Device and method for classification of a patient into at least two groups of subjects having a gait pathology

Abstract

The present invention relates to a device for obtaining a trained machine learning classification model for classifying a patient into at least two groups of subjects, wherein the subjects of said group have a gait pathology. The invention also relates to a device and method for classification of a patient in at least two groups of subjects, wherein the subjects of said group have a gait pathology.

Description

DEVICE AND METHOD FOR CLASSIFICATION OF A PATIENT INTO AT LEAST TWO GROUPS OF SUBJECTS HAVING A GAIT PATHOLOGYFIELD OF INVENTION

[0001] The present invention relates to the technical field of patient classification in a group of interest. More precisely, the invention relates to a device for obtaining a trained machine learning classification model for classifying a patient into at least two groups of subjects, wherein the subjects of said group have a gait pathology. The invention also relates to a device and method for classification of a patient in at least two groups of subjects, wherein the subjects of each group have a common gait pathology.BACKGROUND OF INVENTION

[0002] Classifying patients into specific groups based on gait pathology presents significant challenges in the field of healthcare and clinical diagnostics. These challenges arise from the inherent variability in gait patterns among individuals affected by similar pathologies, as well as the subjective nature of clinical assessment methods commonly used for diagnosis.

[0003] Gait patterns can exhibit substantial variability due to factors such as age, weight, coexisting medical conditions, and the severity of the underlying pathology. This variability makes it difficult to identify consistent and discriminative features for accurate classification. Moreover, certain gait pathologies may manifest overlapping symptoms, complicating the differentiation between different conditions.

[0004] The quality and quantity of gait data collected for analysis further compound these challenges. Factors such as sensor accuracy, placement variability, and environmental conditions can introduce noise and inconsistency into the data, hindering the development of robust classification models.

[0005] Furthermore, the subjective nature of clinical assessment poses a significant hurdle to accurate patient classification. Gait assessment often relies on subjective evaluation by clinicians, leading to variability and inconsistency in labeling patients with specific pathologies. This subjectivity undermines the reliability of labeled data used for model training and validation.

[0006] Therefore, there is a pressing need for improved devices and methods for classifying patients with gait pathology accurately.SUMMARY

[0007] This invention thus relates to a device for obtaining a trained machine learning classification model for classifying a patient into at least two groups of subjects, wherein the subjects of said group have a gait pathology, said device comprising: at least one input adapted to receive for each subject of a plurality of subjects, at least one Inertial Measurement Unit (IMU) recording previously obtained using at least one IMU sensor, wherein said IMU recording is representative of a gait pattern of said subject; at least one processor configured to: obtain said trained machine learning classification model by feeding said training dataset to a trainable machine learning classification model, and at least one output configured to provide the obtained trained machine learning classification model.

[0008] According to one embodiment, the invention relates to a device for obtaining a trained machine learning classification model for classifying a patient into one of at least two groups, wherein each group is associated to a gait pathology, said device comprising: at least one input configured to receive for each subject of a plurality of subjects, at least one Inertial Measurement Unit (IMU) recording previously obtained using at least one IMU sensor, said IMU recording being representative of a gait pattern of said subject; at least one processor configured to:o generate a training dataset by, defining at least one training sample for each subject of said plurality of subjects by:■ extracting at least one feature representative of said gait pattern from the at least one Inertial Measurement Unit (IMU) recording of said subject;■ labelling said recording into one of said at least two groups ;■ defining said at least one training sample for said subject as comprising the extracted at least one feature and the corresponding label; o obtain said trained machine learning classification model by feeding said training dataset to a trainable machine learning classification model; wherein the trainable machine learning classification model is configured to receive as input the at least one extracted feature and provide as output a classification prediction for the at least two groups; and at least one output configured to provide the obtained trained machine learning classification model.

[0009] According to one embodiment, at least one subject from the plurality of subjects is affected by the gait pathology associated with one of the at least two groups, and at least one other subject is affected by the gait pathology associated with another group of the at least two groups. The plurality of subjects may further comprise healthy subjects.

[0010] Advantageously the device of the present invention allows to obtain a trained classification model configured to classify the patient into a group of subjects suffering from a same gait pathology. The specific construction of the training dataset herein proposed advantageously allows to obtain a robust classification model, capable of effectively differentiating between the different types of gait pathologies.

[0011] Developing a high-performing classification model that can classify a patient into a specific group based solely on an IMU (Inertial Measurement Unit) recording offers several advantages. It can aid in the early detection of gait pathologies using readily accessible IMU recordings. Early diagnosis allows for prompt intervention and treatmentplanning, potentially preventing further deterioration of the condition and improving patient outcomes. By accurately classifying patients into specific groups based on their gait pathologies, healthcare professionals can tailor treatment plans to address individual needs. Personalized interventions can lead to more effective rehabilitation strategies and better patient outcomes. IMU recordings are non-invasive and cost-effective compared to traditional motion analysis systems or diagnostic tools. Utilizing IMU-based classification models can potentially reduce healthcare costs associated with gait assessments while maintaining diagnostic accuracy. Classification models integrated with IMU technology can provide real-time feedback on gait abnormalities, enabling continuous monitoring of patients’ progress during rehabilitation or treatment, while also allowing correction of the pathological gait using a device for electrical neuromuscular stimulation.

[0012] According to other advantageous aspects of the invention, the device comprises one or more of the features described in the following embodiments, taken alone or in any possible combination.

[0013] According to one embodiment, said at least one IMU recording corresponds to movements detected by said at least one IMU sensor while said subject is performing at least 10 steps. In other words, said at least one IMU recording has been acquired while the subject is performing at least 10 steps so the IMU recording comprises a signal representative of movements associated to the action of performing multiple steps (e.g., walking).

[0014] According to one embodiment, said at least one input is further adapted to receive for each subject of said plurality of subjects, at least one electromyogram (EMG) recording previously obtained using at least one EMG sensor, wherein said EMG recording is representative of said gait pattern of said subject. Said at least one EMG sensor being positioned in contact with the skin of the subject, preferably on at least one of the legs of the subject.

[0015] According to one embodiment, said at least one feature is extracted using at least one of a statistical analysis, a frequency domain analysis, and a time domain analysis.According to one embodiment, said at least one feature is extracted from the at least one Inertial Measurement Unit (IMU) recording and / or at least one electromyogram (EMG) recording using at least one of a statistical analysis, a frequency domain analysis, and a time domain analysis.

[0016] According to one embodiment, the machine learning classification model to be trained comprises (i.e., is based on) at least one of: a decision tree, random forest, support vector machine (SVM), and a neural network.

[0017] According to one embodiment, generating said training dataset further comprises extracting at least two features (e.g., from the at least one Inertial Measurement Unit (IMU) recording and / or at least one electromyogram (EMG) recording) and selecting one feature from the at least two features based on at least one of correlation analysis, feature importance ranking, and domain expertise.

[0018] According to one embodiment, said at least two groups of subjects comprises three groups, each group comprising two sub-groups, wherein the subjects from each subgroup comprise at least one of: a reduced dorsiflexion of at least one ankle in swing phase, a reduced dorsiflexion of at least one ankle in stance phase, a reduced flexion of at least one knee in swing phase, a genu recurvatum of at least one knee, a reduced range of motion of at least one hip, weak flexor muscles of at least one hip.

[0019] According to one embodiment, said at least two groups comprises three groups, each group of said three groups comprising two sub-groups; wherein each sub-group is associated to at least one of: a reduced dorsiflexion of at least one ankle in swing phase, a reduced dorsiflexion of at least one ankle in stance phase, a reduced flexion of at least one knee in swing phase, a genu recurvatum of at least one knee, a reduced range of motion of at least one hip, weak flexor muscles of at least one hip.

[0020] According to one embodiment, the trained machine learning classification model is configured to classify the input information (e.g., the patient) into three groups: a first group associated to a gait pathology linked to a knee pathology, a second group associated to a gait pathology linked to a hip pathology and a third group associated to a gait pathology linked to an ankle pathology.

[0021] According to one embodiment, the trained machine learning classification model is configured to classify the input information (e.g., the patient) into six groups associated to a gait pathology: (1) a reduced dorsiflexion of at least one ankle in swing phase, (2) a reduced dorsiflexion of at least one ankle in stance phase, (3) a reduced flexion of at least one knee in swing phase, (4) a genu recurvatum of at least one knee, (5) a reduced range of motion of at least one hip, (6) weak flexor muscles of at least one hip.

[0022] The present invention further relates to device for classification of a patient in one at least two groups, each group being associated to a gait pathology, using the trained machine learning classification model obtained according to the device for obtaining a trained machine learning classification model, said device comprising: at least one input adapted to receive at least one Inertial Measurement Unit (IMU) recording acquired on said patient; at least one processor configured to: extracting at least one feature representative of said gait pattern from the at least one Inertial Measurement Unit (IMU) recording; provide as an input to said trained classification model said so as to obtain a prediction of a classification of said patient into one of said at least two groups, and at least one output adapted to provide said prediction of classification of said patient in said at least two groups.

[0023] According to tone embodiment, the invention relates to a device for classification of a patient in at least two groups of subjects, wherein the subjects of said group have a gait pathology, using the trained machine learning classification model obtained thanks to the device for obtaining a trained machine learning classification model such as described above, said device comprising: at least one input adapted to receive at least one Inertial Measurement Unit (IMU) recording acquired on said patient; at least one processor configured to provide as an input to said trained classification model said at least one IMU recording so as to obtain a prediction of a classification of said patient into said at least two groups, andat least one output adapted to provide said prediction of classification of said patient in said at least two groups.

[0024] According to one embodiment, said prediction of classification is provided as an input to a device for electrical neuromuscular stimulation of said patient so as to obtain a stimulation profile for said patient, said stimulation profile being specific to said group of classification.

[0025] Indeed, Integration of IMU-based gait pathology classification with devices for electrical neuromuscular stimulation offers the advantage of guided therapy. By accurately classifying patients into specific gait pathology groups, clinicians can customize neuromuscular stimulation protocols tailored to address identified abnormalities. This targeted approach ensures that electrical stimulation is applied to the appropriate muscles and at the optimal timing during the gait cycle, maximizing therapeutic effectiveness. Additionally, real-time feedback from IMU recordings can inform adjustments to stimulation parameters, ensuring adaptive and personalized treatment plans that better address patients' specific gait impairments. As a result, guided neuromuscular stimulation therapy based on IMU-based classification not only enhances rehabilitation outcomes but also optimizes the utilization of electrical stimulation devices, leading to more efficient and effective treatment strategies.

[0026] According to another embodiment, the invention relates to a computer- implemented method for classification of a patient in at least two groups of subjects, wherein the subjects of said group have a gait pathology, using the trained machine learning classification model obtained thanks to the device for obtaining a trained machine learning classification model such as described above, said method comprising:- receiving at least one Inertial Measurement Unit (IMU) recording acquired on said patient;- providing as an input to said trained classification model said at least one IMU recording so as to obtain a prediction of a classification of said patient into said at least two groups, and- outputting said prediction of classification of said patient in said at least two groups.

[0027] In addition, the disclosure relates to a computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method for classification of a patient in at least two groups of subjects, wherein the subjects of said group have a gait pathology, compliant with any of the above execution modes.

[0028] The present disclosure further pertains to a computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the method for classification of a patient in at least two groups of subjects, wherein the subjects of said group have a gait pathology compliant with any of the above execution modes.

[0029] The present disclosure further pertains to a non-transitory program storage device, readable by a computer, tangibly embodying a program of instructions executable by the computer to perform the method for classification of a patient in at least two groups of subjects, wherein the subjects of said group have a gait pathology compliant with the present disclosure.

[0030] Such a non-transitory program storage device can be, without limitation, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor device, or any suitable combination of the foregoing. It is to be appreciated that the following, while providing more specific examples, is merely an illustrative and not exhaustive listing as readily appreciated by one of ordinary skill in the art: a portable computer diskette, a hard disk, a ROM, an EPROM (Erasable Programmable ROM) or a Flash memory, a portable CD-ROM (Compact-Disc ROM).DEFINITIONS

[0031] In the present invention, the following terms have the following meanings:

[0032] The terms “adapted” and “configured” are used in the present disclosure as broadly encompassing initial configuration, later adaptation or complementation of thepresent device, or any combination thereof alike, whether effected through material or software means (including firmware).

[0033] The term “processor” should not be construed to be restricted to hardware capable of executing software, and refers in a general way to a processing device, which can for example include a computer, a microprocessor, an integrated circuit, or a programmable logic device (PLD). The processor may also encompass one or more Graphics Processing Units (GPU), whether exploited for computer graphics and image processing or other functions. Additionally, the instructions and / or data enabling to perform associated and / or resulting functionalities may be stored on any processor- readable medium such as, e.g., an integrated circuit, a hard disk, a CD (Compact Disc), an optical disc such as a DVD (Digital Versatile Disc), a RAM (Random- Access Memory) or a ROM (Read-Only Memory). Instructions may be notably stored in hardware, software, firmware or in any combination thereof.

[0034] “Machine learning (ML)” designates in a traditional way computer algorithms improving automatically through experience, on the ground of training data enabling to adjust parameters of computer models through gap reductions between expected outputs extracted from the training data and evaluated outputs computed by the computer models.

[0035] “Datasets” are collections of data used to build an ML mathematical model, so as to make data-driven predictions or decisions. In “supervised learning” (i.e. inferring functions from known input-output examples in the form of labelled training data), three types of ML datasets (also designated as ML sets) are typically dedicated to three respective kinds of tasks: “training”, i.e. fitting the parameters, “validation”, i.e. tuning ML hyperparameters (which are parameters used to control the learning process), and “testing”, i.e. checking independently of a training dataset exploited for building a mathematical model that the latter model provides satisfying results.

[0036] A “neural network (NN)” designates a category of ML comprising nodes (called “neurons”), and connections between neurons modeled by “weights”. For each neuron, an output is given in function of an input or a set of inputs by an “activation function”. Neurons are generally organized into multiple “layers”, so that neurons of one layer connect only to neurons of the immediately preceding and immediately following layers.BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The present disclosure will be better understood, and other specific features and advantages will emerge upon reading the following description of particular and non-restrictive illustrative embodiments, the description making reference to the annexed drawings wherein:

[0038] Figure 1 is a block diagram representing schematically a particular mode of a device for obtaining a trained machine learning classification model compliant with the present disclosure;

[0039] Figure 2 is a is a flow chart showing successive steps executed with the device for obtaining a trained machine learning classification model of figure 1;

[0040] Figure 3 is a block diagram representing schematically a particular mode of a device for classification of a patient in at least two groups of subjects using the trained machine learning classification model obtained from the device represented in Figure 1, compliant with the present disclosure;

[0041] Figure 4 is a flow chart showing successive steps executed with the device for classification of figure 3; and

[0042] Figure 5 is a summary table of the characteristics associated with each group of classification.ILLUSTRATIVE EMBODIMENTS

[0043] The present description illustrates the principles of the present disclosure. It will thus be appreciated that those skilled in the art will be able to devise various arrangements that, although not explicitly described or shown herein, embody the principles of the disclosure and are included within its scope.

[0044] All examples and conditional language recited herein are intended for educational purposes to aid the reader in understanding the principles of the disclosure and the concepts contributed by the inventor to furthering the art, and are to be construed as being without limitation to such specifically recited examples and conditions.

[0045] Moreover, all statements herein reciting principles, aspects, and embodiments of the disclosure, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents as well as equivalents developed in the future, i.e., any elements developed that perform the same function, regardless of structure.

[0046] Thus, for example, it will be appreciated by those skilled in the art that the block diagrams presented herein may represent conceptual views of illustrative circuitry embodying the principles of the disclosure. Similarly, it will be appreciated that any flow charts, flow diagrams, and the like represent various processes which may be substantially represented in computer readable media and so executed by a computer or processor, whether or not such computer or processor is explicitly shown.

[0047] The functions of the various elements shown in the figures may be provided through the use of dedicated hardware as well as hardware capable of executing software in association with appropriate software. When provided by a processor, the functions may be provided by a single dedicated processor, a single shared processor, or a plurality of individual processors, some of which may be shared.

[0048] It should be understood that the elements shown in the figures may be implemented in various forms of hardware, software or combinations thereof. Preferably, these elements are implemented in a combination of hardware and software on one or more appropriately programmed general-purpose devices, which may include a processor, memory and input / output interfaces.

[0049] The present disclosure will be described in reference to a particular functional embodiment of a device 1 for obtaining a trained machine learning classification model 31, as illustrated on Figure 1.

[0050] The device 1 is adapted to produce a trained machine learning classification model 31 for classification of a patient into one or at least one of at least two groups (i.e., groups of subjects), wherein each group is associated to a gait pathology.

[0051] Groups of subjects (the plurality of subjects), notably the medical information associated with them, may be used to obtain said trained machine learning classificationmodel 31. According to the invention, a group of subjects with a gait pathology refers to a collection of individuals (i.e., plurality of subjects) who exhibit abnormal walking patterns or movements due to various musculoskeletal, neurological, or biomechanical issues. These individuals typically share common characteristics or symptoms related to their gait abnormalities, which may include reduced range of motion, altered joint mechanics, muscle weakness, or balance impairments. The classification of a patient into such a group allows for the study, analysis, and treatment of gait pathologies, as well as the development of interventions aimed at improving gait function and overall mobility.

[0052] Notably, the device 1 is configured to generate a training dataset and train a trainable machine learning classification model 20 using said training dataset in order to obtain the trained classification model 31. The device 1 may be configured to receive as input said training dataset (i.e.; from a database 10) or generate the training dataset in a module for constructing the training dataset.

[0053] The device 1 for training the trainable machine learning classification model 20 is associated with a device 2, represented on Figure 3, for classification of a patient in said at least two groups of gait pathologies, using the trained classification model 31 obtained from the device 1, which will be subsequently described. More precisely, the trained machine learning classification model 31 is configured to receive as input at least one Inertial Measurement Unit (IMU) recording 32 acquired on said patient. The trained machine learning classification model 31 generates as an output a prediction 61 of a classification of said patient into said at least two groups of subjects (i.e., at least two classes of gait pathologies).

[0054] Though the presently described devices 1 and 2 are versatile and provided with several functions that can be carried out alternatively or in any cumulative way, other implementations within the scope of the present disclosure include devices having only parts of the present functionalities.

[0055] Each of the devices 1 and 2 is advantageously an apparatus, or a physical part of an apparatus, designed, configured and / or adapted for performing the mentioned functions and produce the mentioned effects or results. In alternative implementations, any of the device 1 and the device 2 is embodied as a set of apparatus or physical parts ofapparatus, whether grouped in a same machine or in different, possibly remote, machines. The device 1 and / or the device 2 may e.g. have functions distributed over a cloud infrastructure and be available to users as a cloud-based service, or have remote functions accessible through an API.

[0056] The device 1 and the device 2 may be integrated in a same apparatus or set of apparatus, and intended to same users. In other implementations, the structure of the device 2 may be completely independent of the structure of the device 1, and may be provided for other users. For example, the device 2 may have a trained machine learning classification model 31 available to operators for classifying a patient into said at least two groups of subjects, wholly set from previous training effected upstream by other players with the device 1.

[0057] In what follows, the modules are to be understood as functional entities rather than material, physically distinct, components. They can consequently be embodied either as grouped together in a same tangible and concrete component, or distributed into several such components. Also, each of those modules is possibly itself shared between at least two physical components. In addition, the modules are implemented in hardware, software, firmware, or any mixed form thereof as well. They are preferably embodied within at least one processor of the device 1 or of the device 2.

[0058] The device 1 comprises a module 11 for receiving IMU recordings 22 obtained from a plurality of subjects and the trainable machine learning classification model 20, stored in one or more local or remote database(s) 10. The latter can take the form of storage resources available from any kind of appropriate storage means, which can be notably a RAM or an EEPROM (Electrically-Erasable Programmable Read-Only Memory) such as a Flash memory, possibly within an SSD (Solid-State Disk).

[0059] According to one embodiment, the subjects comprised in said plurality of subjects are health subject (i.e., presenting no gait pathology), or subjects affected by at least one of the gait pathology associated to one of the at least two groups of gait pathologies.

[0060] One group / class may be associated to one of the following gait pathology:an antalgic gait: a limping gait often caused by pain, where the affected individual minimizes weight-bearing on the affected limb; an ataxic gait: a staggering or unsteady gait associated with cerebellar dysfunction, characterized by poor coordination and balance; a parkinsonian gait: a shuffling gait with reduced arm swing and stooped posture, typical of individuals with Parkinson's disease; a spastic gait: a stiff, jerky gait often seen in individuals with conditions affecting the central nervous system, such as cerebral palsy or multiple sclerosis; a scissors gait: a crossing of the legs while walking, resembling the movement of scissors, commonly seen in individuals with cerebral palsy or other conditions affecting muscle tone; a steppage gait: a high-stepping gait where the foot is lifted excessively during the swing phase, often seen in individuals with foot drop due to nerve damage or muscle weakness; a waddling gait: a rolling or swaying gait characterized by excessive side-to-side movement of the pelvis, often seen in individuals with muscular dystrophy or hip dislocation; a hemiplegic gait: a gait pattern observed in individuals with hemiplegia, characterized by dragging one side of the body while walking, typically due to stroke or cerebral palsy affecting one side of the body; a cerebellar gait: a wide-based, unsteady gait with irregular steps and poor coordination, associated with lesions or dysfunction of the cerebellum; a dystonic gait: a gait pattern characterized by involuntary muscle contractions and twisting movements, often seen in individuals with dystonia or other movement disorders; a myopathic gait: a waddling or wide-based gait associated with muscle weakness or wasting, typically seen in individuals with muscular dystrophy or other myopathies; propulsive gait: A stooped, forward-leaning gait with reduced arm swing and rigid posture, often seen in individuals with Parkinson's disease or other basal ganglia disorders.

[0061] More precisely, each subject of the plurality of subjects may have its movements recorded by at least one IMU sensor. Such IMU recordings 22 may be performed for a predetermined duration, such as the number of steps taken by the subject. For example, the recording may comprise at least ten consecutive steps performed by the subject. In other words, the recording(s) may have been acquired while the subject was walking, the duration of the recording comprising at least 10 steps of the subject.

[0062] To illustrate, the IMU recordings 22 may be obtained using at least one of an accelerometer, a gyroscope, or a magnetometer. Advantageously, such sensor may be used in combination with a camera and the IMU recordings 22 may be synchronized with the camera recordings to help with the subsequent annotation process (e.g. extraction of features and labelling). In order to synchronize the recordings, standardization may be performed by converting the recordings into a common format or data structure, normalizing the recordings to a common scale or range to eliminate differences in magnitude or units, aligning the recordings to synchronize them in time and the like.

[0063] Optionally, module 11 may be further configured to receive at least one EMG recording obtained from a plurality of subjects, stored in one or more local or remote database(s) 10. The EMG recording may be obtained using EMG sensors either positioned on the skin surface, above muscles of interest to detect and record electrical activity generated by muscle contractions. Alternatively, intramuscular EMG electrodes may be used. These electrodes are inserted directly into the muscle tissue to record electrical activity within specific muscles with greater precision. More precisely, when a muscle contracts, it generates electrical signals known as action potentials or electromyographic signals, which can be detected by the EMG sensors. The EMG sensors converts these electrical signals into analog or digital data that can be recorded and analyzed.

[0064] The device 1 further comprises optionally a module 12 for preprocessing the IMU recordings 22. Preprocessing may include data cleaning, normalization and temporal alignment, resampling, filtering and noise reduction.

[0065] It may also transform the recordings 22, (e.g. by signal compression or decompression).

[0066] According to various configurations, the module 12 is adapted to execute only part or all of the above functions, in any possible combination, in any manner suited to the following processing stage.

[0067] In advantageous modes, the module 12 is configured for preprocessing the recordings 22 so as to have the recordings standardized. Standardizing the recordings 22 from different sensors (e.g. IMU and EMG sensors) involves ensuring that the recordings 22 from each sensor is transformed or processed in a consistent manner, such that they can be directly compared or combined for further analysis. This may enhance the efficiency of the downstream processing by the device 1. Such standardization may be particularly useful when exploited recordings originate from different sources (e.g. different sensors), including possibly different sensors types.

[0068] The device may comprise a module 13 for the construction of the training dataset using the received IMU recordings 22 and optionally the EMG recording.

[0069] Construction of the training dataset may involve extracting at least one feature characteristic of the gait pathology of the subject from each IMU recording and / or optionally from each EMG recording.

[0070] The at least one feature may further be extracted using at least one of a statistical analysis, a frequency domain analysis, and a time domain analysis.

[0071] Statistical analysis involves the examination and interpretation of recordings (e.g. IMU recordings 22 and / or EM recordings) through the application of statistical methods and techniques. It aims to describe, summarize, and draw inferences from the recordings 22 by quantifying various aspects of its distribution, central tendency, variability, and relationships between variables.

[0072] Frequency domain analysis involves the examination of the recordings (e.g. IMU recordings 22 and / or EM recordings) in terms of their frequency components. It typically involves transforming time-domain signals into the frequency domain using techniques such as Fourier analysis. This analysis reveals the presence and characteristics of periodic components, oscillations, and spectral features within the recordings (e.g. IMU recordings 22 and / or EM recordings).

[0073] Time domain analysis involves the direct analysis of the recordings (e.g. IMU recordings 22 and / or EM recordings) in the time domain. It focuses on the behavior and characteristics of signals over time, including amplitude variations, temporal patterns, and dynamic changes. Time domain analysis provides insights into the temporal aspects of the recordings (e.g. IMU recordings 22 and / or EM recordings) without considering frequency components.

[0074] More precisely, such feature may be at least one of the following: mean acceleration: average acceleration values along each axis (x, y, z) during a gait cycle. peak acceleration: maximum acceleration values observed along each axis during a gait cycle. root mean square (RMS) acceleration: a measure of the overall magnitude of acceleration, calculated as the square root of the average of the squares of acceleration values. jerk: rate of change of acceleration over time, indicating abrupt changes in movement. stride length: distance covered during a single gait cycle, typically calculated from the integration of acceleration recordings 22. stride time: time duration of a single gait cycle, representing the time taken to complete a step. step time: time duration of a single step, providing insights into the timing of foot placement and lift-off. step symmetry: comparison of temporal parameters (e.g., step time) between left and right steps, indicating asymmetry in gait. cadence: number of steps per unit time, often expressed as steps per minute, step width: lateral distance between the feet during walking, indicating stability and balance. foot clearance: minimum height of the foot during the swing phase, reflecting the risk of tripping or foot drop. range of motion: maximum and minimum angles of joint movements during the gait cycle, indicating flexibility and mobility.gait variability: variability in gait parameters such as step length, step time, and stride time, providing insights into gait stability and consistency. frequency components: analysis of frequency domain characteristics, such as dominant frequencies or spectral features, revealing oscillatory patterns or rhythmicity in movement. temporal patterns: identification of specific patterns or sequences of movements within the gait cycle, such as heel strike, toe-off, and mid-stance phases.

[0075] When multiple features are identified, a feature selection may be performed to identify the most informative and discriminative features for classification.

[0076] In other words, when two or more features are extracted (e.g., from the at least one Inertial Measurement Unit (IMU) recording and / or at least one electromyogram (EMG) recording) and a sub-set of features may be selected from said at least two features based on at least one of correlation analysis, feature importance ranking, and domain expertise. The term domain expertise refers to the specialized knowledge or experience that an expert has in a particular field. In the context of the present invention, it means that the selection of one feature (from at least two) can be based on insights or judgment from a user (e.g., a professional) who have deep knowledge of the specific domain of gait pathologies.

[0077] Feature importance ranking may be determined using various techniques, including model-based methods, statistical analyses, and feature selection algorithms. Model-based methods may include decision tree-based approaches, such as Gini importance (mean decrease in impurity) and permutation importance, which evaluate the contribution of each feature to the model’s performance. Additionally, SHAP (SHapley Additive exPlanations) values may be utilized to assess both global and local feature importance based on game-theoretic principles. In the case of linear models, feature importance may be determined using coefficient magnitudes or LASSO regression, where less relevant features are shrunk toward zero. Statistical methods such as correlation analysis (e.g., Pearson, Spearman, Kendall) and mutual information analysis may further be employed to quantify the relationship between individual features and the target variable. Moreover, techniques such as ANOVA (Analysis of Variance) may be used toassess categorical feature relevance in regression tasks. Feature selection algorithms, including Recursive Feature Elimination (RFE), Principal Component Analysis (PCA), and Minimum Redundancy Maximum Relevance (mRMR), may also be applied to identify features that maximize predictive power while minimizing redundancy.

[0078] In tree-based models like random forests, feature importance scores may be computed based on how much each feature contributes to the model's accuracy. For IMU recordings 22, feature importance from trees could identify features such as range of motion, cadence, and step width as the most influential for classifying gait pathologies.

[0079] Correlation-based feature selection evaluates the pairwise correlations between features and selects the least correlated subset. For IMU recordings 22, features such as stride length and step time may be highly correlated, so only one of these features would be retained to avoid redundancy in the classification model.

[0080] After or before identification of the features, the IMU recordings 22 are classified (i.e., associated to a label) into said at least two groups. Said classification may be performed manually by an operator or a physician or automatically, using a classification model. In other words, module 13 is further configured to perform labeling of said recordings (e.g., IMU and / or EMG) so as to associate each recording to one of said at least two groups.

[0081] In one example, the labelling may be based on a classification system such as the one illustrated in Figure 5. In this case six groups are defined. The table consists of three main groups labeled as Group 1, Group 2, and Group 3. Under each group, there are subgroups labeled as Subgroup la, Subgroup lb, Subgroup 2a, Subgroup 2, Subgroup 3a, and Subgroup 3b.

[0082] For each combination of group and subgroup, there are specific criteria listed related to gait pathologies or abnormalities. Here's a summary of the criteria listed under each subgroup: reduced dorsiflexion of at least one ankle in swing phase: this refers to a limitation in the upward movement of the foot at the ankle joint during the swing phase of walking. Dorsiflexion is the movement that brings the top of the footcloser to the shin, and reduced dorsiflexion during the swing phase may affect the clearance of the foot while walking; reduced dorsiflexion of at least one ankle in stance phase: this indicates a limitation in the upward movement of the foot at the ankle joint while the foot is bearing weight during the stance phase of walking. It differs from reduced dorsiflexion in the swing phase as it occurs during the weight-bearing phase of the gait cycle. reduced flexion of at least one knee in swing phase: this refers to a limitation in the bending of the knee joint during the swing phase of walking. Reduced knee flexion may affect the clearance of the foot during swing, potentially leading to tripping or dragging of the foot. genu recurvatum of at least one knee: genu recurvatum is a condition characterized by hyperextension of the knee joint beyond its normal range of motion, resulting in the knee bending backward. This abnormality may lead to instability during walking and may be associated with other musculoskeletal issues. reduced range of motion of at least one hip: this refers to a limitation in the movement range of the hip joint. Reduced hip range of motion can affect various aspects of gait, including stride length, balance, and overall mobility. weak flexor muscles of at least one hip: this indicates weakness in the muscles responsible for flexing the hip joint. Weakness in hip flexor muscles can result in difficulties with lifting the leg during walking, leading to alterations in gait patterns and potential compensatory movements.

[0083] In Figure 5, each group of subjects is characterized by at least one of the above criteria. Notably: subjects in Group 1, Subgroup la show a reduced dorsiflexion of at least one ankle in the swing phase; subjects in Group 1, Subgroup lb show a reduced dorsiflexion of at least one ankle in the swing phase. Similarly, reduced dorsiflexion of at least one ankle in the stance phase is also present. Additionally, genu recurvatum of at least one knee is present.subjects in Group 2, Subgroup 2a show, optionally, reduced dorsiflexion of at least one ankle in the swing phase. Additionally, reduced flexion of at least one knee in the swing phase is present; subjects in Group 2, Subgroup 2b show, optionally, reduced dorsiflexion of at least one ankle in the swing phase. Similarly, reduced flexion of at least one knee in the swing phase is present. Additionally, genu recurvatum of at least one knee is present; subjects in Group 3, Subgroup 3a show, optionally, reduced dorsiflexion of at least one ankle in the swing phase may be present. Optionally, reduced dorsiflexion of at least one knee in the swing phase may also be present. Moreover, reduced range of motion of at least one hip, and weak flexor muscles of at least one hip are also present in this subgroup; subjects in Group 3, Subgroup 3b show, optionally, reduced dorsiflexion of at least one ankle in the swing phase. Reduced dorsiflexion of at least one ankle in the stance phase is present. Optionally, reduced flexion of at least one knee in the swing phase may be present. Additionally, genu recurvatum of at least one knee is present. Reduced range of motion of at least one hip and weak flexor muscles of at least one hip are also present in this subgroup.

[0084] Module 13 may be therefore configured to define for each subject at least one training sample. One training sample may be defined to comprise at least one extracted feature from at least one recording of said subject and the corresponding label. In one example, if IMU and EMG recordings (i.e. signals) had been acquired simultaneously, one training sample may be defined to comprise one or more extracted feature from at least one IMU recording and one or more extracted feature from at least one EMG recording and the corresponding label associates to at least one of the two groups. Multiple training samples may be defined, notably when the recording used had been acquired with days or months apart.

[0085] A subject may for example present two gait pathologies so that in that case the training sample could be associated with the two gait pathologies labels.

[0086] The device 1 further comprises a module 14 configured to train the trainable machine learning classification model 20 using the training dataset constructed (orreceived) by module 13. The architecture of the trainable machine learning classification model 20 is configured to receive as input the extracted at least one feature and provide as output a classification prediction for the at least two groups.

[0087] The trainable machine learning classification model 20, trained in module 14, comprises at least one of a decision tree, random forest, support vector machine (SVM), and a neural network.

[0088] A decision tree is a supervised machine learning algorithm used for classification and regression tasks. It recursively partitions the feature space into regions and assigns a class label or predicts a continuous value in each region based on the values of input features. It is represented as a tree structure where each internal node represents a decision based on a feature, and each leaf node represents a class label or regression value.

[0089] Random forest is an ensemble learning method that constructs a multitude of decision trees during training and outputs the class label or regression value that is the mode of the classes or mean prediction of the individual trees. It introduces randomness both in the sampling of the data and the features used for training each tree, leading to improved generalization and reduced overfitting compared to individual decision trees.

[0090] For multiclass classification of the present invention, each decision tree independently predicts a class label, and the final classification output is determined through majority voting. In the case of multilabel classification, module 13 is configured to train multiple decision trees, each specialized in predicting the presence or absence of a specific label, with final predictions aggregated across the ensemble. During training, the Random Forest optimizes a Gini impurity or entropy-based information gain criterion at each node to determine the most discriminative features for splitting. Feature selection is further enhanced by feature importance ranking, computed based on metrics such as mean decrease in impurity (MDI) or permutation importance. The hyperparameters of the Random Forest, including the number of trees, maximum depth, and minimum samples per split, may be tuned using optimization techniques such as grid search or Bayesian optimization to improve classification accuracy for medical datasets.

[0091] Support Vector Machine (SVM) is a supervised machine learning algorithm used for classification and regression tasks. It works by finding the hyperplane that bestseparates the classes in the feature space. The hyperplane is chosen to maximize the margin between the classes, and data points closest to the hyperplane, called support vectors, influence its position. SVM can be linear or non-linear, and it can handle both linearly separable and non-linearly separable data using techniques like the kernel trick. The SVM is trained to optimize a decision boundary that maximizes the margin between different classes in a high-dimensional feature space. Strategies such as one-vs-one (OvO) or one-vs-all (OvA) may be used to extend the binary SVM framework. In the case of multilabel classification, the system applies an ensemble of binary SVM classifiers, each trained to predict the presence or absence of a specific label. The training process may be configured to minimize a hinge loss function, optionally incorporating class-weighted penalties to address data imbalance commonly encountered in medical datasets. The optimization of the SVM model may be performed using techniques such as Sequential Minimal Optimization (SMO) or Quadratic Programming (QP) solvers to efficiently determine the optimal support vectors and classification hyperplane. Kernel functions, including linear, polynomial, radial basis function (RBF), or sigmoid kernels, may be selected based on dataset characteristics to enhance model performance and capture complex decision boundaries.

[0092] A neural network is a computational model inspired by the structure and function of the human brain. It consists of interconnected nodes, called neurons, organized in layers. Information flows from the input layer through one or more hidden layers to the output layer. Each connection between neurons is associated with a weight, and neurons apply an activation function to the weighted sum of inputs to produce an output. Neural networks can learn complex patterns and relationships in data through iterative optimization algorithms, such as gradient descent. In this example, to obtain the trained classification model 31 each training sample of the training dataset is iteratively provided so that the extracted feature(s) are feed to the machine learning classification model 20 under training and the label is used as ground truth, still convergence.

[0093] In the example of neural network, the training process may utilize a loss function adapted to handle multiple target labels per input training sample, such as binary crossentropy for multilabel classification or categorical cross-entropy for multiclass classification. In some embodiments, when addressing class imbalance commonlypresent in medical datasets, the system employs focal loss, wherein a modulating factor is applied to reduce the impact of well-classified samples and focus on harder- to-classify cases. The neural network training further incorporates an optimization algorithm such as foe example an Adam optimizer, which dynamically adjusts learning rates based on the first and second moments of the gradient, thereby improving convergence and stability. Alternative optimization methods, including RMSprop or SGD with momentum, may also be employed to enhance training efficiency depending on computational constraints and dataset characteristics.

[0094] Once the training completed, the module 14 is configured to output the trained classification model 31. The trained classification model 31 may then by stored in one or more local or remote database(s) 10. The latter can take the form of storage resources available from any kind of appropriate storage means, which can be notably a RAM or an EEPROM (Electric ally-Eras able Programmable Read-Only Memory) such as a Flash memory, possibly within an SSD (Solid-State Disk).

[0095] Performances of the trained machine learning classification model may be evaluated using a validation dataset. Performances checking may involve calculating metrics such as accuracy, precision, recall, and Fl -score to assess the model's classification capabilities.

[0096] In its automatic actions, the device 1 may for example execute the following process (Figure 2):- receiving for each subject of said plurality of subjects, at least one IMU recording 22 and optionally at least one EMG recording previously obtained using at least one IMU sensor, wherein said IMU recording is representative of a gait pattern of said subject (step 41),- preprocessing the at least one IMU recording 22 and optionally the at least one EMG recording (step 42),- constructing the training dataset by, for each recording of each subject of said plurality of subjects, extracting at least one feature characteristic of said gait pattern; and classifying said recording into said at least two groups of subjects based on said extracted feature (step 43),obtaining said trained machine learning classification model 31 by feeding said training dataset to the trainable machine learning classification model 20 (step 44).

[0097] The present invention also relates to a device 2 for classification of a patient in at least two groups of subjects, wherein the subjects of said group have a gait pathology, using the trained machine learning classification model 31 obtained from the device 1, as described above. The device 2 will be described in reference to a particular function embodiment as illustrated in Figure 3.

[0098] The device 2 is adapted to receive as input at least one IMU recording acquired on a patient. Optionally, the device 2 is further adapter to receive at least one EMG recording.

[0099] The device 2 is adapted to provide as an output a prediction 61 of a classification of said patient into said at least two groups.

[0100] The device 2 comprises a module 15 for receiving the trained machine learning classification model 31 and the at least one IMU recording 32 acquired on said patient, and optionally the at least one EMG recording stored in one or more local or remote database(s) 10. The latter can take the form of storage resources available from any kind of appropriate storage means, which can be notably a RAM or an EEPROM (Electrically- Erasable Programmable Read-Only Memory) such as a Flash memory, possibly within an SSD (Solid-State Disk). In advantageous embodiments, the trained machine learning classification model 31 and all its parameters have been previously generated by a system including the device 2 for training. Alternatively, the trained machine learning classification model 31 and its parameters are received from a communication network.

[0101] The device 2 further comprises optionally a module 16 for preprocessing the at least one IMU recording 32 acquired on said patient and optionally the at least one EMG recording. Preprocessing may include data cleaning, normalization and temporal alignment, resampling, filtering and noise reduction.

[0102] The device 2 further comprises a module 17 configured to provide the at least one IMU recording 32 and optionally the at least one EMG recording acquired on saidpatient to said trained machine learning classification model 31 so as to generate a corresponding prediction 61.

[0103] The device 2 may interact with a user interface 18, via which information can be entered and retrieved by a user. The user interface 18 includes any means appropriate for entering or retrieving data, information or instructions, notably visual, tactile and / or audio capacities that can encompass any or several of the following means as well known by a person skilled in the art: a screen, a keyboard, a trackball, a touchpad, a touchscreen, a loudspeaker, a voice recognition system.

[0104] In its automatic actions, the device 2 may for example execute the following process (Figure 4):- receiving the at least one IMU recording 32 acquired on said patient and optionally the at least one EMG recording (step 51),- optionally preprocessing the at least one IMU recording 32 and / or the at least one EMG recording (step 52),- providing the IMU recording 32 and optionally the at least one EMG recording to said trained machine learning classification model 31 so to obtain a prediction 61 of a classification of said patient into said at least two groups (step 53).

[0105] In its automatic actions, the device 2 may for example execute the following process (not illustrated):- receiving the at least one IMU recording 32 acquired on said patient and optionally the at least one EMG recording (step 51),- optionally preprocessing the at least one IMU recording 32 and / or the at least one EMG recording (step 52),- extracting at least one feature representative of said gait pattern from the at least one Inertial Measurement Unit (IMU) recording 32 and / or the at least one EMG recording;- providing the at least one extracted feature to said trained machine learning classification model 31 so to obtain a prediction 61 of a classification of said patient into said at least two groups (step 53).

[0106] Further to the generation of the prediction 61, the device 2 may be configured to provide said prediction 61 to a device for electrical neuromuscular stimulation of said patient so as to obtain a stimulation profile for said patient that is specific to the gait pathology group in which said patient has been classified.

[0107] Such device for electrical neuromuscular stimulation may include several elements.

[0108] A first element is a multi-channel electrostimulator configured to send electric pulses to the muscles of the patient to stimulate them.

[0109] The electro stimulator includes a portable electrical muscle stimulation (EMS) generator that may be housed within a housing such as a bag or a vest. The bag may be carried on the back or on the front or on one shoulder, or around the waist. Alternatively, the generator may be housed in a pocket of an article of clothing, such as trousers.

[0110] The generator may deliver a voltage comprised between 0 and 350 V and a current intensity comprised between 0 and 170 mA. The generator is associated with an intensity variator to deliver a pulse sequence with a pre-determined intensity, width and frequency.

[0111] The generator may have several channels that can be each configured to deliver a different pulse sequence.

[0112] The generator is in electrical connection with a set of electrodes configured to be positioned on the skin of the patient, either on the motor points of muscles, where nerve endings enter the muscle or on sensory receptors in the skin that are innervated or at sensory nerves endings.

[0113] The electrodes may be connected to the generator via conductive wires, electrical traces, conductive fibers, or a combination thereof. The conductive wires, electrical traces, conductive fibers, or a combination thereof can be embedded within the article of clothing, for instance in a layer of the article of clothing or interwoven with fibers used to make the article of clothing.

[0114] The electrodes may comprise several layers including a contact layer configured to attach to the patient’s skin, a connector layer configured to be connected to thegenerator and conductive layers positioned between the contact layer and connector layer. The contact layer can be made of a biocompatible polymeric layer. It may include an adhesive and / or a hydrogel. Alternatively, the contact layer may be dry and may require that a conductive gel or a hydrating lotion hydrates the skin surface of the patient.

[0115] The electrodes may be coupled to an inner surface of the article of clothing by adhesives, clips, straps, hook-and-loop fasteners, stitches or a combination thereof. The electrodes may be positioned such that when the patient puts the article of clothing on, the contact layer of the electrodes is automatically positioned in contact with the skin over the motor points, the sensory receptors or the sensory nerves endings that needs to be stimulated. Advantageously, the electrodes can be detached from the article of clothing to allow replacing a defective electrode.

[0116] A second element of the device for electrical neuromuscular stimulation is at least one IMU sensor and at least one pressure detection unit. The IMU sensor may be the as the one used to capture the IMU recordings 32.

[0117] The IMU sensor includes at least one motion sensor such as an accelerometer, a gyroscope, a magnetometer and / or a combination thereof. Advantageously, the IMU sensor includes three motion sensors for each leg and one motion sensor for the hip. A first motion sensor may be positioned on the thigh, a second motion sensor may be positioned on the calf and a third motion sensor may be positioned on the foot. The motion sensors are configured to sense the three-dimensional movements of the leg throughout a gait cycle. The motion sensors may be embedded within the article of clothing, for instance in a layer of the article of clothing or interwoven with fibers used to make the article of clothing.

[0118] Optionally, the device for electrical neuromuscular stimulation may further comprise an EMG sensor.

[0119] The pressure detection unit includes pressure sensors that may be positioned under the feet to sense the strength of a contact of the different parts of the patient’s feet with the ground. The expression “strength of a contact” refers to the pressure exerted by a at least one region of one foot on the ground, the pressure being defined as a force per unit of area. Advantageously, the pressure sensors may be included in an insole or a soleof a shoe. For instance, the pressure detection unit may include between 1 and 1000 sensors distributed on the sole to collect data on the pressure applied, the position and the pressure changes when the patient is walking, running, jumping, climbing, descending, sitting or standing. In a preferred embodiment, there are five pressure sensors per foot. Two pressure sensors may be positioned under the heel, on the medial and lateral side, and a third pressure sensor may be positioned under the big toe. The fourth and fifth sensors may be positioned under the metatarsal bones.

[0120] The pressure sensors may be capacitive sensors, resistive sensors, piezoelectric sensors, piezoresistive sensors or a combination thereof. The pressure sensors provide an electrical signal output, which is either a voltage or a current, that is proportional to the pressure exerted on said pressure sensors.

[0121] Advantageously, the device for electrical neuromuscular stimulation may include other sensors such as electromyography sensors to access muscle fatigue and movement intention or encoders positioned at the legs joints to access the angular position of the different parts on the legs.

[0122] For instance, the patient motion data coming from the motion sensors and the patient foot plantar pressure data coming from the pressure sensors may be used to determine the foot strike pattern, the foot inclination angle, the tibia angle, the hip flexion and extension, the trunk lean, the ankle inversion and eversion, the foot progression angle, the pelvic drop, the knee flexion and extension, the stride length, or the displacement of the center of mass, the speed of gait, the cadence, etc.

[0123] A third element of the device for electrical neuromuscular stimulation is a device for computing a start timepoint and an end timepoint of at least one sequence of electrical pulses to apply to at least one of a nerve and a muscle of a patient using the electrodes. Such device may receive the prediction 61 and the additional sensor data (such as pressure data) to compute the start timepoint and an end timepoint of the at least one sequence of electrical pulses. Such device may be housed within the housing with the generator. Such device may communicate with a control device such as a tablet, a PC or a smartphone, hosting the user interface. The communication may be wireless, using Bluetooth, WiFi or other protocols. The control device may be used to parameter the device for electricalneuromuscular stimulation and visualize information on the state of the device for electrical neuromuscular stimulation and on how the patient is faring.

[0124] A particular apparatus may embody the device 1 as well as the device 2 described above. It corresponds for example to a workstation, a laptop, a tablet, a smartphone, or a head- mounted display (HMD).

[0125] That apparatus comprises the following elements, connected to each other by a bus of addresses and data that also transports a clock signal:- a microprocessor (or CPU);- a graphics card comprising several Graphical Processing Units (or GPUs) and a Graphical Random Access Memory (GRAM); the GPUs are quite suited to image processing, due to their highly parallel structure;- a non-volatile memory of ROM type;- a RAM;- one or several I / O (Input / Output) devices such as for example a keyboard, a mouse, a trackball, a webcam; other modes for introduction of commands such as for example vocal recognition are also possible;- a power source; and- a radiofrequency unit.

[0126] According to a variant, the power supply is external to the apparatus.

[0127] The apparatus also comprises a display device of display screen type directly connected to the graphics card to display synthesized images calculated and composed in the graphics card. According to a variant, a display device is external to the apparatus and is connected thereto by a cable or wirelessly for transmitting the display signals. The apparatus, for example through the graphics card, comprises an interface for transmission or connection adapted to transmit a display signal to an external display means such as for example an LCD or plasma screen or a video -projector. In this respect, the RF unit can be used for wireless transmissions.

[0128] It is noted that the word "register" used hereinafter in the description of memories can designate in each of the memories mentioned, a memory zone of low capacity (some binary data) as well as a memory zone of large capacity (enabling a whole program to bestored or all or part of the data representative of data calculated or to be displayed). Also, the registers represented for the RAM and the GRAM can be arranged and constituted in any manner, and each of them does not necessarily correspond to adjacent memory locations and can be distributed otherwise (which covers notably the situation in which one register includes several smaller registers).

[0129] When switched-on, the microprocessor loads and executes the instructions of the program contained in the RAM.

[0130] As will be understood by a skilled person, the presence of the graphics card is not mandatory, and can be replaced with entire CPU processing and / or simpler visualization implementations.

[0131] In variant modes, the apparatus may include only the functionalities of the device1, and not the learning capacities of the device 2. In addition, the device 1 and / or the device 2 may be implemented differently than a standalone software, and an apparatus or set of apparatus comprising only parts of the apparatus may be exploited through an API call or via a cloud interface.

Claims

CLAIMS1. A device (1) for obtaining a trained machine learning classification model (31) for classifying a patient into one of at least two groups, wherein each group is associated to a gait pathology, said device (1) comprising: at least one input configured to receive for each subject of a plurality of subjects, at least one Inertial Measurement Unit (IMU) recording (22) previously obtained using at least one IMU sensor, said IMU recording being representative of a gait pattern of said subject; at least one processor configured to: o generate a training dataset by, defining at least one training sample for each subject of said plurality of subjects, wherein defining the training sample comprises:■ extracting at least one feature representative of said gait pattern from the at least one Inertial Measurement Unit (IMU) recording (22) of said subject;■ labelling said recording into one of said at least two groups ;■ defining said at least one training sample for said subject as comprising the extracted at least one feature and the corresponding label; o obtain said trained machine learning classification model (31) by feeding said training dataset to a trainable machine learning classification model (20); wherein the trainable machine learning classification model (20) is configured to receive as input the at least one extracted feature and provide as output a classification prediction for the at least two groups; and at least one output configured to provide the obtained trained machine learning classification model (31).

2. The device according to claim 1, wherein said at least one IMU recording (22) corresponds to movements detected by said at least one IMU sensor while said subject is performing at least 10 steps.

3. The device according to claim 1 or 2, wherein said at least one input is further configured to receive for each subject of said plurality of subjects, at least one electromyogram (EMG) recording previously obtained using at least one EMG sensor, wherein said EMG recording is representative of said gait pattern of said subject.

4. The device according to any of claims 1 to 3, wherein said at least one feature is extracted using at least one of a statistical analysis, a frequency domain analysis, and a time domain analysis.

5. The device according to any of claims 1 to 4, wherein the trained machine learning classification model (31) comprises at least one of a decision tree, random forest, support vector machine (SVM), and a neural network.

6. The device according to any of claims 1 to 5, wherein generating said training dataset further comprises extracting at least two features and selecting one feature from the at least two features based on at least one of correlation analysis, feature importance ranking, and domain expertise.

7. The device according to any of claims 1 to 6, wherein said at least two groups comprises three groups, each group of said three groups comprising two subgroups; wherein each sub-group is associated to at least one of: a reduced dorsiflexion of at least one ankle in swing phase, a reduced dorsiflexion of at least one ankle in stance phase, a reduced flexion of at least one knee in swing phase, a genu recurvatum of at least one knee, a reduced range of motion of at least one hip, weak flexor muscles of at least one hip.

8. A device (2) for classification of a patient in one at least two groups , each group being associated to a gait pathology, using the trained machine learning classification model (31) obtained according to any of claims 1 to 7, said device comprising: at least one input configured to receive at least one Inertial Measurement Unit (IMU) recording (32) acquired on said patient; at least one processor configured to:extracting at least one feature representative of said gait pattern from the at least one Inertial Measurement Unit (IMU) recording (32); provide as an input to said trained classification model (31) said so as to obtain a prediction (61) of a classification of said patient into one of said at least two groups, and at least one output configured to provide said prediction (61) of classification of said patient in said at least two groups.

9. The device according to claim 8, wherein said prediction (61) of classification of said patient in said at least two groups is provided as an input to a device for electrical neuromuscular stimulation of said patient so as to obtain a stimulation profile for said patient, said stimulation profile being specific to said group of classification.

10. A computer-implemented method for classification of a patient in one at least two groups , each group being associated to a gait pathology, using the trained machine learning classification model (31) obtained from the device according to any of claims 1 to 7, said method comprising: receiving at least one Inertial Measurement Unit (IMU) recording (32) acquired on said patient; extracting at least one feature representative of said gait pattern from the at least one Inertial Measurement Unit (IMU) recording (32); providing as an input to said trained classification model (31) said so as to obtain a prediction (61) of a classification of said patient into one of said at least two groups, and outputting said prediction (61) of classification of said patient in said at least two groups.

11. A non-transitory program storage device, readable by a computer, tangibly embodying a program of instructions executable by the computer to perform the method according to claim 10.

12. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method according to claim 10.

13. A computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the method according to claim 10.

Images