System and method for early detection and monitoring of health state

Abstract

A monitoring system (100) comprising a plurality of monitoring devices (10-12) measuring a monitoring signal (Rmi) associated with a physiological or physical characteristic of the user, during a monitoring time period (Tm). A reference signal (Rri) for each monitoring device is associated with a normal physiological or physical characteristic of the user. A processing unit (20) calculates a difference between a time series of monitoring features (Fmi), from each monitoring signal (Rmi), and a time series of reference features (Fri), from each a reference signal (Rri). The processing unit (20) compares the difference with a threshold value (Tv), an exceeding period of time (Te) when the difference is above the threshold value (Tv), and an exceeding frequency (Fe) indicating the number of occurrences the difference is above the threshold value (Tv). The processing unit (20) determines an onset feature (Op) when the exceeding period of time (Te) is above the reference exceeding period of time (Tre) for a subset (STm) of the monitoring time period, and / or when the exceeding frequency (Fe) is above the reference exceeding frequency (Fre).

Description

System and method for early detection and monitoring of health stateTechnical domain

[0001] The present disclosure concerns a monitoring system and a computerized method for the anonymized early detection of an onset of a pathological state of a user.Background

[0002] Strokes are cerebrovascular events that cause temporary or permanent loss of brain functions. It is estimated that there are approximately 800,000 cases of symptomatic strokes per year in the United States alone, and ten or more million asymptomatic, or silent, strokes per year. When a stroke occurs, it is critical that prompt medical attention is received in order to potentially reduce the severity of the stroke and / or to prevent permanent tissue loss. Patients who experience a stroke, however, often do not seek prompt medical attention due to a variety of factors, including not recognizing that a stroke has occurred, not realizing the importance of prompt treatment, misperceiving the symptoms of stroke as temporary or due to other causes, or for still other reason.

[0003] Parkinson's disease, Alzheimer's and other dementias, stroke, heart failure, heart attack and other coronary diseases, epilepsy, sepsis, Asthma attack, etc., could be detected early in order to take preventive action that could save lives.

[0004] With the expansion of artificial intelligence, intelligent monitoring of people's behavioural tracking can detect the occurrence of disorders in gestures or language, allowing to predict a suspicion of the onset of a targeted pathology in the future. This intelligent monitoring must remain discreet and must respect the privacy of individuals. Existingartificial intelligence algorithms can be used to monitor individuals, but in order to be accurate enough, these algorithms must identify the individuals, typically through biometrics. This allows for fine-grained detection of changes in the gestures or language of the targeted individual.

[0005] Each individual has their own unique gestures and facial / vocal expressions. Without integrating these personal characteristics into monitoring algorithms for early detection of a targeted pathology, this lack of information will result in a higher number of false positives.

[0006] It is recognized that 95% of people who are considered normal have a slight facial asymmetry, which can be more or less pronounced. Multi-neuronal classifiers in artificial intelligence, through deep learning between two populations, one without stroke and the other with stroke, will be able to predict if a person has a stroke. However, with this approach the detection as having a stroke can be a person without a stroke who has a natural facial asymmetry. Moreover, it can be difficult to have a large enough population, one without pathology and the other with a pathology (especially at the time where the stroke happens). In early detection, facial asymmetry can evolve.Summary

[0007] The present disclosure concerns a monitoring system for the anonymized early detection of an onset of a pathological state of a user, the monitoring system comprising: a plurality of monitoring devices, each being configured to measure at least a monitoring signal during a monitoring time period, each of said at least monitoring signal being associated with a physiological or physical characteristic of the user; at least a reference signal for each of said plurality of monitoring device, each reference signal extending for the monitoring time period andbeing associated with a physiological or physical characteristic of the user considered as normal; at least a processing unit configured to calculate at least a time series of monitoring features from each said at least a monitoring signal for the monitoring time period, and calculate at least a time series of reference features from each said at least a reference signal for the monitoring time period; a database unit configured to store said at least a time series of monitoring features and said at least a time series of reference features, for each of said plurality of monitoring device; wherein said at least a processing unit is further configured to calculate a difference between each of said at least a time series of monitoring features and said at least a time series of reference features, during the monitoring time period, and comparing the calculated difference with a threshold value; wherein said at least a processing unit is further configured to calculate an exceeding period of time wherein the calculated difference is above the threshold value and an exceeding frequency indicating the number of occurrences wherein the calculated difference is above the threshold value; wherein the database unit is further configured to store a reference exceeding period of time and a reference exceeding frequency; and wherein said at least a processing unit is further configured to determine an onset feature corresponding to an onset of pathological state of the user when the exceeding period of time is above the reference exceeding period of time for a predetermined subset of the monitoring time period, and / or when the exceeding frequency is above the reference exceeding frequency.

[0008] The monitoring system disclosed herein comprises a plurality of monitoring devices allowing for an earlier determination of the onset feature. The determination of the onset feature is also more accurate andreliable. Since the monitoring system calculates the onset feature based on the time series of monitoring features and time series of reference features, without the user being identified, the determination of the onset feature is anonymized.

[0009] The present disclosure further concerns a non-transitory computer readable medium storing a program causing a computer to execute a method for determine the onset feature.Brief description

[0010] Exemplar embodiments of the invention are disclosed in the description and illustrated by the drawings in which:Fig. 1 shows schematically a monitoring system for early detection and monitoring of health state of a user, according to an embodiment;Fig. 2 shows the monitoring system comprising a remote unit, according to an embodiment;Figs. 3a and 3b illustrate an example where the monitoring devices comprise an image capture device associated with facial characteristics of the user;Fig. 4 illustrates the movements of both ankle joints of a user walking normally; andFig. 5 reports the same type of measurement of the ones of Fig. 4 but on a post -stroke user.Detailed description

[0011] Fig. 1 shows schematically a monitoring system 100 for early detection and monitoring of health state of a user, according to an embodiment. The monitoring system comprises a plurality of monitoring devices 10-12, each being configured to measure at least a monitoringsignal Rmi during a monitoring time period Tm. Each monitoring signal is associated with a physiological or physical characteristic of the user.

[0012] The monitoring system further comprises at least a reference signal Rn for each of said plurality of monitoring device 10-12. Each reference signal Rn is associated with a physiological or physical characteristic of the user considered as normal.

[0013] The monitoring system further comprises at least a processing unit 20 configured to calculate at least a time series of monitoring features Fmi from each said at least a monitoring signal Rmi for the monitoring time period Tm. Each time series of monitoring features Fmi comprises a plurality of monitoring features F.

[0014] The processing unit 20 further calculates at least a time series of reference features Fn from each said at least a reference signal Rn.

[0015] The monitoring system further comprises a database unit 30 configured to store said at least a time series of monitoring features Fmi and said at least a time series of reference features Fn, for each of said plurality of monitoring device 10.

[0016] The processing unit 20 is further configured to calculate a difference Diff between each of said at least a time series of monitoring features Fmi and said at least a time series of reference features Fn, during the monitoring time period Tm, and to compare the calculated difference Diff with a threshold value Tv.

[0017] The processing unit 20 is further configured to calculate an exceeding period of time Te corresponding to a time period where the calculated difference is above the threshold value Tv. The processing unit 20 is further configured to calculate an exceeding frequency Fecorresponding to the number of occurrences where the calculated difference is above the threshold value Tv.

[0018] In an aspect, the difference Diff is calculated by using mathematical equations such as linear equation (see first example below) or by using Al algorithms. The Al algorithms can comprise multi-neuronal classifiers using deep learning between two populations, one without pathology and the other with pathology. The Al algorithms can be complemented by labels provided by specialized personnels capable of diagnosing the pathological cases and non-pathological cases from the time series of monitoring feature Fmi and the time series of reference feature Fn.

[0019] In an embodiment, a reference exceeding period of time Tre and a reference exceeding frequency Fre can be stored in the database unit 30.

[0020] In one aspect, the processing unit 20 is further configured to determine an onset feature Op corresponding to an onset of pathological state of the user when the exceeding period of time Te is above the reference exceeding period of time Tre for a predetermined subset (STm) of the monitoring time period Tm.

[0021] Alternatively, or in combination, the processing unit 20 is further configured to determine the feature Op when the exceeding frequency Fe is above the reference exceeding frequency Fre.

[0022] In an embodiment, one of the monitoring devices comprises at least an image capture device 10 with the monitoring signal corresponding to a plurality of images of the user. The image capture device 110 can comprise a camera having a field-of-view (FOV). The image capture device 110 can be configured to perform one or a combination of panning, tilting, optically zooming, and digitally zooming making possible altering the apparent FOV of the image capture device 110. The image capture device 110 can be panned to direct the FOV towards the user. The image capturedevice 110 can comprise a pan-tilt-zoom (PTZ) camera, preferably provided with an optical zoom.

[0023] In one aspect, the plurality of monitoring devices 10-12 comprises a network of image capture devices distributed over a defined geographic area. The measurement of the monitoring signal can then be performed by one of the image capture devices 10-12 that has detected the presence of the user within the FOV of the image capture device 110, or by more than one image capture devices simultaneously. Here, the monitoring signal can correspond to images comprising facial characteristics of the user or specific movements of the user measured by the image capture device 110. The time series of the monitoring feature Fmi and the time series of reference feature Fn can comprise an asymmetry in the facial characteristics.

[0024] Alternatively, the monitoring signal can correspond to images comprising characteristics of the user specific movements of the user.

[0025] In another aspect, the plurality of monitoring devices 10-12 comprises at least a heart rate monitoring sensor with the monitoring signal corresponding to a heart rate signal.

[0026] In another aspect, the plurality of monitoring devices 10-12 comprises at least an audio capture device with the monitoring signal corresponding to an audio signal. The audio signal can correspond to speech data.

[0027] The threshold value Tv can be a predetermined value or can be calculated from the time series of monitoring features Fmi and the time series of reference features Fn.

[0028] If the calculated difference is below the threshold value Tv during the whole monitoring time period Tm, the threshold value Tv can belowered to get a more precise and efficient detection of the onset feature Op.

[0029] Calculating the difference between the time series of monitoring features Fmi and the time series of reference features Fn allows for a quantitative measurement of the onset feature Op. For instance, in the first example and based on the time series of monitoring features Fmi comprising 104 images results in an accuracy higher than 92%.

[0030] Calculating the difference between the time series of monitoring features Fmi and the time series of reference features Fn further allows for a kinetic measure, in other words, assessing the variation (evolution) of the difference between the time series of monitoring features Fmi and the time series of reference features Fn in time (such as during the monitoring time period Tm). The processing unit 20 can be further configured to determine the onset feature Op of the user based on the variation of the difference Diff.

[0031] In one aspect, the predetermined subset STm of the monitoring time period Tm corresponds to above 10% of the monitoring time period Tm. The predetermined subset STm of the monitoring time period Tm can further correspond to above 20%, 25% or 50% of the monitoring time period Tm.

[0032] In one aspect, the reference exceeding frequency Fre is equal to above 10% of the number of monitoring features F in the time series of monitoring features (Fmi). The reference exceeding frequency Fre can be equal to above 20%, 25% or 50% of the number of monitoring features F in the time series of monitoring features (Fmi).

[0033] In an embodiment, the monitoring signals of the plurality of monitoring devices 10-12 are measured sequentially.

[0034] In an embodiment, a subset of the monitoring signals of the plurality of monitoring devices 10-12 are measured sequentially and another subset of the plurality of monitoring devices 10-12 are measured simultaneously.

[0035] In an embodiment, the monitoring signals of the plurality of monitoring devices 10-12 are measured simultaneously.

[0036] In one aspect, the processing unit 20 can be configured to determine the onset feature Op of the user when the exceeding period of time Te is above the reference exceeding period of time Tre for all of the monitoring devices 10-12 or a subset of the monitoring devices 10-12.

[0037] In one aspect, the processing unit 20 can be configured to determine the onset feature Op of the user when the exceeding frequency Fe is above the reference exceeding frequency Fre for all the monitoring devices 10-12 or a subset of the monitoring devices 10-12.

[0038] In another aspect, the processing unit 20 can be configured to determine the onset feature Op of the user based on the value of the exceeding period of time Te and / or exceeding frequency Fe averaged for all the time series of monitoring features Fmi calculated for the monitoring signals of each of the plurality of monitoring devices 10-12.

[0039] Determining the onset feature Op of the user when the exceeding frequency Fe is above the reference exceeding frequency Fre and / or based on the value of the exceeding period of time Te and / or exceeding frequency Fe averaged for all the time series of monitoring features Fmi calculated for the monitoring signals using the plurality of monitoring devices 10-12 (or a subset of the monitoring devices 10-12) allows for an earlier determination of the onset feature.

[0040] In another aspect, the processing unit 20 can be configured to use the monitoring signal of at least one of the plurality of monitoring devices 10-12 as a trigger to start or stop the measurement with the plurality of monitoring devices 10-12 and / or to start or stop determining the onset featureOp.

[0041] In yet another aspect, the time series of reference features Fn are validated by an expert and / or by using an Al algorithm.

[0042] In some aspects, the monitoring system 100 can be further configured to trigger the use of a biometric recognition, if it has determined the onset featureOp.

[0043] For instance, the monitoring system 100 can comprise a recording a biometric authentication device 120 configured to perform an authentication of the user. The biometric authentication device 120 can comprise an optical biometric authentication unit recording an image of the user. For example, the optical biometric authentication unit can comprise the image capture device 110 with facial recognition and / or palm recognition capabilities. Alternatively, the biometric authentication device 120 can comprise a capture device with automatic voice recognition, an iris recognition device, a fingerprint recognition device, etc.

[0044] As shown in Fig. 2, the monitoring system 100 can be configured to output the exceeding period of time Te, the exceeding frequency Fe and the onset featureOp and transmit theses values Te, Fe, Op to a remote unit 200. The remote unit can comprise an external database 200 where the exceeding period of time Te, the exceeding frequency Fe and the onset featureOp are stored.

[0045] In some aspects, the remote unit 200 can be accessed by specialized personnels in order to evaluate the pathological state of the user and possibly a degree of emergency (such as absolute medicalemergency) based on the exceeding period of time Te, the exceeding frequency Fe and the onset featureOp stored in the remote unit 200.

[0046] The remote unit 200 can comprise an interface unit 210. Based on the evaluation of the pathological state of the user and degree of emergency, the specialized personnels can use the interface unit 210 to transmit a command signal Au to the monitoring system 100, instructing the latter to activate the biometric authentication device 120.

[0047] Once the user has been identified by using the biometric authentication device 120, the monitoring system 100 can be configured to retrieve a file 220 containing user's information. For example, the file 220 can comprise the user's medical history. The file 220 can be stored in the remote unit 200 such as to be accessible to the specialized personnels.First Example

[0048] Figs. 3a and 3b illustrate an example where the monitoring devices comprise at least one image capture device 10 and the time series of monitoring features Fmi with the monitoring signal corresponding to a plurality of images of the user. Fig. 3a shows the geometry of the mouth in a normal state, i.e., non-pathological sate, and Fig. 3b shows the geometry of the mouth in a pathological sate. The plurality of images is associated with facial characteristics of the user. The time series of the monitoring feature Fmi comprises an asymmetry in the facial characteristics.

[0049] Asymmetry in the facial characteristics can comprise a plurality of landmark points LP (coordinates) corresponding to characteristics of the user's face. The time series of monitoring features Fmi can be calculated from the landmark points LP. The time series of monitoring features Fmi can correspond to facial asymmetry of at least a portion of the user's face.

[0050] In this example, the time series of reference features Fn can correspond to the geometry of the mouth in a normal state, i.e., non- pathological sate (fig. 3a). The difference between the time series of monitoring features Fmi and the time series of reference features Fn can be calculated from the angular deviation between the time series of monitoring features Fmi and the time series of reference features Fn. For example, the difference Diff can correspond facial asymmetry expressed in angular differences and / or distance ratios.

[0051] In one aspect, a linear equation can be used to describe the time series of monitoring features Fmi, such as equation 1 :Wi*Fi + W2*F2+ W3*F3+ ... WN-I*FN-I + WN*FN (1) where where Wi is a weighting factor multiplying the feature Fi, in the case of i=N features. The monitoring features Fi to FN may take different values. For example, if the monitoring feature Fi corresponds to the ratio between the ends of the mouth and the pupils, then Fi can be equal to 0.05. If the monitoring feature F2corresponds to the ratio of the distance between the middle of the mouth and lip ends, then F2can be equal to 0.23, etc. In this example, the monitoring feature F2is about four times greater than the monitoring feature Fi and will dominate in equation 1. Equation 1 would then be unbalance yielding a misleading result. The weighting factor Wi multiplying the monitoring feature Fi thus allows the features to have about the same weight in the linear equation. For example, if Fi is equal to 0.05 and F2is equal to 0.23, Wi can take the value of 4.6 and W2can take the value of 1. As mentioned above, based on the time series of monitoring features Fmi comprising 104 images results, an accuracy higher than 92% could be obtained.Second Example

[0052] In another example, the monitoring devices comprise at least one image capture device 10 with the monitoring signal corresponding to a plurality of images of the user. The plurality of images is associated withmovement characteristics of the user. The time series of the monitoring feature Fmi comprises an asymmetry in the movement characteristics.

[0053] Fig. 4 illustrates the movements of both ankle joints of a user walking normally, as captured using the Mediapipe Holistic library and an image capture device 10 (a camera). Fig. 4 shows two curves, one corresponding to the kinetic readings of the right ankle and the other of the left ankle. Fig. 4 clearly demonstrates the synchronized movement between the left ankle and the right ankle. Fig. 5 reports the same type of measurement of the ones of Fig. 4 but on a post-stroke user. The curves in Fig. 5 clearly show a marked asymmetry between the left and the right ankle joints of the user's legs.

[0054] The time series of monitoring features Fmi corresponds to the asymmetry that is quantified by calculating the standard deviation of the intersection of the two curves representing the movement kinetics on the left and right legs. The time series of reference features Fn corresponds to the asymmetry corresponding to a normal walking. The standard deviation has a constant and low level in normal walking, while it is significantly higher in unbalanced walking in case of pathology.

[0055] The calculated difference between the series of monitoring features Fmi and the time series of reference features Fn provides a means of quantifying gait quality during the monitoring time period. The calculated difference can further allow for measuring the kinetics of unbalanced walking in case of pathology via the evolution of the calculated difference during the monitoring time period. This can be used, for example, to intervene before a user falls. This monitoring capability has the potential to improve patient outcomes by enabling timely interventions.Third Example

[0056] In another example (not shown), the monitoring devices comprise at least one motion sensor with the monitoring signal corresponding to a motion signal, where the motion signal is associated with movement characteristics of the user. For example, the movement characteristics of the user comprise the gait of the user. Here, the time series of the monitoring feature and the time series of reference feature comprises comprise an asymmetry in the gait.

[0057] The motion sensor can comprise a three-dimensional accelerometer such as a MEMS-based accelerometer adapted to deliver an acceleration (or motion) signal along three axes.Fourth Example

[0058] In another example where the monitoring devices comprise at least one image capture device 10 and the time series of monitoring features Fmi with the monitoring signal corresponding to a plurality of images of the user. The plurality of images is associated with facial characteristics of the user. The time series of the monitoring feature Fmi comprises an asymmetry in the facial characteristics.

[0059] As in the first example, symmetry in the facial characteristics comprises a plurality of landmark points LP (coordinates) corresponding to characteristics of the user's face. The time series of monitoring features Fmi can be calculated from the landmark points LP. The time series of monitoring features Fmi can correspond to facial asymmetry of at least a portion of the user's face.

[0060] Here, the processing unit 20 can be configured to find landmark points in each of the plurality of images. Finding the landmark points canbe performed by the processing unit 20 using a face detection algorithm, for example by using DLIB or MEDIAPIPE libraries.

[0061] The processing unit 20 can then be configured to recalculate targeted points corresponding to landmark points of the user's face most relevant for the pathology under monitoring. For example, landmark point related to the user's mouth for stroke monitoring.

[0062] The processing unit 20 can then be configured to determining a point in 3D space (triangulation, for example Delaunay triangulation) of the measured images.

[0063] The processing unit 20 can then be configured to calculate the triangulation of the destination image with the recalculated points.

[0064] The processing unit 20 can then be configured to extract and warp triangles and link the warped triangles together.

[0065] The processing unit 20 can then be configured to replace the face on the destination image.

[0066] The processing unit 20 can then be configured to perform seamless cloning.

[0067] Other morphing methodologies can be used. The recalculation of the targeted points corresponding to landmark points of the user's face most relevant for the pathology under monitoring should be included in the morphing methodology.

[0068] Conversely, it is possible to use images of users having a pathology and recalculate the landmark points to obtain a transition with a symmetrical face (an image corresponding to the absence of pathology).

[0069] The present disclosure further concerns a non-transitory computer readable medium storing a program causing a computer to execute a method for determine the onset feature. The method comprises the steps of:Using the plurality of monitoring device 10-12 to measure said at least a monitoring signal Rmi during the monitoring time period Tm, each of said at least monitoring signal being associated with a physiological or physical characteristic of the user; using the processing unit 20 to calculate said at least a time series of monitoring features Fmi from each said at least a monitoring signal Rmi for the monitoring time period Tm, and calculating at least a time series of reference features Fn from each said at least a reference signal Rn for the monitoring time period Tm; storing in the database unit 30 said at least a time series of monitoring features Fmi and said at least a time series of reference features Fn, for each of said plurality of monitoring device; using the processing unit 20 to calculate the exceeding period of time Te wherein the calculated difference is above the threshold value Tv and the exceeding frequency Fe indicating the number of occurrences wherein the calculated difference is above the threshold value Tv; storing in the database unit 30 the reference exceeding period of time Tre and a reference exceeding frequency Fre; and using the processing unit 20 to determine the onset feature Op when the exceeding period of time Te is above the reference exceeding period of time Tre for a predetermined subset STm of the monitoring time period, and / or when the exceeding frequency Fe is above the reference exceeding frequency Fre.Reference numbers and symbols100 monitoring system10-12 monitoring device110 image capture device120 biometric authentication device20 processing unit200 remote unit210 interface unit220 file30 database unitAu command signalDiff differenceF monitoring featureFe exceeding frequencyFrrii time series of monitoring featureFr time series of reference featureFre reference exceeding frequencyLP landmark pointOp onset featureRfi reference signalRm i monitoring signaSTm predetermined subset of the monitoring time periodTe exceeding period of timeTre reference exceeding period of timeTm monitoring time periodTv threshold value

Claims

Claims1. Monitoring system (100) for the anonymized early detection of an onset of a pathological state of a user, the monitoring system comprising: a plurality of monitoring devices (10-12), each being configured to measure at least a monitoring signal (Rm during a monitoring time period (Tm), each of said at least monitoring signal being associated with a physiological or physical characteristic of the user; at least a reference signal (Rn) for each of said plurality of monitoring device (10-12), each reference signal (Rn) being associated with a physiological or physical characteristic of the user considered as normal; at least a processing unit (20) configured to calculate at least a time series of monitoring features (Fmi) from each said at least a monitoring signal (Rm for the monitoring time period (Tm), and calculate at least a time series of reference features (Fn) from each said at least a reference signal (Rn) for the monitoring time period (Tm); a database unit (30) configured to store said at least a time series of monitoring features (Fmi) and said at least a time series of reference features (Fn), for each of said plurality of monitoring device; wherein said at least a processing unit (20) is further configured to calculate a difference between each of said at least a time series of monitoring features (Fmi) and said at least a time series of reference features (Fn), during the monitoring time period (Tm), and comparing the calculated difference with a threshold value (Tv); wherein said at least a processing unit (20) is further configured to calculate an exceeding period of time (Te) wherein the calculated difference is above the threshold value (Tv) and an exceeding frequency (Fe) indicating the number of occurrences wherein the calculated difference is above the threshold value (Tv);wherein the database unit (30) is further configured to store a reference exceeding period of time (Tre) and a reference exceeding frequency (Fre); and wherein said at least a processing unit (20) is further configured to determine an onset feature (Op) corresponding to an onset of pathological state of the user when the exceeding period of time (Te) is above the reference exceeding period of time (Tre) for a predetermined subset (STm) of the monitoring time period, and / or when the exceeding frequency (Fe) is above the reference exceeding frequency (Fre).

2. The monitoring system according to claim 1, wherein said a plurality of monitoring devices (10-12) comprises at least an image capture device (110) with the monitoring signal corresponding to a plurality of images of the user.

3. The monitoring system according to claim 1 or 2, wherein said a plurality of monitoring devices (10-12) comprises at least a heart rate monitoring sensor with the monitoring signal corresponding to a heart rate signal.

4. The monitoring system according to any one of claims 1 to 3, wherein said a plurality of monitoring devices (10-12) comprises at least an audio capture device with the monitoring signal corresponding to an audio signal.

5. The monitoring system according to any one of claims 1 to 4, wherein said a plurality of monitoring devices (10-12) comprises at least a motion sensor with the monitoring signal corresponding to a motion signal.

6. The monitoring system according to claim 2, wherein said plurality of images of the user are associated with movement characteristics of the user.

7. The monitoring system according to claim 5, wherein the motion signal is associated with movement characteristics of the user.

8. The monitoring system according to claim 6 or 7, wherein said movement characteristics of the user comprise the gait of the user; and wherein the time series of the monitoring feature and the time series of reference feature (Fn) comprises comprise an asymmetry in the gait.

9. The monitoring system according to claim 2, wherein said plurality of images is associated with facial characteristics of the user.

10. The monitoring system according to claim 9, wherein the time series of the monitoring feature and the time series of reference feature (Fmi, Fn) comprises an asymmetry in the facial characteristics.

11. The monitoring system according to claim 10, wherein said at least a processing unit (20) is further configured to calculate a variation of the time series of the monitoring feature (Fmi) during the monitoring time period (Tm); and wherein said at least a processing unit is further configured to determine the onset of pathological state of the user based on the calculated variation.

12. The monitoring system according to any one of claims 1 to 11, wherein the monitoring signals of the plurality of monitoring devices (IQ- 12) are measured simultaneously.

13. The monitoring system according to claim 12, wherein the predetermined subset (STm) of the monitoring time period corresponds to above 10% of the monitoring time period (Tm).

14. The monitoring system according to claim 12 or 13, wherein said at least a processing unit is further configured to determine the onset of pathological state of the user based on the value of the exceeding period of time (Te) and / or exceeding frequency (Fe) averaged for all the time series of monitoring features (Fm ) calculated for the monitoring signals of each of the plurality of monitoring devices (10-12).

15. The monitoring system according to any one of claims 12 to 14, wherein said at least a processing unit is further configured to use the monitoring signal of at least one of the plurality of monitoring devices as a trigger to start or stop the measurement with the plurality of monitoring devices and / or to start or stop the determine an onset of pathological state.

16. The monitoring system according to any one of claims 1 to 15, wherein the time series of reference features (Fn) are validated by an expert.

17. The monitoring system according to any one of claims 1 to 16, further comprising a remote unit (200); and wherein the monitoring system (100) is configured to output the exceeding period of time (Te), the exceeding frequency (Fe) and the onset feature (Op) and transmit them to the remote unit (200).

18. A non-transitory computer readable medium storing a program causing a computer to execute a method for determine the onset feature, the method comprises the steps of: using the plurality of monitoring device (10-12) to measure said at least a monitoring signal (Rm during the monitoring time period (Tm),each of said at least monitoring signal being associated with a physiological or physical characteristic of the user; using the processing unit (20) to calculate said at least a time series of monitoring features (Fm from each said at least a monitoring signal (Rmi) for the monitoring time period (Tm), and calculating at least a time series of reference features (Fn) from each said at least a reference signal (RH) for the monitoring time period (Tm); storing in the database unit (30) said at least a time series of monitoring features (Fm and said at least a time series of reference features (Fn), for each of said plurality of monitoring device (10-12); using the processing unit (20) to calculate the exceeding period of time (Te) wherein the calculated difference is above the threshold value (Tv) and the exceeding frequency (Fe) indicating the number of occurrences wherein the calculated difference is above the threshold value (Tv); storing in the database unit (30) the reference exceeding period of time (Tre) and a reference exceeding frequency (Fre); and using the processing unit (20) to determine the onset feature (Op) when the exceeding period of time (Te) is above the reference exceeding period of time (Tre) for a predetermined subset (STm) of the monitoring time period, and / or when the exceeding frequency (Fe) is above the reference exceeding frequency (Fre).

Images