WO2012025622A2 - Monitoring method and system for assessment of prediction of mood trends - Google Patents

Monitoring method and system for assessment of prediction of mood trends Download PDF

Info

Publication number
WO2012025622A2
WO2012025622A2 PCT/EP2011/064742 EP2011064742W WO2012025622A2 WO 2012025622 A2 WO2012025622 A2 WO 2012025622A2 EP 2011064742 W EP2011064742 W EP 2011064742W WO 2012025622 A2 WO2012025622 A2 WO 2012025622A2
Authority
WO
WIPO (PCT)
Prior art keywords
data
patient
system
unit
features
Prior art date
Application number
PCT/EP2011/064742
Other languages
French (fr)
Other versions
WO2012025622A3 (en
Inventor
Rita Paradiso
Anna M. Bianchi
Sergio Cerutti
Olaf Schleusing
Philippe Renevey
Enzo Pasquale Scilingo
Antonio LANATÀ
Original Assignee
Smartex S.R.L.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority to CH13842010 priority Critical
Priority to CH01384/10 priority
Application filed by Smartex S.R.L. filed Critical Smartex S.R.L.
Publication of WO2012025622A2 publication Critical patent/WO2012025622A2/en
Publication of WO2012025622A3 publication Critical patent/WO2012025622A3/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F19/00Digital computing or data processing equipment or methods, specially adapted for specific applications
    • G06F19/30Medical informatics, i.e. computer-based analysis or dissemination of patient or disease data
    • G06F19/34Computer-assisted medical diagnosis or treatment, e.g. computerised prescription or delivery of medication or diets, computerised local control of medical devices, medical expert systems or telemedicine
    • G06F19/3418Telemedicine, e.g. remote diagnosis, remote control of instruments or remote monitoring of patient carried devices
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/50ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H10/00ICT specially adapted for the handling or processing of patient-related medical or healthcare data
    • G16H10/60ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records

Abstract

There is provided a method for remote identification of signal trends indicating detection and prediction of critical events for patients affected by mood disorders, wherein the method comprises the steps of: - acquiring data of the patient; - processing the acquired data; - interpreting the processed data by means of a comparison with predefined data sets; and - providing a feedback concerning the mental state of the patient.

Description

TITLE

Monitoring Method and System for assessment of prediction of mood trends

TECHNICAL FIELD

The present invention relates to a method and a system identification of signal trends indicating detection and prediction of critical events for patients affected by mood disorders according to the preamble of claim 1 and 5.

SUMMARY OF THE INVENTION

There is provided a method for remote identification of signal trends indicating detection and prediction of critical events for patients affected by mood disorders, wherein the method comprises the steps of:

- acquiring data of the patient;

- processing the acquired data;

- interpreting the processed data by means of a comparison with predefined data sets; and

- providing a feedback concerning the mental state of the patient.

Furthermore a Monitoring system is provided for remote evaluation of the mental state of a patient, wherein this system comprising:

- a data acquisition unit to acquire data of the patient;

- a processing unit to process the acquired data, whereby processed data is obtainable;

- an interpretation unit having a reference database with predefined data sets to interpret the processed data by means of a comparison; and

- a feedback unit to provide a feedback concerning the mental state of the patient.

Further embodiments of the invention are laid down in the dependent claims. BRIEF DESCRIPTION OF THE DRAWINGS

Preferred embodiments of the invention are described in the following with reference to the drawings, which are for the purpose of illustrating the present preferred embodiments of the invention and not for the purpose of limiting the same. In the drawings,

Fig. 1 shows an embodiment of a monitoring system;

Fig. 2a shows a diagram for assessing the status of a patient;

Fig. 2b shows the overall system

Fig. 3 a shows a model representing the relation between the factors influencing patient state and the observed signals;

Fig. 3b shows an embodiment of the data acquisition unit;

Fig. 4a shows the extraction of the features related the control of the nervous system on the observed signals;

Fig. 4b shows the overall scheme of processing and interpretation unit; and

Fig. 5 shows the extraction of the indicator/predictor of bipolar state from features extracted from unsupervised monitoring signals.

DESCRIPTION OF PREFERRED EMBODIMENTS

The present invention relates to a method and a system for remote identification of signal trends indicating detection and prediction of critical events for patients affected by mood disorders, in particular to evaluate the mental state of a patient. Reference is made to figures 1 to 5.

The method comprises the following steps:

1. acquiring data of the patient;

2. processing the acquired data;

3. interpreting the processed data by means of a comparison with predefined data sets; and

4. providing a feedback concerning the mental state of the patient.

Data acquisition according to the first step is achieved by means of a personal multi- parametric monitoring system. Such a personal, multi-parametric monitoring system or platform is preferably based on textile platforms and portable sensing devices for the long term and short-term acquisition of data from patients affected by mood disorders. The system is based on wearable devices for monitoring of physiological signals and portable devices for voice analysis, behavioural index correlated to patient state, and biochemical markers. The devices that are used to acquire data can be designated as data acquisition unit.

The acquired data are then processed and analyzed in a platform, according the second step. Preferably said platform can optionally take into consideration the electronic health records (EHR) of the patient, a personalized data reference system, as well as medical analysis in order to verify the diagnosis and help in prognosis of the disease. Such a platform can be designated as processing unit. The processing unit can process and analyze the signals off-line, in real time or by stages. In case an off-line scheme is used the data can be stored within a storage element in the processing unit for further use.

With the third step, said processed data are compared by means of an interpretation unit that comprises a reference annotated database. The reference database comprises a plurality of predefined data sets from patients with mental disorders and healthy individuals, that allow the step of comparing the processed data with the reference data, in order to provide a feedback based on the interpretation of the acquired data to the physician and the patient.

The reference database with annotated physiological and behavioural signals can be used and integrated in the platform offering a basis for content-based searches, tools for feature extraction and signal processing, as well as tools for integration with Electronic Health Record information. Moreover, it constitutes the basis for the extraction of correlations between combinations of signals and disorder status leading to the identification of signal patterns and trends predicting a critical state of the disease. The reference database is thereby part of the interpretation unit.

Communication and feedback to the patient and physician according the fourth step, is preferably done through a closed loop approach that facilitates disease management by giving patients support and assuring interaction between patient and physician. The feedback is provided through a feedback unit, which can be a computer screen, a mobile device such as a mobile phone or a tablet computer. Any other device can also be used. However, the device should be equipped with signalling means showing the present state of the patient.

The data acquisition unit can be provided with a plurality of devices, which can be used to acquire the following data:

- physiological data

- behavioural data and other parameters correlated to the general status of the patient voice

Physiological data that are collected by wearable devices include Electrocardiogram (ECG) signal, Heart Rate Variability (HRV), and Respiratory Signal (RS).

Behavioural data or behavioural indexes are based on the correlation of data deriving from inertial sensors (e.g. accelerometers in the sensorized T-Shirt, in the smartphone etc.), biochemical measurements and clinical information from the Electronic Health Record and the detection of attitudinal indicators (social interaction, sleep quality, activity and gesture, number, duration and activity during phone calls) that are implemented to extrapolate predictive indexes. Moreover personal information reported on the patient diary, like for instance the amount of smoked cigarettes, the frequency of social events, the increasing of superfluous expenses, the variation of weight, or other information that are relevant for the detection of changes in the patient habits, are also considered as variable for the processing unit.

Other parameters regarding the general behaviour of the patient that are taken into consideration include the study of sleep pattern alteration, peripheral measures in cardiovascular and respiratory functioning, as well as the secretion of stress-related hormones, including change in the diurnal variations of all these measures (circadian rhythms).

Patients interact with the system through a user interface unit such as smart phones or similar platforms or devices. Preferably the user interface has to be fed or updated by means of the interpretation unit in case it is noted that further information about the patient is necessary, thereby the patient is provided with a request for feedback such as questionnaires or with advices such as daily diary. This functionality can also comprise administrating physician advices, combined with traditional functionality.

Figure 3 a shows a possible model representing the relation between influencing factors and observed signals. The response of the body to the external/internal conditions can be represented as an input-output model. Such model is useful to represent how the different input factors and regulation processes interact to produce the observed signal changes. In a very simplified embodiment two main influencing factors can be applied:

Firstly the environment refers to the whole surrounding of the subject that produces voluntary or involuntary reaction of the body. It includes all the events that can occur in everyday life as well as all the changes of ambient conditions that modify regulatory mechanisms of the body functions. Voluntary reactions include for instance body activity, displacement, interaction with other people and dialogue.

Secondly, the circadian rhythms refer to the natural rhythms of body that are synchronised to the 24 hours light-dark periodicity.

These influencing factors modify the measured signals via the control that is operated by the nervous system. Two different mechanisms take part in this control: the autonomic nervous system and the central nervous system. The autonomic nervous system (ANS) is involved in the non-voluntary or unconscious control of the body state. It adapts the physiological parameters to ensure homeostasis (stable conditions) in the whole body when some change occurs or to prepare the body to react to some situation. Part of the central nervous system (CNS) is involved in the voluntary or conscious control of the body state. The central and autonomic systems are linked and influence each other; for example an emotional stimulation, which is only perceived by the conscious part of central nervous system, has an influence on the regulation mechanisms of the autonomic nervous system. Both the controls operated by autonomic and central nervous systems operate on time scales ranging from fraction of second to hours. For longer time scales (hours to days) the endocrine system (hormonal regulation) is involved in the regulation of body state. This system is in relation with the autonomic and central nervous systems and influences their behaviours.

This description of the control operated by the nervous system is simplistic and incomplete but the objective is to highlight and discover the relations that exist between mood state and the observed signals and not to analyze the mechanisms that take place in the nervous system. The aim of this discussion about the control operated by the nervous system is to motivate the features, extracted from the observed signals that are based on the actual knowledge about nervous system control.

Finally, the output of the proposed model consists of the signals that are recorded and that are expected to be influenced by the mechanisms described in the proposed model.

During the long-term acquisition, the system collects data from patients in different states of the illness (mania or depression episodes, remission). The assessment of the patient status is done by the medical professionals in a controlled or remote environment and used to annotate the recorded signals, which means that the medical professionals obtain for each of the patients further data to be integrated with the annotated database and used as reference data. The collected data along with the subjective annotations is also recorded in said reference database, where information from the EHR such as medication, patient history and exams, are integrated.

Automatic comparison with the reference data-base is done with a double aim: i) to put into evidence differences between the present status of the user and the control normal group; ii) to put into evidence any deviations of the user's status from his/her previous status. Methodologies of clustering and automatic classification will be used at this purpose. Semantic technology is used to enable content-based searches.

The assessment of the patient is done by professionals in clinics or through a medical visit. This is useful for the definition of a baseline for the specific patient and to provide a classified status that will be the reference for other classifications.

However, the assessment of the patient status by the medical professional is optional. In fact, the system allows the estimation of a reference status from the patient. In such a case, data are acquired from a patient and compared with control groups in the database, in order to define his/her state. Automatic unsupervised clustering is done day- by-day as soon as new data will be available, in order to detect any changes from the previous status and to define personalized statuses of a specific subject. The latter is particularly advantageous in case reference data from patients are collected in different states of illness.

Data are processed, classified and correlated to assess patient's status (from health professional annotations and other clinical findings) on the base of the measured parameters.

A professional environment is used for monitoring, through easily formulated queries; medical professionals are able to view current patient data as well as information extracted from electronically stored medical files. The physicians are able to follow patient response to treatment and to be alerted in case of critical predicting indexes. The professional loop helps the psychiatrics in preventing relapse by the early detection of change in behaviours, sleep, physiological or biochemical signs.

Reference is now made to figure 2a. The system foresees the acquisition of patient baseline during the initial phase of use. With other words: before the method is applied continuously an initial data acquisition step is executed, whereby a baseline will be determined which serves as reference for the predefined data sets. Said baseline is obtained by acquiring data in a controlled environment, typically in a hospital or medical room, or in a naturalistic environment, data have to be checked by physicians that identify the status of the patient. The baseline data is stored in the database as reference data.

With regard to figure 2b the procedure for assessing the mental state of the patient will now be explained. Preferably a system according to figure 2b is used to assess the mental state. The system here comprises a data unit to acquire data from the patient, a processing unit and an interpretation unit. Furthermore the system may optionally comprise a feedback unit which provides the medial professional or any other person including the patient itself with various feedback and/or reminder messages. The feedback unit may also comprise a professional portal which provides the person to be reminded with reminder messages or any other related information. The procedure comprises mainly three steps:

1) Data acquisition: data acquisition is performed remotely at home and in a naturalistic environment; the recording system allows the user to conduct his/her daily life, without interfering with his/her usual activities. The users wear parts of the sensors, parts are portable (electronics unit, smart phone) and other information can be acquired remotely (biochemical screening). Thereby devices can be used for the management of diseases such as heart diseases, chronic obstructive pulmonary disease, metabolic disorders and diabetes. The use of such techniques for psychiatric and mental disorders comprises a variety of advantages as the full spectrum of different data is acquired which leads to an integral data cloud, therefore this is an innovative application. Further, the recorded parameters and signals allow clinical evaluation of metal disorders and their use in such an application is another innovative element of the system.

2) Processing of the acquired data: the use of multiple sensors for long periods of time produces huge amounts of data (i.e. a data cloud) that require to be properly processed in order to provide the useful information. Part of the processing procedures is implemented on the electronics embodied in the sensors, while part of the data are transmitted to an external device (PC, PDA, smart phone, etc.) and preferably to a central server for further processing (i.e. data-mining) . After a generic and an ad-hoc pre-processing step, the processing procedures are intended to extract parameters able to provide descriptive information on the patient status. Such parameters are obtained through mono-variate and multi-variate analysis of the recorded data, including also results from questionnaires, clinical and medical information, patient characteristics (age, gender, etc.).

The methodologies used for signal processing and parameter extraction depend on the specific signals.

• Heart rate variability is analyzed through the procedures for the assessment of the autonomic regulation of the heart rate: time domain and frequency domain parameters that have already been proven to be related to the sympatho-vagal balance.

Further, more sophisticated techniques are also used: beat-to-beat time variant spectral analysis, for the evaluation of the non-stationary conditions and especially for the sleep classification and analysis; cross-spectral analysis of heart rate variability and respiration allows the calculation of the frequency coherence between the two signals. This is related to the sympatho-vagal balance and to the "stress" conditions of the user; long term correlation analysis (based on the calculation of non linear indices, such as entropy, cross entropy, detrended fluctuation analysis, complexity indices, power low slope, etc,) allows to put into evidence long term regulatory mechanisms that may play a role in mood disorders. Respiration is analysed in relation to its amplitude variations and its frequency, these parameters are evaluated in different period lengths during day and night, in order to obtain trends. Further, respiration is analysed in relation to the heart rate variability.

Electro-dermal response is analyzed in order to identify relevant events during a day. Since this signal is strictly event-related, it has to be analyzed during specific time window (e.g. during the periodic voice acquisition). However, the mean value is retained for further evaluation during the long-term acquisition.

Movement and activity are recorded through accelerometers embedded on the wearable system (t-shirt) and integrated in the portable device (smart phone). The quantity of activity can be calculated as total amount during the day and during the night (this is related to the sleep quality), as well as trend during the 24h in intervals of different length (lh, l/2h, etc.). Moreover the inertial sensors in the electronic unit provide information related to the posture during sleep, while the inertial sensors in the mobile platforms provide information about the gesture during the phone calls.

Heart rate variability, respiration and movement, recorded during the night, are processed in time and in frequency domain, through spectral and cross-spectral analysis, for the assessment of sleep.

Voice signal is recorded through smart phone and/or a personal computer; mainly two indicators of the patient's mood influence speech: the social interaction and the psychomotor control. In particular, speech features are extracted from temporal and spectral characteristics of the speech as well as from the voice quality. This comprises the duration of speech and pause segments, the intensity of speech and its dynamic range, information related to the spectrum and temporal structure of articulation, the fundamental frequency of the voice and the voice quality (e.g. harshness, tonality, etc.).

• Biochemical data could be acquired as measurements in order to put into evidence circadian trends, while processing is only related to calibration and normalization before statistical analysis.

As mentioned above, coupling measures and multivariate techniques are implemented in order to extract information, otherwise kept hidden.

3) Interpretation of the processed data. A model is used in order to properly correlate the parameters obtained at point 2) with the mental status of the user. The model is based on previously acquired data considering the intra- subject and the inter-subject records. The inter-subject data are processed according to the clinical evaluation, i.e. definition of patient clusters that shows the same clinical status. Basically, the model is constituted by a parameter selection phase followed by supervised and unsupervised expert systems able to perform the clinical classification (i.e. status identification). Therefore, the model is constituted by linear and non-linear transformations of the multi-dimensional parameter space obtained at point 2) along with several association rules and/or fuzzy rules. In this phase, the state-of-the-art of the pattern recognition techniques is implemented in order to obtain robust automatic classification. Physician's classification is used to train the supervised part of the model.

The model is based on average data coming from literature and from previously acquired data classified by physicians. The model is identified through multiparametric statistical analysis of calculated parameters and features, recorded data, clinical data, patient's data and physician's classification. The approach for the definition of the model is supervised classification. The model is thereby integrated in the reference database. The patient status is defined as a point in a multi-dimensional space defined by the features identified during the supervised classification phase. Modification of one or of a set of features will identify a trajectory towards a new point in the multi-dimensional space and then a trend towards a new state.

Furthermore the generic model is personalized for each specific user according the two different procedures:

a. before hospital delivery (controlled environment) the patient undergo an "assessment" procedure, aimed to provide his/her baseline parameters and his/her variability intervals. The average model is then modified accordingly;

b. at home (naturalistic environment), at the beginning of monitoring period, proper procedures, under the physician's control, allows the adaptation of the generic model and the definition of the personalized baseline.

Based on the model an automatic classification of the patient's status is implemented. The classification is performed through classification procedures like neural networks, linear or quadratic discriminants, nearest neighbours, etc.

During the monitoring phase significant deviations from the baseline, indicative of deviations from the reference status, are notified to the physician and, if needed or required, all the data related to an alarm event (raw signals, features, score of questionnaires, etc.) are transmitted to a specialized centre for further and more accurate evaluations.

The reference database comprising said predefined data sets with annotated physiological and behavioural signals from patients is automatically updated. The database allows data mining for searching associations of the recorded data and calculated parameters with mental status and for the further development of the interpretative model. Further, the database allows a clustering of patients into subgroups, according to their different characteristics (i.e. age, gender, therapy, etc.) and is used for a more focused personalization of the interpretative model.

Data acquisition unit

With regard to figure 3b the data acquisition unit combines wearable devices 1 and portable devices 2. The monitoring platform is based on the integration of sensors 3 for physiological and behavioural data into the remote monitoring system. The system includes portable devices for voice analysis, in particular for analysing voice behaviour, and for the extrapolation of behavioural indexes correlated to mental illness. A typical configuration for the data acquisition unit are clothes with integrated fabric electrodes 3, with a minimal number of two for ECG monitoring, four for ECG and respiration through impedance or pneumography monitoring, six for ECG, respiration through impedance and GSR monitoring. On the shirt a piezoresistive fabric sensor is integrated in the fabric to monitor the movement of the thorax due to respiration, the sensor is realised with polyamide fibers coated with carbon, combined with elastane, the preferred configuration is at thorax level at the edge of the sternum. The system is realized with a cut and sews process, where the sensing region is made through seamless technology, the fabric electrodes are preferably placed on the thorax and a multilayer structure is used to increase the pressure and the amount of sweat. The electrodes are realised with metal fibers, preferably stain steel fibers or textile fibers coated with silver. The sensing region is more elastic and is combined with regions that are lighter and more comfortable. The electronic device and the connectors are embedded in the system; the preferred position is on the sternum in a pocket.

The textile sensors garment is connected through a standard jack connector to the portable electronics that is able to acquire, evaluate and transmit the following signals and parameters: 1 lead ECG, Respiratory signal through bio impedance measurement, plehtysmography through piezoresistive sensors and 3D accelerometers, from these signals the following parameters are evaluated: Heart rate, Breathing rate, Breathing Amplitude, Heart rate variability, Activity classification, posture during night. Furthermore it is also possible to acquire data concerning the activity of the patient, such as energy or classification of activity.

Signals can be transmitted to a smart phone, to a pc, or to the server through a gateway.

Patients interact with the system through a user interface unit such as smart phones or similar platforms or devices. Preferably the user interface is fed or updated by means of the interpretation unit in case it is noted that further information about the patient is necessary, thereby the patient is provided with a request for feedback such as questionnaires or with advices such as daily diary. This functionality can also comprise administrating physician advices, combined with traditional functionality. Processing and interpretation unit

The acquired data are processed, classified and correlated to assess patient's status (from health professional annotations and other clinical findings) on the base of the measured parameters. Software system platform consists of two lines of activity. The first one is based on extracting significant parameters from raw signals, and determining the correlation between them. The second line is constituted of algorithms dedicated to extracting from the generic features an indicator/predictor that is related to the state of bipolar patients and their evolution.

With reference to Figure 4b, the method foresees that once data are acquired, a preprocessing is done in order to remove the noise/artefact. Part of the pre-processing is performed in the wearable system, such as basic filtering, HRV extraction and format conversion (i.e. ASCII conversion).

In addition, a robust automatic artefact removal algorithm is applied in order to enhance the signal-to-noise ratio. This algorithm removes automatically movement artefacts from long time duration signals. It is based on filtering techniques and relies on the knowledge of the bandwidth of movement artefacts, such as reported in the literature. The algorithm uses an arbitrary window, whose length must be submultiples of the entire signal dynamics. Inside the window, the algorithm detects the envelop of maximum a minimum points of the signal, respectively, and from these it determines the smoothed mean envelope. A histogram is obtained from the mean envelope and a threshold is chosen as its 95th percentile. By analyzing the smoothed mean envelop and using the calculated threshold the movement artefacts are detected and removed. All operations are repeated for each window up to cover the whole signal length. After the pre-processing step, the mono- variate and the multivariate analysis are performed.

Regarding the first line of software activity, it is proposed to do features extraction in a two-stage process. The first step, named generic feature extraction, consists in the extraction of "standards features", known to be correlated with the control of nervous system, from the signals. During this step, all the signals are processed separately. The second step, named specific feature extraction, consists in the combination of previously extracted features to determine new features that are directly related to the control of nervous system and therefore should be strongly related to mood disorders.

One possible embodiment of the feature extraction is shown in figure 4a. In general words the feature extraction can be described as follows, whereas reference is made to all of the figures.

Generic feature extraction

A pre-processing stage is performed before feature extraction in order to remove noise when possible and to discard segments where no information is present. The extraction of generic features is divided into four main groups of features.

The first group of features, the extraction of physiological features, concerns the signals that are related to the autonomic nervous system control, namely the ECG, the GSR (galvanic skin response), the respiration, activity during the night and during the day. The features that are extracted from the signals are those that are known to be correlated with the autonomic nervous system control; for the ECG, the heart rate based on RR intervals as well as the variability of these intervals. Similarly, frequency, amplitude and variations of these two quantities is extracted from the respiration signal and GSR

The second group of features, the estimation of body activity, is related to the classification of physical activity. It is based on the signals of the body-worn accelerometers and gyroscopes sensors. This classification results in an identification of subject position (lying, sitting or standing), as well as a classification of the activity into several classes (at rest, non-rhythmical activity, walking, running, etc.). Additionally, some features related to the psychomotor control are extracted. These features are obtained during rhythmical activities (walking and running) and include motion frequency, stability of motion frequency, amplitude of the motion signal and stability of this amplitude.

Finally, the third group of features, the extraction of speech features, concerns the extraction of features from speech signals. Speech contains two kinds of information relative to the mood of the patients: the social interaction and the psychomotor control. Features related to the social interaction are based on the estimation of the pro-activity of the patient during conversations and the dynamical range of the speech. Additionally, this information could be combined with those of noise sensor, when is included in the data unit, to differentiate direct interaction with other people from indirect ones as for example telephone calls. Discarding the actual content of the speech preserves the privacy of the patient.

Secondly, the psychomotor control has a major influence on various features extracted from speech. The respective parameters related to psychomotor control are extracted from the answers given to a questionnaire, recorded on a daily basis or even several times per day. Cognitive fatigue can be observed in the temporal characteristics of speech through a delay in answers, reduced speech segment duration and prolonged pause segment duration. Furthermore, the value of the fundamental frequency of the vocal fold vibration is conditioned by the tension of the vocal folds and therefore related to psychomotor control. The values of this frequency as well as the dynamics of its variation (the distribution of the fundamental frequency) are extracted during voiced sounds (e.g. vowels). This lack of precision also leads to a reduced harmony in the voice, which is reflected in the voice HNR (harmonics-to-noise ratio) feature. Other features concern the articulation of speech. The use of autoregressive modelling allows the separation of the excitation signal (produced by the lungs and the larynx) from the articulation (obtained by modifying dynamically the shape and size of the oral cavity). The articulation of speech contains information about psychomotor control of the articulators. The features that are extracted are the resonant frequencies and the dynamic of change between different positions of the articulator. It is expected to have a reduction of this dynamic as well as a reduction of reparability of the different phonemes when patients are depressed.

Specific feature extraction

The objective of the specific feature extraction is to combine the information from previously extracted features in order to obtain a new set of features that are directly related to nervous system control. Five different categories of features are defined.

The first category, the estimation of circadian rhythms, includes features that are related to 24-48 h periodicities that can be observed in physiological signals. For example it is possible to detect a modulation in the sympatho-vagal balance during day-night cycle. This modulation can be altered by pathologies and stress conditions. There are many features extracted form HRV and respiration signals that are able to quantify this circadian modulation. However many of them may be affected by different activities and events occurring during the daily life. In order to ensure repeatability, all the features have to be extracted during resting periods according to specific protocols (i.e. in the morning after wake up, after lunch, before going to sleep, etc.), or during more reproducible and controlled situations such as during sleep. The recorded signals related to circadian rhythms are the ECG, the respiration, the activity. Sleep is not directly observed by one sensor but sleep periods, sleep wake alternance, sleep architecture, etc. can be deduced from the combination of activity, ECG, respiration and ambient condition signals. For all the features derived from these signals, the estimation of circadian rhythms consists in the estimation of periodicity and range of variation of these features along the 24-h and in comparing this periodicity with reference 24 hour periodicities. Notably, the phase difference and phase shift are promising features for the determination of a bipolar state indicator.

The second category of features, the estimation of autonomic nervous system control, consists in the extraction of relations between the environment and the reaction of the body. For example, the heart rate increases when starting to walk. This variation of heart rate is controlled by the autonomic nervous system. If the functions of the autonomic nervous system are affected by bipolar state, one could expect a change of the relation between environmental conditions and regulated physiological values. Envisaged features are the rest heart rate and heart rate variability, relation between heart rate and the intensity of activity, variation of heart rate changes at activity onset and offset, respiration rate and intensity during activity and GSR change, HRV variations measured during different sleep stages (i.e. differences between REM sleep and delta sleep).

The third category of features, the estimation of social interactions, aims at determining an indicator of social interaction. This estimation is based on the signal from user microphone and from the external noise sensor. The user microphone signal allows the determination of the time when the patient is speaking. From this detection of vocal activity, statistics about average sentence length, duration of speech pauses, total duration of vocal activity per day, etc. can be established. Combining this information with those of the external noise sensor allow the determination of direct dialogue with someone from indirect dialogue {e.g. telephone calls). Additionally, the combination of activity and environment sensor allows the detection of social interaction taking place in a single location (the home of the patient) or if the patient moves to have social interactions.

Finally, the fourth category of features, the estimation of activity level, consists in the extraction of the average activity performed on daily time scale. This estimation is based on activity and ambient condition signals. Features extracted are representative of duration and intensity of activity as well as the ratio of inside/outside activity. Such information allows the estimation of the average energy expenditure per day; the addition of features from physiological signals, like ECG, could also be used to improve this estimation.

Detailed description of system classification of mood disorders is explained in the following and with reference to figure 5:

The extraction of bipolar state indicator is made in two steps: 1) a reference indicator is first derived from the annotated database. This consists in dividing the multi-dimensional space defined by signal features in regions each one related to a different clinical status. The reference indicator can be further personalized on the basis of clinician's diagnosis and self-evaluation tests. The definition of the reference indicator is firstly done in supervised way, meaning that the clinical observations of bipolar state are used to determine the regions in the multi-dimensional feature space that are associated to different clinical states. 2) In a second phase, (i.e. the proper monitoring phase), using recordings from patients not used in the development of the reference indicator, and without clinical evaluation, the indicator is extracted in an unsupervised manner and the result is compared with the reference indicator to estimate the present status of the user or to predict trends towards a new status.

Creation of reference indicator from clinicians' diagnoses and self-evaluation questionnaire

In order to extract an indicator of bipolar state from unsupervised monitoring signals, it is first required to determine a personalized reference indicator of patient state. The only available and validated estimation of patient state results from clinicians' diagnoses. Unfortunately, it is not possible to perform such an evaluation on every day or several times per day intervals. Personalized reference indicators can be obtained during the subjects' assessment phase in controlled environment, or during the first stages of the monitoring. Otherwise to overcome this limitation, a reference indicator is created from self evaluation results, by comparing these results with clinicians' diagnoses. The analysis of correlation between self evaluation test results and clinicians' diagnoses permits the creation of the personalized reference index of patient state. This reference indicator is obtained by finding the combination of answers to the self evaluation test that best matches the clinicians' diagnoses. This determination of the personalized reference indicator is performed separately for each patient because, due to the inter-patient differences, it will not be the same answers and the same combination of them that will give the best match. The advantage of this index is that it allows the determination of the evolution of patient state once a day or several times per day.

Interpretation unit: extraction of mood state indicator from recorded signals

The modification of mood state will modify the regulation process performed by the nervous system. It is expected that this effect is observable on specific features presented before and that variations induced by mood disease will be superposed to normal ones. In order to highlight these variations, the first step consists in the analysis of the variations of the specific features over short (hours), mid (days) and long (weeks) terms. Then the correlation or relation that exists between these variations and the reference indicator of bipolar state is analyzed. The results of this analysis is then used to construct an indicator of patient state based only on features extracted from recorded signals, and to use it to predict the evolution of the patient state.

Preferably such a system is worn 24 hours by the patient. This means that monitoring of the patient can take place in a natural setting, through a naturalistic approach will provide parameters, indexes and trends; that will be used to assess mood status, to support patients, to predict and anticipate treatment response at its earlier phases, to prevent relapse and to alert physicians in case of critical events, as depicted in Figure 1.

In summary a system or platform for remote evaluation of the mental state of a patient, wherein said system comprising

- a data acquisition unit to acquire data of the patient; - a processing unit to process the acquired data, whereby processed data is obtainable;

- an interpretation unit having a reference database with predefined data sets to interpret the processed data by means of a comparison; and

- a feedback unit to provide a feedback concerning the mental state of the patient.

Claims

1. A method for remote identification of signal trends indicating detection and prediction of critical events for patients affected by mood disorders, wherein the method comprises the steps of:
- acquiring data of the patient;
- processing the acquired data;
- interpreting the processed data by means of a comparison with predefined data sets; and
- providing a feedback concerning the mental state of the patient.
2. Method according to claim 1, characterized in that the step of processing of the acquired data comprises a sub-step of feature extraction.
3. Method according to claim 2, characterized in that the feature extraction comprises the extraction of generic features and the extraction of specific features, wherein the generic feature extraction consist in the extraction of standard features and wherein the specific feature extraction consists in the extraction of dynamic determined features, in particular in combination with the generic feature extraction.
4. Method according to any of the preceding claims, characterized in that before the method is applied continuously, an initial data acquisition step is executed, whereby a baseline will be determined which serves as reference for the predefined data sets.
5. Monitoring system for remote evaluation of the mental state of a patient, wherein this system comprising:
- a data acquisition unit to acquire data of the patient;
- a processing unit to process the acquired data, whereby processed data is obtainable;
- an interpretation unit having a reference database with predefined data sets to interpret the processed data by means of a comparison; and
- a feedback unit to provide a feedback concerning the mental state of the patient.
6. System according to claim 5, wherein the data acquiring unit comprises a textile platform and portable sensing devices.
7. System according to one of the claims 5 to 6, wherein the data acquiring unit acquires data chosen from the group of physiological signals, biochemical markers, voice analysis and behavioural index correlated to patient state.
8. System according to one of the claims 5 to 7, wherein the system comprises an interface unit which allows the user to provide feedback to the system, wherein said user feedback is also processed by the data acquiring unit.
9. System according to one of the claims 5 to 8, wherein the data acquiring unit or the processing unit comprises storage means to store the acquired data for further use.
10. System according to one of the claims 5 to 9, wherein the interpretation unit provides a warning in case of an abnormal deviation from the processed data and the reference data, wherein the warning is submitted to the feedback unit.
11. System according to one of the claims 5 to 10, wherein the system includes portable devices allowing remote monitoring.
PCT/EP2011/064742 2010-08-27 2011-08-26 Monitoring method and system for assessment of prediction of mood trends WO2012025622A2 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CH13842010 2010-08-27
CH01384/10 2010-08-27

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
EP11772904.6A EP2609533A2 (en) 2010-08-27 2011-08-26 Monitoring method and system for assessment or prediction of mood trends

Publications (2)

Publication Number Publication Date
WO2012025622A2 true WO2012025622A2 (en) 2012-03-01
WO2012025622A3 WO2012025622A3 (en) 2012-06-07

Family

ID=45090911

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2011/064742 WO2012025622A2 (en) 2010-08-27 2011-08-26 Monitoring method and system for assessment of prediction of mood trends

Country Status (2)

Country Link
EP (1) EP2609533A2 (en)
WO (1) WO2012025622A2 (en)

Cited By (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2014182930A2 (en) 2013-05-09 2014-11-13 Otsuka America Pharmaceutical, Inc. Systems and methods for administering health care systems
WO2015003247A1 (en) * 2013-07-10 2015-01-15 1Datapoint Diagnostics Inc. Systems and methods relating to subject monitoring, assessment and treatment
CN104584017A (en) * 2012-08-16 2015-04-29 金格输入输出有限公司 Method for modeling behavior and health changes
US20150148621A1 (en) * 2013-11-22 2015-05-28 Grant Joseph Sier Methods and systems for creating a preventative care plan in mental illness treatment
EP3072446A1 (en) 2015-03-26 2016-09-28 Digital for Mental Health Mental suffering monitoring system
EP2660745A3 (en) * 2012-05-01 2016-10-26 Bernie Almosni Mental health digital behavior monitoring system and method
WO2017062544A1 (en) * 2015-10-06 2017-04-13 University Of Pittsburgh-Of The Commonwealth System Of Higher Education Method, device and system for sensing neuromuscular, physiological, biomechanical, and musculoskeletal activity
US10014077B2 (en) 2012-08-16 2018-07-03 Ginger.io, Inc. Method and system for improving care determination
US10068059B2 (en) 2012-08-16 2018-09-04 Ginger.io, Inc. Method and system for providing automated conversations
US10068060B2 (en) 2012-08-16 2018-09-04 Ginger.io, Inc. Method for modeling behavior and psychotic disorders
US10068670B2 (en) 2012-08-16 2018-09-04 Ginger.io, Inc. Method for modeling behavior and depression state
US10102341B2 (en) 2012-08-16 2018-10-16 Ginger.io, Inc. Method for managing patient quality of life
US10242754B2 (en) 2012-08-16 2019-03-26 Ginger.io, Inc. Method for providing therapy to an individual
US10265028B2 (en) 2012-08-16 2019-04-23 Ginger.io, Inc. Method and system for modeling behavior and heart disease state
US10269448B2 (en) 2012-08-16 2019-04-23 Ginger.io, Inc. Method for providing patient indications to an entity
EP3397146A4 (en) * 2015-12-29 2019-08-07 Lifeq Global Ltd Cardio-kinetic cross-spectral density for assessment of sleep physiology

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9189596B2 (en) * 2005-06-29 2015-11-17 National Ict Australia Limited Measuring cognitive load
US8655817B2 (en) * 2008-02-20 2014-02-18 Digital Medical Experts Inc. Expert system for determining patient treatment response

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
None

Cited By (22)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2660745A3 (en) * 2012-05-01 2016-10-26 Bernie Almosni Mental health digital behavior monitoring system and method
US10068670B2 (en) 2012-08-16 2018-09-04 Ginger.io, Inc. Method for modeling behavior and depression state
CN104584017A (en) * 2012-08-16 2015-04-29 金格输入输出有限公司 Method for modeling behavior and health changes
US10269448B2 (en) 2012-08-16 2019-04-23 Ginger.io, Inc. Method for providing patient indications to an entity
EP2884888A4 (en) * 2012-08-16 2016-04-20 Ginger Io Inc Method for modeling behavior and health changes
US10265028B2 (en) 2012-08-16 2019-04-23 Ginger.io, Inc. Method and system for modeling behavior and heart disease state
US10242754B2 (en) 2012-08-16 2019-03-26 Ginger.io, Inc. Method for providing therapy to an individual
US10276260B2 (en) 2012-08-16 2019-04-30 Ginger.io, Inc. Method for providing therapy to an individual
US10068060B2 (en) 2012-08-16 2018-09-04 Ginger.io, Inc. Method for modeling behavior and psychotic disorders
US10102341B2 (en) 2012-08-16 2018-10-16 Ginger.io, Inc. Method for managing patient quality of life
US9836581B2 (en) 2012-08-16 2017-12-05 Ginger.io, Inc. Method for modeling behavior and health changes
US10014077B2 (en) 2012-08-16 2018-07-03 Ginger.io, Inc. Method and system for improving care determination
US10068059B2 (en) 2012-08-16 2018-09-04 Ginger.io, Inc. Method and system for providing automated conversations
US10068672B2 (en) 2012-08-16 2018-09-04 Ginger.io, Inc. Method for modeling behavior and health changes
EP2994876A4 (en) * 2013-05-09 2017-01-04 Otsuka America Pharmaceutical, Inc. Systems and methods for administering health care systems
WO2014182930A2 (en) 2013-05-09 2014-11-13 Otsuka America Pharmaceutical, Inc. Systems and methods for administering health care systems
WO2015003247A1 (en) * 2013-07-10 2015-01-15 1Datapoint Diagnostics Inc. Systems and methods relating to subject monitoring, assessment and treatment
US20150148621A1 (en) * 2013-11-22 2015-05-28 Grant Joseph Sier Methods and systems for creating a preventative care plan in mental illness treatment
WO2016151135A1 (en) 2015-03-26 2016-09-29 Digital For Mental Health Mental suffering monitoring system
EP3072446A1 (en) 2015-03-26 2016-09-28 Digital for Mental Health Mental suffering monitoring system
WO2017062544A1 (en) * 2015-10-06 2017-04-13 University Of Pittsburgh-Of The Commonwealth System Of Higher Education Method, device and system for sensing neuromuscular, physiological, biomechanical, and musculoskeletal activity
EP3397146A4 (en) * 2015-12-29 2019-08-07 Lifeq Global Ltd Cardio-kinetic cross-spectral density for assessment of sleep physiology

Also Published As

Publication number Publication date
WO2012025622A3 (en) 2012-06-07
EP2609533A2 (en) 2013-07-03

Similar Documents

Publication Publication Date Title
Parkka et al. Activity classification using realistic data from wearable sensors
US8795170B2 (en) Residual based monitoring of human health
CA2809764C (en) Systems and methods for respiratory event detection
US8306610B2 (en) Method and apparatus for analysis of psychiatric and physical conditions
Hravnak et al. Defining the incidence of cardiorespiratory instability in patients in step-down units using an electronic integrated monitoring system
US7558622B2 (en) Mesh network stroke monitoring appliance
US8622899B2 (en) Continuous monitoring of stress using self-reported psychological or behavioral data
Wilhelm et al. Emotions beyond the laboratory: Theoretical fundaments, study design, and analytic strategies for advanced ambulatory assessment
US8725462B2 (en) Data aggregation platform
US9189599B2 (en) Calculating and monitoring a composite stress index
US20140121540A1 (en) System and method for monitoring the health of a user
US20080208015A1 (en) System, apparatus and method for real-time health feedback on a mobile device based on physiological, contextual and self-monitored indicators of mental and physical health states
KR101850855B1 (en) Contactless and minimal-contact monitoring of quality of life parameters for assessment and intervention
Alberdi et al. Towards an automatic early stress recognition system for office environments based on multimodal measurements: A review
Marsland et al. Stimulated production of proinflammatory cytokines covaries inversely with heart rate variability
Sharma et al. Objective measures, sensors and computational techniques for stress recognition and classification: A survey
JP2006263054A (en) Acquisition method of respiratory disease related analysis data, oxymeter system, its operation program, oxymeter and oxygen supply system
RU2568354C2 (en) Behaviour therapy method and system applicable for insomnia management
US20140257058A1 (en) Automated personal medical diagnostic system, method, and arrangement
US20150366518A1 (en) Apparatuses, Methods, Processes, and Systems Related to Significant Detrimental Changes in Health Parameters and Activating Lifesaving Measures
US8622901B2 (en) Continuous monitoring of stress using accelerometer data
US8388530B2 (en) Personalized monitoring and healthcare information management using physiological basis functions
JP5862226B2 (en) Method for tracking, recording or analyzing body sensations in 3D
US20120289788A1 (en) Triggering User Queries Based on Sensor Inputs
EP2265173B1 (en) Method and system for sleep/wake condition estimation

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 11772904

Country of ref document: EP

Kind code of ref document: A2

WWE Wipo information: entry into national phase

Ref document number: 2011772904

Country of ref document: EP

NENP Non-entry into the national phase in:

Ref country code: DE