent.

In some embodiments, the antidepressant agent described herein is an antidepressant agent selected from the class of SSRI, SNRI, SMS, SARI, NRI, NDRI, TCA, TeCA, MAOI. In some embodiments, the antidepressant agent described herein is an antidepressant agent selected from the group of Agomelatine, Esketamine, Ketamine, Tandospirone, Tianeptine, Metralindole, Moclobemide, Pirlindole, Toloxatone, Caroxazone, Selegiline, Isocarboxazid, Phenelzine, Tranylcypromine, Amoxapine, Maprotiline, Mianserin, Mirtazapine, Setiptiline, Amitriptyline, Amitriptylinoxide, Clomipramine, Desipramine, Dibenzepin, Dimetacrine, Dosulepin, Doxepin, Imipramine, Lofepramine, Melitracen, Nitroxazepine, Nortriptyline, Noxiptiline, Opipramol, Pipofezine, Protriptyline, Trimipramine, Bupropion, Atomoxetine, Reboxetine, Teniloxazine, Viloxazine, Trazodone, Vilazodone, Vortioxetine, Desvenlafaxine, Duloxetine, Levomilnacipran, Milnacipran, Venlafaxine, Citalopram, Escitalopram, Fluoxetine, Fluvoxamine, Paroxetine, and Sertraline.

The term “anxiolytic agent”, as used herein, refers to a therapeutic agent used in the treatment of symptoms in patients with anxiety or emotional disorders including stress, anxiety, neurosis, and obsessive-compulsive disorder. Anti-anxiety drugs are usually divided into two broad categories: benzodiazepines and non-benzodiazepines. In some embodiments, the anxiolytic agent described herein is benzodiazepine. In some embodiments, the anxiolytic agent described herein is a benzodiazepine selected from the group of clonazepam, diazepam, estazolam, flunitrazepam, lorazepam, midazolam, nitrazepam, oxazepam, triazolam, temazepam, chlordiazepoxide, alprazolam, clobazam, clorazepate, etizolam.

In some embodiments, the anxiolytic agent described herein is a non-benzodiazepine. In some embodiments, the anxiolytic agent described herein comprises at least one non-benzodiazepine selected from the class of serotonin 1A agonists, barbiturates, carbamates, antihistamines, opioids, and Z-drugs. In some embodiments, the anxiolytic agent described herein comprises at least one non-benzodiazepine selected from the group of buspirone, amobarbital, aprobarbital, butabarbital, mephobarbital, methohexital, pentobarbital, phenobarbital, primidone, secobarbital, thiopental, meprobamate, carisoprodol, tybamate, lorbamate, zaleplon, zolpidem, zopiclone, eszopiclone, chlorpheniramine, dexchlorpheniramine, dimenhydrinate, diphenhydramine, promethazine, trimeprazine, gabapentin, pregabalin, tramadol, tapentadol, morphine, diamorphine, hydromorphone, oxymorphone, oxycodone, hydrocodone, methadone, propoxyphene, meperidine, fentanyl, codeine, carfentanil, remifentanil, alfentanil, sufentanil, phenibut, mebicar, and gamma-hydroxybutyric acid.

The means and method described herein enable the identification of patient populations that would not have been detected by methods known in the art. For example, the means and methods of the invention enable early detection of patients in need of a treatment without the requirement of a doctor’s appointment.

Accordingly, the invention is at least in part based on unique subject population that is identifiable by the method of the invention to be treated with the pharmaceutical product described herein.

In certain embodiments, the invention relates to a method for monitoring a mood disorder the method comprising the steps of:

1 ) determining a score indicative of the probability of a subject to develop a mood disorder using the method of the invention at a first timepoint;

2) determining a score indicative of the probability of a subject to develop a mood disorder using the method of the invention at a second timepoint;

3) comparing the score of step (1 ) with the score of step (2); and

4) monitoring the mood disorder in the subject based on the comparison of step (3).

Therefore, the method for monitoring described herein can be implemented by determining a score indicative for the risk of developing a mood disorder over a prolonged period of time (see e.g., example 2 months). As such the for example early development, progression and/or treatment success of mood disorders can be monitored.

In certain embodiments, the invention relates to a library comprising a score indicative of the probability of the subject to develop a mood disorder determined according to the invention.

In certain embodiments, the invention relates to a storage device comprising computer- readable program instructions to execute the method according to the invention, preferably additionally comprising the library according to the invention.

In some embodiments, the storage device described herein is at least one selected from the group of electronic storage device, magnetic storage device, optical storage device, electromagnetic storage device, semiconductor storage device, any suitable combination thereof. A non-exhaustive list of more specific examples of the storage device includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A storage device, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

In certain embodiments, the invention relates to a system for determining a score indicative for a mood disorder comprising at least one sensor for obtainment of at least one parameter indicative for a mood disorder and the storage device of the invention.

In certain embodiments, the invention relates to a server comprising the storage device of the invention, at least one processing device, and a network connection for receiving at least one parameter indicative of the state of mood of the subject and/or at least one weighted set of data points of the parameter(s).

The term “network connection”, as used herein, refers to a communication channel of a data network. A communication channel can allow at least two computing systems to communicate data to one another. In some embodiments, the data network is selected from the group of the internet, a local area network, a wide area network, and a wireless network. The network may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers. A network adapter card or network interface in each computing/processing device receives computer readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium within the respective computing/processing device. The server described herein, can receive at least one parameter indicative of the state of mood of the subject and/or at least one weighted set of data points of the parameter(s), process it according to the method of the invention, and provide a result. Sending the parameter/weighted set to a server reduces the requirements for processing power in the device that acquires the parameter/weighted set and enables the efficient processing of large datasets (e.g. large reference signatures). In embodiments, wherein the invention relates to a server, the parameter/weighted set can be acquired by any device that has a network connection.

The server may be connected to the device for the acquirement of the vascular image through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider).

In certain embodiments, the invention relates to the method of the invention, the pharmaceutical product for use the invention, the storage device of the invention, the system of the invention, the server of the invention or the library according to the invention, wherein the mood disorder is a stress-associated mood disorder.

In certain embodiments, the invention relates to the method of the invention, the pharmaceutical product for use the invention, the storage device of the invention, the system of the invention, the server of the invention or the library according to the invention, wherein the mood disorder is at least one disorder selected from the group of burnout, generalized anxiety disorder, and depression.

The means and methods described herein enable to determine at same time scores indicative of the risk of several mood disorders. Certain mood disorders are frequently comorbid with other mood disorders. Determining scores indicative of the risk to develop several mood disorders from the same or overlapping data points is particularly useful to effectively estimate risk, disease progression, and/or plan therapy.

In certain embodiments, the invention relates to the method of the invention, the storage device of the invention, the pharmaceutical product for use according to the invention, the system of the invention, the server of the invention or the library according to the invention, wherein the mood disorder is burnout. In certain embodiments, the invention relates to the method of the invention, the storage device of the invention, the pharmaceutical product for use according to the invention, the system of the invention, the server of the invention or the library according to the invention, wherein the mood disorder is an symptom of burnout, preferably at least one symptom selected from the group of emotional exhaustion, depersonalization, and reduced personal accomplishment.

Brief description of Figures

Brief description of Figure 1

Figure 1 describes a schematic overview of some of the steps involved in the method of the invention.

Fig.2: Daily ML predicted score of the EE dimension of MBI of a healthy person with a single stressful event

Fig.3: Daily ML predicted score of the EE dimension of MBI of a healthy person under repetitive stress

Fig.4: Daily ML predicted score of the EE dimension of MBI of a person diagnosed with depression

Fig. 5: Illustration of timeline of the training data collection: data collection from wearables - daily (illustrated by the dotted line on top); Stroop, Wisconsin - at the start and at the end of 2 weeks window (indicated by the dotted arrow); PHQ-9, GAD-7, MBI - at the start and at the end of 2 weeks window (indicated by solid line arrow)

Fig. 6: Illustration of timeline of the training of the machine learning model

Fig. 7: Illustration of the estimation of daily ML predicted score (separately for each questionnaire) using trained ML model

Fig. 8: Illustration of the thresholding of the ML output over 14 days - sliding window is used (separately for each questionnaire)

Detailed description of the invention

The term “state of mood”, as used herein, refers to any affective state or factors of an individual, which influence the affective state. State of mood includes but is not limited to level of stress, exhaustion, anger, sadness, anxiety, happiness, aggression, selfesteem, sexual arousal, lack of sleep, nervousness, excitement.

The term “weighting”, as used herein, refers to a mathematical method of assigning a weight to a data point, i.e. to factor in the relevance of data points. While data points with a weight of 0 are excluded from a calculation step, data points with a high weight are of high relevance for a calculation step. A “predefined pattern”, as used herein, refers to a pattern that is defined before the method of the invention is applied. In certain examples of the invention a “predefined pattern” can refer to a set of at least two values, such as threshold values, and/or the output of a machine learning algorithm, such as a model, in particular a pretrained model.

A “reference signature”, as used herein, refers to a value or a pattern that is defined using data from the reference subjects.

The term “time of obtainment”, as used herein, refers to the point in time in which a data point, such as data point of a parameter, was obtained.

A “score”, as used herein, refers to a value, a category and/or a classification.

As used herein, “mood disorder” refers to a group of conditions where a disturbance in the subject's mood is an underlying feature. The classification may be done according to the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association (2013). Diagnostic and Statianual (5th ed.). Arlington, VA: American Psychiatric Association) and/or International Classification of Diseases (World Health Organization. (2018). International classification of diseases for mortality and morbidity statistics (11th Revision)). In some embodiments, the mood disorder described herein is characterized by pervasive, prolonged, and/or disabling exaggerations of mood and affect that are associated with behavioral, physiologic, cognitive, neurochemical and/or psychomotor dysfunctions. The mood disorder described herein can include, but is not limited to major depressive disorder, bipolar disorder, dysthymic disorder, psychotic major depressive disorder, melancholic major depressive disorder, seasonal pattern depression, postpartum depression; brief recurrent depression; late luteal phase dysphoric disorder and cyclothymic disorder, post-traumatic stress disorder, obsessive-compulsive disorder, burnout, and anxiety disorder such as social anxiety disorders and generalized anxiety disorder.

The term “burnout”, as used herein, refers to stress- or exhaustion related mood disorders, for which the Maslach Burnout Inventory is useful in assessing the mood disorders and/or symptoms of the mood disorders. In some embodiments, the burnout described herein refers to the state of vital exhaustion in meaning of International Classification of Disease, ICD-10 or subsequent editions. Frequently, burnout is the state of physical, emotional and mental exhaustion that results from chronic occupational stress. The nonexclusive examples of well-known methods of measurement of burnout are the Maslach Burnout Inventory, Shirom-Melamed Burnout Questionnaire, and burnout test of Jan Boettcher. The burnout syndrome includes nonexclusively emotional exhaustion, depersonalization, fatigue, and reduced personal accomplishment.

The term “generalized anxiety disorder”, as used herein, refers to anxiety-related mood disorders, for which the GAD-7 questionnaire is useful in assessing the mood disorders and/or symptoms of the mood disorders.

The term “depression”, as used herein, refers to mood disorders characterized by low- mood related, for which the PHQ-9 questionnaire is useful in assessing the mood disorders and/or symptoms of the mood disorders. An "individual" or "subject" as used herein, refers to a human.

The terms “questionnaire” and “scale” are used herein interchangeably and refer to a set of questions with a choice of answers.

The term “obtained data”, as used herein, refers to any information obtained and/or used in the method in raw or processed form. Obtained data may therefore comprise results from tests, interpreted sensor datasets, autonomic nervous system biomarkers, parameters, predefined pattern, weighted set of parameters, reference signature, score indicative of the probability of the subject to develop a mood disorder.

A “data source”, as used herein, refers to a source of data such as a sensor, a questionnaire and /or a test.

The term “training data”, as used herein, refers to a data that can be used for training of a machine learning technique.

A “sensor” as used herein is a device, module, machine, or subsystem whose purpose is to detect events or changes in its environment and send the information to other electronics, such as a computer processor. In one or more examples the sensor is a skin conductance sensor, a gyroscope, an optical sensor, a heartbeat sensor, a temperature sensor, a respiratory rate sensor, a sonic sensor and/or a microphone.

A “heartbeat sensor”, as used herein, refers generally to any sensor capable of measuring the patient’ s heart rate and/or waveforms associated with the heartbeat and may be worn by a patient and/or applied to a patient. A common type of heartbeat sensor operates on the principle of photo plethysmography, which involves illuminating the skin (e.g., with a light-emitting diode and measuring the light reflected or transmitted to a photodiode to detect blood volume changes in the microvascular tissue in and/or immediately underneath the skin. They can be found integrated, for example, in mobile devices such as in Fitbit devices as well as in some smartphones. Other types of heartbeat sensors measure electrical signals indicative of cardiac activity. Electrocardiogram sensors including a set of one or more electrodes placed in contact with the patient’s skin may also serve as heartbeat sensors.

The term “heart rate variability” or “HRV”, as used herein, refers to the variance in time between heartbeats. Heart rate variability includes but is not limited to pNN50, AMo50, Mean RR, SDNN, rMSSD measurements.

An “optical sensor”, as used herein, refers to any opto-electronic component responsive to a band of light wavelengths.

An “accelerometer”, as used herein, refers to any of the wide variety of sensors and/or devices available that can sense acceleration.

A "gyroscope", as used herein, refers to a device for measuring movement about a rotational axis.

A “skin conductance sensor”, as used herein, refers to any of the wide variety of sensors and/or devices available that can sense skin conductance and/or that can determine a score indicative for electrodermal activity.

A “biomarker”, as used herein, refers to a measurable indicator of some biological state or condition. In one or more examples a biomarker is embodied by stored data from a sensor.

The “autonomic nervous system”, as used herein, refers to a division of the peripheral nervous system that acts largely unconsciously and regulates bodily functions, such as the heart rate, digestion, respiratory rate, pupillary response, urination, and sexual arousal. The phrase “autonomic nervous system biomarker”, as used herein, refers to any data that can be obtained from a subject by a sensor, that does not require the subject to enter the data actively and consciously. An “interpreted sensor dataset”, as used herein, refers to dataset comprising data from a combination of data of autonomic nervous system biomarkers and/or by weighting of at least one data point of autonomic nervous system biomarkers. In one or more examples an interpreted sensor dataset is indicative for heart rate, heart rate variability, skin conductance, finger temperature, respiratory rate, sleep quality, insomnia, circadian sleep rhythm, phone usage, ambient sounds, bedtime and wake time, sleep duration, physical activity, and/or facial expression.

The term “physical activity”, as used herein, refers to activity of a subject, in particular movement, steps, calorie usage.

The term “phone usage”, as used herein, refers to time spent for usage of a mobile device and/or a pattern of usage of a mobile device. In particular “phone usage”, may refer to the usage of a mobile device for specific tasks such as games, web surfing and/or communication tools, more particularly to the usage of a mobile device for text messages or phone calls, “facial expression”, as used herein, refers to any data obtained from the face of a subject by a sensor, such as processed data obtained from a certain state of the face of the subject using a camera of a mobile device.

The term “Subject background data”, as used herein, refers to characteristics of a subject that can, i.e. for technical or temporal reasons, not be obtained by the sensors that are used to obtain the other parameters.

The term “BMI” or “body mass index”, as used herein, refers to a value derived from the mass (weight) and height of a person. The BMI is defined as the body mass divided by the square of the body height.

The term “sociodemographic data”, as used herein, refers to characteristics of a subject such as age, sex, education, migration background, ethnicity, religious affiliation, marital status, household, employment, and/or income.

The term “cognitive skill”, as used herein, refers to any brain-based skill which is useful in acquisition of knowledge, manipulation of information and/or reasoning. Cognitive skill or aspects thereof can be measured, inter alia, by the Wisconsin Card Sorting test and Stroop test. Aspects of cognitive skill include, but are not limited to, working memory, executive function, attention, cognitive control and/or cognitive flexibility, such as prefrontal brain network function and/or dysfunction. The term “mobile device”, as used herein, refers to any portable device comprising sensors and/or processing capabilities, such as a wearable device, smartphone, smartwatch, wearable sensor, portable multimedia device and/or tablet computer. The mobile device, in particular the smartphone, smartwatch, wearable sensor, portable multimedia device or tablet computer, may actively or passively collect data from the user, for example using an app installed on the mobile device. It is preferred that the mobile device is able to transmit data to a system, for example a server or cloud-based system, able to preprocess, process and/or analyze the data collected by the mobile device.

“Traditional mood disorder scoring methods”, as used herein, refer to established, validated and/or standardized methods that result in a score indicative for a mood disorder. Traditional mood disorder scoring methods include, but are not limited to standardized mood disorder questionnaires, mental health evaluations diagnosed by health workers.

A “test”, as used herein, refers to a task and/or a method to measure and/or improve a subject’s performance. A test may include, inter alia, a version of a Wisconsin card sorting test and/or a version the Stroop test.

The “Wisconsin card sorting test”, as used herein, refers to a test described in “The Professional Manual” for the WCST was written by Robert K. Heaton, Gordon J. Chelune, Jack L. Talley, Gary G. Kay, and Glenn Curtiss or to an adapted version thereof, i.e. a more challenging version and/or a version suitable to be used offline, online and/or on a mobile device.

The “Stroop test”, as used herein, refers to a test that determines a score and involves by examining the Stroop effect of a subject. The Stroop effect refers to the delay in reaction time between congruent and incongruent stimuli. Scores of the stroop test may be indicative for congruent and incongruent speed in the correct trials of a subject and/or error count of a subject, such as total, preservation and/or non-preservation error count of a subject.

The term “standardized”, as used herein, may refer to any form of standardization. Further, “standardized” may refer to “established”, “repeatedly used by professionals”, “well-known”. Accordingly, standardized tests may be tests administered and scored in a predetermined, standard manner (Popham, W.J. (1999). "Why standardized tests don't measure educational quality". Educational Leadership. 56 (6): 8-15).

A “mood disorder questionnaire”, as used herein, refers to a questionnaire involved in evaluating, predicting and/or diagnosing of a mood disorder, symptoms of a mood disorder, and/or risk factors to develop a mood disorder. Examples of mood disorder questionnaires and/or standardized mood disorder questionnaires are the Maslach Burnout Inventory test, GAD-7 questionnaire, PHQ-9 questionnaire.

The “composite Maslach Burnout Inventory” or “Maslach Burnout Inventory” or “MBI”, as used herein, refer to a questionnaire described in Maslach Burnout Inventory Manual (Fourth Edition) (Maslach, C.; Jackson, S.E.; Leiter, M.P. (1996-2016) Menlo Park, CA: Mind Garden, Inc.) or an adapted version thereof that assesses the same symptoms of a mood disorder.

“GAD-7”, as used herein, refers to a questionnaire described in “A brief measure for assessing generalized anxiety disorder: the GAD-7”( R. L. Spitzer, K. Kroenke, J. W. Williams, B. Lowe:. In: Arch Intern Med. 166, 2006, S. 1092-1097) or an adapted version thereof that assesses the same symptoms of a mood disorder.

“PHQ-9”, as used herein, refers to a questionnaire described in Kroenke, Kurt, and Robert L. Spitzer. "The PHQ-9: a new depression diagnostic and severity measure." Psychiatric annals 32.9 (2002): 509-515.) or an adapted version thereof that assesses the same symptoms of a mood disorder.

The term “oversampling”, as used herein, refers to adjusting the class distribution of multiple classes (or categories) represented in a given data set. Moreover, oversampling generally includes selecting data points from a minority class (that is, a class that is underrepresented in the given data set as compared to one or more other classes) to serve as the basis for the generation of additional and/or synthetic data points in an attempt to balance the class distribution in the given data set.

The term “machine learning technique”, as used herein, can refer to an application of artificial intelligence technologies to automatically learn and/or improve from an experience (e.g., training data and/or obtained data) without the necessity of explicit programming of the lesson learned and/or improved. In one or more embodiments, the machine learning technique is used for classification and may comprise, inter alia, Logistic regression, CART, Bagging, Random Forest, Gradient Boosting, Linear Discriminant Analysis, Gaussian Process Classifier and/or Gaussian NB. In one or more embodiments, the machine learning technique is used for regression and may comprise, inter alia, Linear, Lasso, Ridge, ElasticNet, KNN, DecisionTree, SVR, AdaBoost, GradientBoost, Random Forest, and/or ExtraTrees for regression. Some machine learning techniques can be used for both, regression and classification. Such machine learning techniques include, inter alia, deep learning technique, in particular simple multilayer perceptron (MLP) and/or convolutional neural networks for classification and/or regression.

Furthermore, in the claims the word "comprising" does not exclude other elements or steps, and the indefinite article "a", "the" or "an" does not exclude a plurality. A single unit may fulfill the functions of several features recited in the claims. The terms “essentially”, “about”, “approximately” and the like in connection with an attribute or a value particularly also define exactly the attribute or exactly the value, respectively. Any reference signs in the claims should not be construed as limiting the scope.

While aspects of the invention are illustrated and described in detail in the figures and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive. It will be understood that changes and modifications may be made by those of ordinary skill within the scope and spirit of the following claims. In particular, the present invention covers further embodiments with any combination of features from different embodiments described above and below. Although the foregoing invention has been described in some detail by way of illustration and example for purposes of clarity of understanding, it will be obvious that certain changes and modifications may be practiced within the scope of the appended claims.

Examples

Example 1

This example followed the following procedure:

(1 ) a) Data was collected from a wearable from study participants and questionnaires were used to clinically assess mood disorders (Fig. 5) b) A machine learning model was trained to predict the questionnaire score based on the data collected in (1 )a) (Fig. 6)

(2) The machine learning model of (1)b) was used to estimate daily questionnaire scores of an individual based on data from a wearable (Fig. 7)

(3) Thresholding of the machine learning output over 14 days was used to classify the individual as healthy or having a mood disorder (Fig. 8).

Participants

Six volunteers, randomly chosen from particular demographics (age 32-55, 3 women and 3 men, currently employed), were included in a study by study organizers. 5 of the volunteers had no other (i.e. excluding depression, anxiety, burnout) neurological or psychiatric diagnosed disorder for at least 2 past years, 2 volunteers had no prior history of being diagnosed with depression and/or burnout and/or anxiety, 1 volunteer was in the stage of stable depression remission, 2 volunteers were currently in treatment with diagnosis of depression and/or burnout and had anxiety symptoms (GAD-7 score above 10). Thus, 2 volunteers were healthy without a history of mood disorder, 1 volunteer was healthy with a history of mood disorder, 1 volunteer recently recovered from a mood disorder and 2 volunteers were in treatment. Data was collected from over a 2-month period.

Clinical assessment

At the beginning and at the end of the measurement periods (each period is two consecutive weeks) during the study the mood status of the participants was assessed using the following questionnaires PHQ-9, GAD-7, Maslach Burnout Inventory. Data collection and analyses

Data was collected using an online Wisconsin card sorting test and an online Stroop test, two Fitness trackers (Xiaomi MiBand, Fitbit Versa) and the Welltory FIRV measurement app. The data between the two Fitness trackers were compared and the Fitbit Versa appeared to provide more consistent measurements. The further analysis was done on the Fitbit Versa. FIRV parameters from Welltory measurement app have been used: pNN50, AMo50, Mean RR, SDNN, rMSSD.

Continuous data gathered for FIRV, total sleep duration and the duration of sleep phases (deep, light, REM, awake) and activity per day (automatically recorded sport activity, number of steps, active minutes). No user entry was required and data gathered autonomously from the API of the wearable device.

Subjects performed the Wisconsin card sorting test and the Stroop test 2 times per month (biweekly) during the experiment.

Prior to analyses, autonomic nervous system biomarkers were extracted from fitness- trackers and mobile phones data to form parameters. Parameters included person- level measures of (1) sleep phases and sleep duration, (2) heart rate variability (e.g. pNN50, AMo50, Mean RR, SDNN, rMSSD), (3) performance at Wisconsin card sorting test and Stroop test.

Subsequently the data collected from wearables, Stroop, Wisconsin scores and respective scores from PHQ-9, GAD-7, MBI each for every two weeks during 2 months time (4 periods in total, 2 periods are illustrated in Fig. 6) was averaged.

Averaged PHQ-9, GAD-7, MBI - are used as labels to train the model

Averaged Stroop, Wisconsin, data collection from wearables - are used at data input to the model

Example how we average: data collected from wearables all days from day 1 to day14 (periodl ), all days from day 14 to day 28 (period2) of measurement (equivalent procedure for periods 3 and 4). Stroop, Wisconsin from day1 and day 14 (periodl ), from day 14 and day 28 (period2) of measurement (equivalent procedure for periods 3 and 4).

PHQ-9, GAD-7, MBI from day1 and day 14 (periodl ), from day 14 and day 28 (period2) of measurement (equivalent procedure for periods 3 and 4).

Machine learning model

Libraries used: Python 3.8.3, Pandas 1.0.5, Numpy 1.18.5, Scikit learn (sklearn) 0.23.1 , Miens 0.2.3,

Gather data:

Exporting CSV data from the Fitbit API or directly from the GUI Importing CSV with pandas

Adding the MBI, GAD-7 and PHQ-9 questionnaire scores to the CSV Preprocessing:

Scaling of the numeric data from the CSV file defined above with MinMaxScaler

Split the data obtained in the previous step into train and test (80/20)

Two types of machine learning models have been created: regression and classification.

The following regression models have been used:

Linear, Lasso, Ridge, ElasticNet, KNN (k-nearest neighbors), Decision trees, SVR (support vector regression), AdaBoost, Gradient Boosting, Random Forest, Extratrees.

The following classification models have been used:

Logistic regression, CART, Bagging, Random Forest, Gradient Boosting, Linear Discriminant Analysis, Gaussian Process Classifier, Gaussian NB, SVC (support vector classification).

A score was determined from the parameters. Oversampling techniques such as SMOTE and ADASYN have been used to adjust the class distribution. Primary outcome measures for regression models were the scores on the 3 MBI scales (emotional exhaustion, depersonalization and personal accomplishment) and the score on PHQ-9 and GAD-7 scales.

Multiple regression and classification models described above were trained and their performance compared.

After training multiple regression and classification models, a model ensemble for regression and classification has been trained with SuperLearner from miens.

Training and evaluation

The predefined pattern was trained and validated using the obtained parameters. The parameters contained HRV, total sleep duration and the duration sleep phases (deep, light, REM, awake) and cognitive test performance features. The validation is done in a 5-fold cross-validation scheme.

Model performance is evaluated using metric of RMSE (root mean squared error). RMSE measures the average of the errors — that is, the average difference between the estimated values and the actual value.

Model performance for classification has been evaluated using the accuracy metric. Thresholding:

Thresholding values:

For each assessed mood disorder the threshold was selected as follows:

- for MBI: emotion exhaustion >= 26 (here and after - score value), depersonalisation >= 9, personal accomplishment =< 34 (note that personal accomplishment is the only dimension where the threshold is inverted, meaning that in case if 70% of measurements within the last two weeks are below threshold this is indicative of burnout).

- for PHQ-9: >=10

- for GAD-7: >=10

Thresholding procedure: A daily machine learning (ML) output score above the threshold for equal or more than 70% of the measurements within a time period of 14 consecutive days, indicates that a person is considered at risk of developing a mood disorder or having a mood disorder at the moment.

A daily machine learning output score above the threshold for less than 70% of the measurements within a time period of 14 consecutive days, indicates that a person is considered healthy.

A sliding window was used to determine the 14 consecutive days.

The thresholding procedure was used to determine a score indicative of depression, a score indicative of burnout and a score indicative of anxiety.

Exemplified threshold graphs can be found in Figure 2 to 4.

Model Performance - results

In this example, the determined score of a subject further resulted in an RMSE = 1 .155 for regression for the emotional exhaustion, RMSE = 2.504 for regression for the depersonalization, and RMSE = 1.206 for regression for the personal accomplishment, when compared to the dimension(s) score of the composite Maslach Burnout Inventory score of a subject. Accordingly, the RMSE score for predicting the PHQ-9 scale was 2.511 and for GAD-7 scale the score was 4.989.

Regarding the classification accuracy, the models have achieved the following accuracy: 87.81 % for MBI, 86.46% for PHQ-9 and 87.74% for GAD-7.

Claims

2. The method of claim 1 , wherein at least two data points in step (a) differ at least two weeks in time of obtainment.

3. The method of claim 1 or 2, wherein the method comprises a step of distinguishing between short-term events and long-term events.

4. The method of claim 3, wherein the step of distinguishing between short-term events and long-term events comprises thresholding.

5. The method of claim 1 to 4, wherein the parameters comprise autonomic nervous system biomarkers obtained by at least one heartbeat sensor, optical sensor, accelerometer, gyroscope, skin conductance sensor, temperature sensor, sonic sensor and/or microphone.

6. The method of claim 5, wherein the parameters additionally comprise self- reported subject background data.

7. The method of claim 6, wherein the self-reported subject background data comprises sociodemographic data and/or BMI.

8. The method of claim 5 to 7, wherein the parameters comprise at least one interpreted sensor dataset obtained by combination of autonomic nervous system biomarkers and/or by weighting of at least one data point of autonomic nervous system biomarkers.

9. The method of claim 8, wherein at least one of the interpreted sensor datasets is indicative of heart rate, heart rate variability, skin conductance, finger temperature, respiratory rate, sleep quality, insomnia, circadian sleep rhythm, phone usage, ambient sounds, bedtime and wake time, sleep duration, sleep quality, cognitive skill and/or facial expression .

10. The method of claim one of the previous claims, wherein the parameters comprise at least one test that determines a score indicative of cognitive skill.

11. The method of claim 10, wherein at least one test is indicative of working memory, executive function, attention, and/or cognitive flexibility.

12. The method of claims 11 , wherein cognitive flexibility is indicative of prefrontal brain network dysfunction.

13. The method of claim any one of the previous claims, wherein the parameters comprise at least one standardized test.

14. The method of claim any one of the previous claims, wherein the parameters comprise the Wisconsin card sorting test and/or the Stroop test.

15. The method of claim any one of the previous claims, wherein at least one test is adapted for usage on at least one mobile device.

16. The method of any one of the previous claims, wherein at least one parameter is obtained from at least one sensor comprised in at least one wearable device.

17. The method of claim 16, wherein the wearable device is a smartwatch and/or a fitness tracker.

18. The method of any one of the previous claims, wherein at least two data points in at least one parameter differ in time of obtainment at least: i) 1 week for cognitive skill ii) 12 hours for heart rate variability, iii) and/or 24 hours for sleep.

19. The method of any one of the previous claims, wherein at least one step comprises the use of a computer implemented method.

20. The method of claim 19, wherein step (a) comprises the use of an application programming interface (API) to obtain parameters.

21. The method of claim 19 to 20, comprising oversampling techniques in particular the oversampling techniques SMOTE, ADASYN.

22. The method of claim 19 to 21, wherein the computer implemented method comprises the use of machine4earning.

23. The method of claim 19 to 22, wherein the generation of the predefined pattern comprises using a machine learning technique.

24. The method of claim 22 or 23, wherein the machine learning technique comprises the use of for Logistic regression, CART, Bagging, Random Forest, Gradient Boosting, Linear Discriminant Analysis, Gaussian Process Classifier and/or Gaussian NB for classification.

25. The method of claim 22, 23 or 24, wherein the machine learning technique, wherein the machine learning technique comprises the use of Linear, Lasso, Ridge, ElasticNet, KNN, DecisionTree, SVR, AdaBoost, GradientBoost, Random Forest, and/or ExtraTrees for regression.

26. The method of claim 22, 23, 24 or 25, wherein the machine learning technique comprises a deep learning technique, in particular simple multilayer perceptron (MLP) and/or convolutional neural networks for classification and/or regression.

27. The method of claim 22 to 26, wherein the machine learning technique used for generation of the predefined pattern comprises the use of a score indicative of the probability of the subject to develop a mood disorder.

28. The method of any one of the previous claims, wherein the reference signature obtained from at least one reference subject is based on a standardized mood disorder questionnaire score.

29. The method of claim 28, wherein the standardized mood disorder questionnaire score comprises the composite Maslach Burnout Inventory score, and/or the GAD-7 score, and/or the PHQ-9 score.

30. The method of any one of the previous claims, wherein the method further comprises administering to the subject having a score indicative for a mood disorder a pharmaceutically effective amount of a pharmaceutical product for the treatment of a mood disorder.

31. A pharmaceutical product for use in the treatment of a mood disorder in a subject, wherein the subject has a score indicative for a mood disorder determined according to any one of the claims 1 to 29.

32. A method for monitoring a mood disorder the method comprising the steps of:

1 ) determining a score indicative of the probability of a subject to develop a mood disorder using the method of any one of the claims 1 to 29 at a first timepoint;

2) determining a score indicative of the probability of a subject to develop a mood disorder using the method of any one of the claims 1 to 29 at a second timepoint;

3) comparing the score of step (1 ) with the score of step (2); and

33. A library comprising a score indicative of the probability of the subject to develop a mood disorder determined according to any one of the claims 1 to 29.

34. A storage device comprising computer-readable program instructions to execute the method according to any one of the claims 1 to 29 or 32, preferably additionally comprising the library according to claim 33.

35. A system for determining a score indicative for a mood disorder comprising at least one sensor for obtainment of at least one parameter indicative for a mood disorder and the storage device of claim 34.

36. A server comprising the storage device of claim 34, at least one processing device, and a network connection for receiving at least one parameter indicative of the state of mood of the subject and/or at least one weighted set of data points of the parameter(s).

37. The method of any one of the claims 1 to 30 or 32, the pharmaceutical product for use according to claim 31 , the storage device of claim 34, the system of claim 35, the server of claim 36 or the library according to claim 33, wherein the mood disorder is at least one disorder selected from the group of burnout, generalized anxiety disorder, and depression.

38. The method of claim 37, the storage device of claim 37, the pharmaceutical product for use according to claim 37, the system of claim 37, the server of claim 37 or the library according to claim 37, wherein the mood disorder is burnout.

Figures

Figure 1

IX

Score indicative of the probability of the subject to develop a mood disorder

Ύ X reference weighted set of signature parameters

1 predefined pattern parameters self-reported autonomic subject nervous system interpreted sensor da results from tests background data biomarkers tasets

1/4