WO2023237758A1

WO2023237758A1 - Method for determining a neurological condition in a subject

Info

Publication number: WO2023237758A1
Application number: PCT/EP2023/065534
Authority: WO
Inventors: Julian VARGHESE
Original assignee: Westfälische Wilhelms-Universität Münster
Priority date: 2022-06-08
Filing date: 2023-06-09
Publication date: 2023-12-14

Abstract

In various embodiments, a method for determining a neurological condition in a subject is provided. The method comprises: receiving motion signals which have been obtained from at least one body part on one side of the subject and from the corresponding body part on the other side of the subject; processing the motion signals to determine: i) a first group of parameters for each side of the subject based on predetermined magnitudes of the motion signals; ii) a second group of parameters for each side of the subject based on an analysis of noise component in the singular spectrum analysis of the motion signals; and iii) a third group of parameters for each side of the subject based on an analysis of periodic components of the motion signals; forming a vector, comprising the first group of parameters, the second group of parameters and the third group of parameters; comparing the formed vector to a plurality of reference vectors, wherein the reference vectors comprise at least one reference vector representing a healthy subject and at least one reference vector representing a subject that has been diagnosed with the neurological condition; and determining a probability for the subject having the neurological condition based on the outcome of the comparison.

Description

Method for determining a neurological condition in a subject

The present invention relates to a method for determining a neurological condition in a subject.

Tremor-related diseases as Parkinson's disease (PD) and essential tremor (ET) are two of the most common movement disorders. Disease classification is primarily based on clinical criteria and remains challenging. Since both neurological conditions affect the way how a subject moves, precise detection of the motion of a subject, such as his hands, for example, and its analysis, in particular in comparison to heathy subjects, has become more and more important in the diagnosis thereof. Smart wearables have become particularly useful for that purpose, as they are readily available and come equipped with sophisticated motion sensors. When used in a clinical setting, their multi-sensor technology provides a source of precise and objective movement monitoring which allows for greater precision in recording subtle changes in motion patterns of the wearer as compared to current clinical rating scales in hospital routine.

So far, a smartphone-based analysis to determine whether a subject has Parkinson's or not is known from prior art. Furthermore, Parkinson's Kinetigraph from Global Kinetics needs to be mentioned, which may be referred to as Parkinson's clock. This system is a clock that only monitors and visualizes movement measurements in patients, but without generating diagnosis probability or a neurological finding with an evaluation of specific Parkinson's characteristics.

The existing diagnostic systems that work with movement or acceleration sensors only allow a distinction between Parkinson's and the state of being healthy (i.e., not having Parkinson's), but not between other similar movement disorders, such as multiple sclerosis (MS). This is because those known systems have only been trained with Parkinson's and healthy individuals, but not with individuals having other movement disorders. As a result, the diagnostically relevant, more comprehensive classification into the three classes - Parkinson's disease, other movement disorders, and healthy - is missing. This is particularly clinically relevant, since there are many diseases similar to Parkinson's and a new or additional movement disorder can be diagnosed in a patient after a diagnosis of Parkinson's.

In addition to the purely simplified diagnostic output (Parkinson's, yes or no), an automatically generated detailed examination report is desirable and currently not provided by the known systems, which explains the diagnostic output and lists neurologically relevant movement characteristics. Such a detailed examination report would provide a precise, objective movement documentation as an examination finding and would facilitate an objective documentation of the course of the disease in a diagnosed patient. Such an objective movement documentation would be of immeasurable value, as it would provide a clear indication for the effectiveness of medication which would supplement or could even replace the rather subjective and possibly too crude assessment of the patient.

In various embodiments, the present invention aims to provide a detection method for determining a neurological condition in a subject, the neurological condition preferably encompassing Parkinson's Disease (PD) and Essential Tremor (ET), which can distinguish between at least two neurological conditions and thus provides an improved diagnosis.

The method of the present invention is based on a synchronous and parallel measurement of a motion pattern of a subject with two motion sensors, such as two smart watches, each of them being attached to a subject's wrist. Such a measurement allows for a better spatial resolution with direct side comparison. Based on a two-sided measurement using two motion sensors, during which the subject follows instructions and undergoes a series of movement patterns, high-resolution characteristics of the slowing down (deceleration) of movement and the trembling behavior, including side comparison, are determined using mathematical methods. Based thereon, a clinical- neurological finding for the diagnosis and for an objective analysis of the course of movement restrictions is provided. As noted above, smart watches can be preferably used for the purpose of the method according to the invention. The smart watches are connected to a master electronic device, such as a smartphone or a tablet, which processes the data from the smartwatches and provides the desired output in the preferable form of a documentation of diagnostic findings and diagnostic classification into three diagnostic areas - Parkinson's, other neurological movement disorder, or healthy.

The method for determining a neurological condition in a subject according to various embodiments is based on a two-sided measurement, preferably on both wrists of a subject, and allows for an automated generation of a clinical-neurological finding/report. In an embodiment of the method, the provided finding lists, based on a mathematical evaluation, refer to Parkinson-specific movement abnormalities: akinesis, rigor and tremor. For each movement abnormality, a quantitative and illustrative statement is preferably made by displaying the respective physical unit (akinesis, rigor and tremor) and a corresponding percentage that shows the severity of the movement abnormality. Based on these abnormalities, a classification into Parkinson's, other movement disorder and no movement disorder is also generated as support for a diagnostic decision.

Due to the two-sided measurement and the nature of the applied mathematical evaluation of the method of the present invention, highly specific Parkinson's characteristics may be automatically recorded and visualized for the patient and/or the clinician. Therefore, the method of the present invention can provide a diagnostic decision support with at least 80% accuracy. There are no other smart watch-based systems for detection of neurological conditions with such a high accuracy.

In addition, the method of the present invention is helpful in solving another problem. Current diagnosis of Parkinson's in the context of a manual neurological examination is time-consuming, regularly necessary and requires the expertise of a movement disorder specialist. The method of the present invention can enable generation of an automated examination report by those affected themselves and from home. This report may support an early detection of Parkinson's or another neurological condition and enables objective monitoring of the course of the tremor behavior and the degree of slowing of movement for the patient and/or clinician.

In various embodiments a method for determining a neurological condition in a subject is provided. The method is based on motion data gathered from a two-sided measurement using two motion sensors, such as smart watches, with each one being attached to the subject's wrist. Each of the motion sensors may comprise a three-axis accelerometer which measures acceleration along each one of the three spatial axes in space. The method of the present invention may be a computer- implemented method in which motions signals received from sensors are evaluated by a corresponding algorithm which, in the end, outputs at least a probability for the subject having a neurological condition.

In a first step, the method of the present invention comprises receiving motion signals which have been obtained from at least one body part on one side of the subject and from the corresponding body part on the other side of the subject a subject. The received motion signals may comprise a plurality of signals, each motion signal being indicative of recorded acceleration of the left and right wrist of the subject, for example, in a respective direction, such an x- ,y- or z-axis of a rectangular coordinate system. Additionally, an orientation signal may be used in the analysis, provided by a gyroscope which provides orientation information of the device (e.g. smart watch) based on Earth's gravity. In the context of this application, if not explicitly stated otherwise, a motion signal may refer to the actual signal from a motion sensor and to the signal from the gyroscope which allows to determine movement (rotation and/or translation) of the corresponding measuring device (e.g. a smart watch) in space from which the movement of the corresponding body part of the subject (e.g. a wrist) may be derived.

In a next step, the method of the present invention comprises processing the motion signals to determine groups of parameters for each side of the subject. For the following processing steps, a sampling step may be applied to the motion signals in order to convert the (quasi) continuous-time motion signals into discrete or quantized time signals. The sampling frequency may be, for example, 100 Hz or more. Each of the group of parameters may comprise one or more parameters, such as 5, 10, 20, 50 or 1000 parameters. The parameters in each group of parameters may belong to the same class of parameters. The parameters in one group of parameters may be determined in the same mathematical manner, e.g. by means of the same mathematical steps or correspond to values or coefficients obtained from the same mathematical function.

The first group of parameters is determined (calculated) based on predefined acceleration magnitudes of the motion signals for each side of the subject. A predefined acceleration magnitude may correspond to the acceleration in g at a predetermined percentile, e.g. the 90th percentile, within a specific neurological assessment step (when the patient is instructed to straighten and raise and lower his arms, for example). In other words, this step includes determining at least one parameter, such as the 90th percentile, which is representative of an upper spectrum of the acceleration magnitudes in each of the motion signals. More than one parameter may be determined at this step, e.g. the 95th percentile, the 90th percentile and the 75th percentile. The extraction of the predefined acceleration magnitude(s) may be performed globally for every motion signal or locally, i.e. in predefined segments of the motion signals, each segment being related to the same specific neurological assessment step. The motion signals may all be synchronized.

The second group of parameters is determined based on an analysis of noise component in a singular spectrum analysis (SSA) of the motion signals for each side of the subject. By means of SSA, any signal comprising a time series, such as the quantized motion signal, can be decomposed and then reconstructed into an oscillatory time series component, a trend component, and a noise component. In this method step, the second group of parameters is determined from the noise component of the SSA in each of the motion signals. For the purpose of the method disclosed herein, any other suitable method by which a time series may be decomposed into the three subseries components trend, seasonal and noise can be used, SSA being one example thereof. The second group of parameters may comprise, for example, a parameter which quantifies the energy of the noise component throughout the entire motion signal or in predefined segments of the motion signal (e.g. segments correlated with neurologically relevant motion sequences of the patient) or a plurality of parameters, each one corresponding to the magnitude of noise at a predetermined time point. In a further embodiment, the second group of parameters may comprise all discrete values of a quantized noise component for a given motion signal. The third group of parameters is determined based on an analysis of periodic components of the motion signals for each body side of the subject. In this method step, the periodic components may be obtained from a Fourier transformation (e.g. FFT) of the motion signals. The third group of components may comprise, for example, a plurality of coefficients of the Fourier transform or other values which are calculated based on a plurality of those coefficients.

In the above method steps, in which the motion signals are analyzed to determine the first, the second, and the third group of parameters, predetermined segments in each of the motion signals can be considered. That is, the motion signals may be divided into neurologically relevant segments, each one representing a neurological region of interest. Each neurologically relevant segment may correspond to a segment in a respective motion signal. Each segment may correspond to a phase during which the subject was performing a predetermined movement pattern, for example based on instructions received from the master electronic device, such as a tablet or a smartphone. In that manner, the evaluation of the motion signals may be applied to neurologically neurologically relevant segments of the motion signals, thus improving accuracy of the results provided by the method according to the invention.

In a next step, the method of the present invention comprises forming a vector, comprising the first group of parameters, the second group of parameters and the third group of parameters, which quantify characteristics of the movement pattern of the subject, in particular allowing a comparison between both sides of the patient. In the simplest exemplary case, in which two smart watches are used, each of them being affixed to one wrist of the subject, the motion vector may be a six dimensional vector, with the first three coefficients representing the first, second, and third group of parameters for the left wrist and the last three coefficients representing the first, second, and third group of parameters for the right wrist. In this very simplified example, each group of parameters comprises one parameter. In the more general case of the method according to various embodiments of the invention, further motion signals may be obtained from motion sensors attached to other locations on the body of the subject, such as the elbows and/or middle portion on the upper arms. In that case, the groups of parameters (i.e. the first, second, and third group of parameters) will be determined for each of the n designated body parts, for both of the two sides of the body separately, and the resulting vector may comprise n‘2‘3 = 6‘n groups of parameters. As noted earlier, each of the 6‘n groups of parameters may include one or more parameters, wherein each pair of groups representing the same body part on both sides of the subject's body will have the same number of parameters. Overall, in practice the vector will be highly dimensional. In any case, the vector allows collective processing of the obtained motion signals in a condensed form, i.e. represented by the first, second, and third parameters.

In a next step, the method of the present invention comprises comparing the formed vector to a plurality of reference vectors, wherein the reference vectors comprise at least one reference vector representing a healthy subject and at least one reference vector representing a subject that has been diagnosed with the neurological condition. The plurality of reference vectors correspond to a reference group to which the testes subject's vector is compared to. The comparison may, for example, include calculation of a distance between the subject's vector and each one of the reference vectors based on a predefined metric. By evaluating the distances, the position of the subject's vector in the multidimensional vector space may be determined, relative to the reference vectors. Additionally, the comparison may include comparing portions of the subject's vector to corresponding portions of the reference vectors. For example, the comparison may be made based on partial comparisons, i.e. by comparisons made in subspaces of the multidimensional vector space, which are then ultimately evaluated collectively, e.g. by a weighted addition, to obtain a final result of the overall comparison. The subspaces may be chosen in any suitable way. For example, the subspaces may be chosen in accordance with portions of the vector defined by a respective group of parameters for a respective body part or by the first groups of parameters of all the considered body parts.

The method may include different measures of similarity being applied different subspaces to extract quantitative statements about the similarity of the formed vector to a vector representing a healthy subject and a subject affected by a neurological condition other than Parkinson's. This may be particularly the case when the vector comprises a large number of coefficients, wherein the first parameter is represented by a first group of coefficients, the second parameter is represented by a second group of coefficients and the third parameter is represented by a third group of coefficients, each group of coefficients being determined in a respective one of the above noted steps. For example, the measure of similarity between the portion of the vector containing coefficients which have been determined based on the evaluation of acceleration magnitudes and the portion of the vector containing coefficients which have been determined based on periodic components and between corresponding portions of "healthy vectors" and vectors affected by other neurological conditions may be determined by calculating their distance, i.e. by applying a similarity metric to the data, such as the Euclidean distance or the Mahalanobis distance. The measure of similarity between the portion of the vector containing coefficients which have been determined based on an analysis of the noise component and between corresponding portions of "healthy vectors" and vectors affected by other neurological conditions may be determined by dynamic time warping or a BOSS transformation. Based on the calculated partial similarity measures, a total similarity measure may be determined.

The classification of the vector, i.e. the determination whether the vector represents a healthy subject or a subject affected by a neurological disorder other than Parkinson's, together with the respective probabilities that that is the case, may be made based on the total similarity, as determined. For example, the k-nearest neighbors algorithm may be used for that purpose.

In further embodiments, comparing the formed vector to the plurality of reference vectors may be performed by means of an artificial neural network which has been trained with data obtained from reference subjects.

In a further, final step the method of the present invention comprises determining a probability for the subject having the neurological condition based on the outcome of the comparison. The probability is calculated collectively from the result(s) obtained in the previous step, in which the subject's vector has been compared to a plurality of reference vectors. In further embodiments of the method, as noted before, the probability may be determined by the artificial neural network.

According to further embodiments, the method may further include processing the groups of parameters into an automated examination report. The automated examination report which is prepared based on the evaluation of the groups of parameters is a time saving element when it comes to assessing and documenting the subject's neurological state.

According to further embodiments of the method, the first group of parameters may be indicative of akinesia in the subject. In other words, akinesia may be quantified based on the first group of parameters. Akinesia describes the loss of ability of intentional muscle activation in order to move. A predominant sign of akinesia is freezing of a body part, which renders it immovable despite the will to move it. The amount of akinesia in a subject's motion or the severity of akinesia by which the subject's motion is affected may be evaluated based on the predetermined magnitudes of the motion signals. The smaller the acceleration value(s) at the predetermined percentile(s), e.g. at the 90th percentile, the more pronounced akinesia can be expected to be in the subject.

According to further embodiments of the method, the second group of parameters may be indicative of rigor in the subject. In other words, rigor may be quantified based on the third group of parameters. Rigor is a further movement abnormality normally occurring in subjects affected by PD. Its main cause is increased muscle tension. A muscle that is flexed by an external force resists movement. This resistance is canceled abruptly and the intended movement can be partially carried out. Then resistance sets in again and after a while ends just as abruptly as before. The result is a jerky motion - it gives the impression as if the patient was moved by a cogwheel, hence the name cogwheel rigidity. The amount of rigor in a subject's motion or the severity of rigor by which the subject's motion is affected may be evaluated based on the noise in each of the motion signals which reflects the erratic and, in particular, non-periodic nature of the rigor.

According to further embodiments of the method, the third group of parameters may be indicative of tremor in the patient. In other words, tremor may be quantified based on the third group of parameters. Tremor, defined as involuntary, rhythmic movement of one or more body parts, is a further characteristic symptom of PD. Unlike ET, it is usually seen at rest. The severity of tremor by which the subject's motion is affected may be evaluated based on the periodic components in the motion signals which reflect the rhythmic nature of the oscillations in the movement of the subject caused by PD. Additionally, the symmetry in occurrence of oscillatory components may be used to distinguish PD from ET which, in contrast to PD, affects both sides of the body rather symmetrically, whereas tremor induced by PD tends to affect mostly one of the sides of the body.

According to further embodiments the method may comprise providing a presence indicator for each one of akinesia, rigor and tremor as neurological symptoms, which indicates the severity of the respective neurological symptom in the patient. The presence indicator for each one of the three symptoms may be derived from the comparison of the subject's vector with the reference vectors, in particular its components representing the respective symptom.

According to further embodiments of the method, providing the probability for the subject having the neurological condition may be further based on a comparison between the two first groups of parameters, the two second groups of parameters, and/or the two third groups of parameters as obtained for a respective body part on the two different sides of the subject's body. In doing so, the symmetry of the motion abnormality may be determined in each of the fields representing rigor, akinesia and tremor. The symmetry (or asymmetry) between corresponding groups of parameters for different sides of the subject's body (e.g. the first group of parameters for the left body side and the first group of parameters for the right body side) may be quantified and comprised as further parameters in the vector which is formed based on the first, second and third groups of parameters.

In general, the vector comprising the groups of parameters for each sensor/device attached to the subject and for the left body side and the right body side may comprise further parameters which have been derived from these groups of parameters. Therefore, the formed vector may have a dimension which is significantly larger than the previously mentioned 6 dimensions for a given body part, e.g. the wrist. The inclusion of further coefficients increases the dimensionality of the comparison and may render the method more reliable and accurate.

According to further embodiments of the method, the motion signals may have been obtained during guided movement of the subject based on movement instructions shown to the subject. Consequently, the according to various embodiments the method may include the further step of providing instructions for guided movement to the subject in order to record motion signals. The instructions can be provided to the subject my means of one of the smart watches already attached to a wrist or by means of the master device, such as a smart phone or a tablet. The instructions may cause the subject to rest in between different movement patterns in order to obtain motion signals which may be easily segmented into sections of (neurological) interest, such as segments where the subject is performing a movement in accordance with the instructions and sections in which the subject is deliberately at reset in immediately before and/or after the actively guided movement, for example.

According to further embodiments the method may comprise determining at least one neurologically relevant segment within the received motion signals and processing only the relevant segment(s) of the motion signals to determine the first group of parameters, the second group of parameters and the third group of parameters. The segmentation of the motion signals may be performed automatically by the corresponding algorithm, e.g. based on segments of the motion signals in which the recorded accelerations do not surpass a threshold, thus indicating breaks in between the motion patterns performed by the subject.

According to further embodiments of the method the step of comparing the formed vector to a plurality of reference vectors may include, for the first group of parameters, applying a symbolic Fourier transformation to the motion signals, dividing the frequency domain into frequency bins and mapping the resulting Fourier coefficients to respective bins, and comparing the obtained histogram with histograms obtained in the same manner for the first parameter of the reference vectors. The result of the binning may be represented by parameters which may be comprised by the first group of parameters as parameters derived from the amplitudes of acceleration in the motion signals.

According to further embodiments of the method, comparing the obtained histograms may include application of a BOSS transformation to the motion signals.

According to further embodiments of the method, the analysis of periodic components of the motion signals may include an analysis of the spectral composition of the motion signals. The spectral composition of the motion signals may be obtained from a Fourier transform of the obtained motion signals.

According to further embodiments of the method, the neurological condition may include Parkinson's disease.

The foregoing and other features of the present invention will become more fully apparent from the following description and appended claims, taken in conjunction with the accompanying drawings. It is understood that the accompanying drawings depict only an exemplary embodiment in accordance with the present invention. Aspects of the invention will be described with additional specificity and detail through use of the accompanying drawings, such that the advantages of the present invention can be more readily appreciated. Fig. 1 shows a flow chart which illustrates an embodiment of the method according to the invention.

Fig. 2 shows an exemplary examination finding which may be provided by the method according to the invention.

Fig.l shows a flowchart 1 which illustrates an exemplary process flow of the method according to various embodiments of the invention. In an initial step 10 of the method motion signals are received, which have been obtained from at least one body part on one side of the subject and from the corresponding body part on the other side of the subject. The motion signals may have been recorded beforehand, stored in a database, and may be provided to the algorithm or program operating in accordance with the method according to the invention. Alternatively, the method according to various embodiments may comprise a further step, which precedes the initial step 10 and in which the motion signals from at least one body site on both sides of the patient are recorded. The motion signals may be recorded during a neurological test, which may be a self-test or a test which is conducted by medical personnel.

An exemplary test protocol of the neurological test may be as follows. The patient is in a sitting position, with palms on placed on the thighs. This position is the basis position in which the patient is at rest. The sensors acquire motion signals of the patient at rest, e.g. in the basis position, and in a number of consecutive movement patterns which may be shown to the patient on an electronic device. The movement patterns may include, for example, raising and lowering arms on both sides, raising both hands and pretending to screw in and screw out lightbulbs, directing a finger of the hand to the nose (simultaneously and alternately on both sides). After going through each motion pattern, the subject may be instructed to hold still for a predetermined period of time to generate a baseline in the motion data which indicates a non-moving subject. The acquired data which is received in the initial step 10 may comprise motion signals form the motion sensors and orientation signals from gyroscopes. As noted before, in order to conduct the measurement phase, smart watches attached to the subject's wrists may be used, which comprise the required sensors.

Once the motion signals are readily available, they are pre-processed in a second step 11, which is an optional step, in order to segment the motion signals. The segmentation aims at flagging neurologically relevant segments of the motion signals which correspond to segments of the motion signals which represent the subject's movement in accordance with the provided instructions in the context of a neurological test, for example during phases in which the subject's lifts his/her arm or holds it still in a lifted position. This can be done by detecting baseline sections in the motion signals, in which the subject was at rest intentionally (due to corresponding instructions). Alternatively, the neurologically relevant segments can be flagged by synchronizing the motion signals with the instructions which have been provided to the subject on the electronic device.

In a next third step 12, which encompasses the three steps 13-15, the pre-processed motion signals are further processed in order to determine the first, second, and third groups of parameters. Each of those groups of parameters is determined in a process flow represented by one of the steps 13-15, which may be executed in series, in parallel or in any other suitable sequence. Individual calculation steps or intermediate results within the process flows may be shared between the overall process flows within each of the steps 13-15, if expedient.

In the fourth step 13 the first group of parameters for each side of the subject may be determined based on predetermined magnitudes of the motion signals, as described earlier. In addition, a singular spectrum analysis (SSA) of the motion signals may be performed to determine the components representing oscillation, a trend, and noise.

In the fifth step 14 the second group of parameters for each side of the subject may be determined based on an analysis of the noise component determined by an SSA of the motion signals. Since an SSA of the motion signals has already been performed in the fourth step 13, the results obtained in that step can be used in the fifth step 14 (or vice versa, depending on the order in which the results are obtained) in order to determine the noise component. The obtained noise component is evaluated and provides the basis for the determination of the second group of parameters.

In a sixth step 15 the third group of parameters for each side of the subject is determined based on an analysis of periodic components in the motion signals. The periodic components may be determined by applying a fast Fourier transform to the motion signals. Here, the results may be shared with the process flow of the third step 12 (or vice versa), since the SSA also requires calculation of a discrete Fourier transform.

Furthermore, in order to present the information extracted from the time series in a more compact fashion, the motion signals may be approximated using symbolic Fourier approximation (SFT). This process includes the steps for applying discrete Fourier transformation (DFT) to a signal and subsequently quantizing it using multiple coefficient binning (MCB). The SFT approximated motion signals may be then evaluated using a BOSS (Bag-of-SFA-Symbols) model. This model is based on word extraction (i.e. combinations of previously used coefficients which were used for binning in the SFT approximation), and generating features in a graph which represent frequencies of the extracted words. The results obtained from these process steps may be allocated to the third group of parameters and also stored in the vector. Therefore, these results may correspond to data which is used in the comparison of the formed vector to healthy vectors and to vectors affected by other neurological conditions than Parkinson's.

Once the groups of parameters have been calculated for the at least one body site on the left and right side on the body of the subject, a vector is formed in a seventh step 16, which comprises the first group of parameters, the second group of parameters and the third group of parameters. The vector is formed based on parameters which have been specifically obtained from motion signals of the subject and is therefore characteristic of the examined subject. The formed vector may include further parameters or coefficients which correspond to further characteristic values extracted from the motion signals.

In a next eighth step 17, the formed vector is compared to a plurality of reference vectors, wherein the reference vectors comprise at least one reference vector representing a healthy subject and at least one reference vector representing a subject that has been diagnosed with the neurological condition. Exemplary methods for performing the comparison have been already described above in the general part of this description. In particular, an artificial neural network which has been trained with data obtained from reference subjects may be used for the comparison.

In a last ninth step 18 of the exemplary method, a probability for the subject having the neurological condition is determined based on the outcome of the comparison. In this step, the result obtained in the previous eighth step 17 is attributed a probability which may be displayed as a result of the method for determining a neurological condition in a subject as a whole. At the same time, the result of the comparison performed in the eighth step 17 may be evaluated in the context of a probability that the subject is suffering from another movement disorder, such as ET, and/or is healthy.

In Fig. 2, a schematic of an exemplary examination finding 2 is depicted which may be provided by the method according to various embodiments, an exemplary embodiment of which is illustrated in Fig. 1. The examination finding 2 comprises a first field 21, in which the amount or severity of akinesia as the first parameter is indicated. In a second field 22, the amount or severity of rigor as the second parameter is indicated. Finally, in a third field 23, the amount or severity of tremor as the third parameter is indicated. The three fields 21-23 display the presence indicators which have been mentioned above in the general part of the description.

In a further field 24, the probability of the subject having Parkinson's is indicated. The presence indicators displayed in the first three fields 21-23 quantify characteristics which have been calculated from the motion signals alone and correspond to absolute values. The presence indicators may be also helpful for a clinician to formulate a diagnosis based on his own clinical experience or to monitor disease progress, purely on the merit of the severity of akinesia, rigor and tremor which have been determined based on the motion data from the subject. The probability of the subject having Parkinson's, as well as the probability of the subject having a movement disorder other than Parkinson's (displayed in a second further field 25) and the probability of the subject being healthy (displayed in a third further field 26), on the contrary, are obtained by collective evaluation of the groups of parameters by means of the comparison with reference subjects (see eighth step 17 in flowchart 1 in Fig. 1).

Claims

1. Method for determining a neurological condition in a subject, comprising: receiving motion signals which have been obtained from at least one body part on one side of the subject and from the corresponding body part on the other side of the subject; processing the motion signals to determine: i) a first group of parameters for each side of the subject based on predetermined magnitudes of the motion signals; ii) a second group of parameters for each side of the subject based on an analysis of noise component in the singular spectrum analysis of the motion signals; and iii) a third group of parameters for each side of the subject based on an analysis of periodic components of the motion signals; forming a vector, comprising the first group of parameters, the second group of parameters and the third group of parameters; comparing the formed vector to a plurality of reference vectors, wherein the reference vectors comprise at least one reference vector representing a healthy subject and at least one reference vector representing a subject that has been diagnosed with the neurological condition; determining a probability for the subject having the neurological condition based on the outcome of the comparison.

2. Method of claim 1, wherein the first group of parameters is indicative of akinesia in the subject.

3. Method of claim 1 or 2, wherein the second group of parameters is indicative of rigor in the subject.

4. Method of any one of the preceding claims, wherein the third group of parameters is indicative of tremor in the patient.

5. Method of any one of claims 2 to 4, further comprising: providing a presence indicator for each one of akinesia, rigor and tremor as neurological symptoms, which indicate the severity of the respective neurological symptom in the patient.

6. Method of any one of claims 1 to 5, wherein providing probability for the subject having the neurological condition is further based on a comparison between the first groups of parameters, the second groups of parameters, and/or the third groups of parameters as obtained for the body part for the two sides of the subject's body.

7. Method of any one the preceding claims 1 to 6, wherein the motion signals have been obtained during guided movement of the subject based on movement instructions shown to the subject.

8. Method of any one of claims 1 to 7, further comprising: specifying at least one neurologically relevant segment within the received motion signals and processing only at least one relevant segment of the motion signals to determine the first group of parameters, the second group of parameters and the third group of parameters.

9. Method of any one of claims 1 to 8, wherein the step of comparing the formed vector to a plurality of reference vectors comprises, for the first group of parameters, applying a symbolic Fourier transformation to the motion signals, dividing the frequency domain into frequency bins and mapping the resulting Fourier coefficients to respective bins, and comparing the obtained histogram with histograms obtained in the same manner for the first group of parameters of the reference vectors.

10. Method of claim 9, wherein comparing the obtained histograms comprises application of a BOSS transformation to the motion signals.

11. Method of any one of claims 1 to 10, wherein the analysis of periodic components of the motion signals comprises an analysis of the spectral composition of the motion signals.

12. Method of any one of the preceding claims 1 to 11, wherein the neurological condition comprises Parkinson's disease.