EP4666289A1

EP4666289A1 - Personalized testing of muscle performance

Info

Publication number: EP4666289A1
Application number: EP23705968.8A
Authority: EP
Inventors: Sofia POMAKI POMAKIDOU
Original assignee: Individual
Current assignee: EVO HUMAN PERFORMANCE S.A.
Priority date: 2023-02-13
Filing date: 2023-02-13
Publication date: 2025-12-24
Also published as: WO2024170061A1

Abstract

A computer-implemented method of personalized testing of a mammal's muscle performance with the use of artificial intelligence models and predictive models, by using as input real time biomechanical data obtainable from well-known wearable electronic devices. The mammal's muscle performance is personalized and corresponds to each muscle group. The system software is stored in cloud.

Description

Personalized testing of muscle performance

TECHNICAL FIELD

A computer-implemented method of personalized testing of a mammal’s muscle performance with the use of artificial intelligence models and predictive models, by using as input real time biomechanical data obtainable from well-known wearable electronic devices. A trained system software artificial intelligence models is used.

BACKROUND OF THE INVENTION

In the recent years the use of wearable electronic devices which monitor and analyze a mammal’s performance are used for the purpose of calculating muscle performance. There are various methods that provide statistical analysis of a mammal’s performance based on various sensor data, deriving from wearables. Nevertheless, the statistical analysis is not personalized for each mammal, cannot make a distinction between the users and furthermore cannot calculate muscle load on each muscle.

US2019046086A1 describes wearable monitoring devices and a method for generating feedback on the performance of the activity by the user based on the comparison. Nevertheless, the invention describes a professional athlete as a benchmark of comparison. The same training cycle may improve user while harming another one. There is a need for a personalized testing of mammal’s performance.

US10314511 B2 describes acquiring image data from a magnetic resonance imaging (MRI) system and determining from the image data if the muscle deviates when compared with a healthy muscle. The healthy muscle is represented by a composite based on individual healthy muscles from a plurality of healthy living subjects in a sample population. There is a need for a method for testing of mammal’s performance, which provides real time results without the use of an expensive laboratory equipment.

In view of the above state of art, the present inventors have come up with a real-time reproducible and accurate personalized testing of a mammal’s muscle performance by avoiding associated high costs and use of laboratory equipment. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 . Flowchart of the training of the artificial neural network of the invention.

FIG. 2. Flowchart of using the artificial neural network of the invention to train a predictive model.

FIG. 3. Flowchart of using the predictive model of the invention with real time biomechanical data.

FIG. 4. Algorithmic flowchart of the invention.

FIG. 5. Flowchart of the connections of the invention.

FIG. 6. MRI of a tendon abnormality in the hamstring of the right leg.

FIG. 7. Processed MRI model for region of injury and plane detail for healthy, initially injured and healed muscle data.

FIG. 8. Stress strain z section plane right bicep femoris for healthy, initially injured and healed muscle data.

FIG. 9. Graphical representation of max stress z on plane for healthy, initially injured and healed muscle data.

FIG. 10. Graphical representation of linear acceleration across the 3 axes vs days of training.

FIG. 11 . A box and whisker plot showing 3 axes acceleration vs injury windows.

FIG. 12. Graphical representation of aggregated time interval during which the player performs at the highest (personalized) zone 6.

FIG. 13. A box and whisker plot showing aggregated acceleration vs injury windows.

DETAILED DESCRIPTION OF THE INVENTION

The present invention relates to a computer implemented method for testing muscle performance as set forth in claims. The present invention relates to a computer implemented method for testing muscle performance in a mammal. Preferably a mammal is selected from a human, a dog and a horse. More preferably the mammal is an athlete. The method is personalized, and the results are customized.

The present invention relates to use of artificial network models and predictive model to corelate the real-time biomechanical data with specialized personalized data obtained by laboratory use.

The present invention relates to input a set of parameters relating to biomechanical data and ground reaction forces data wherein the biomechanical data comprising at least the linear acceleration, the angular velocity, the speed and the instantaneous acceleration impulse of the mammal.

The present invention relates to input a set of parameters relating to biomechanical data and magnetic resonance imaging data. Preferably stored in a cloud.

The present invention further relates to training of artificial network models and predictive model to corelate the real-time biomechanical data with specialized data obtained by laboratory use.

The present invention further relates to all data being stored in a cloud storage.

It is within the scope of this invention, a personalized testing of a mammal’s muscle performance without the use of magnetic resonance imaging data. When magnetic resonance imaging is contraindicated, usually in the presence of internal metallic objects, the system software can be fully operable with reposited magnetic resonance imaging data and/or can be based solely on the biomechanical data of the mammal.

An embodiment of said invention is a computer-implemented method of personalized testing of a mammal’s muscle performance, the method comprising: i) input a set of personalized biomechanical data wherein the biomechanical data comprising at least the linear acceleration, the angular velocity, the speed and the instantaneous acceleration impulse of the mammal into a trained artificial neural network model by using as input biomechanical data, and ground reaction forces and output calculated ground reaction forces ii) optionally, repeat step i) iii) process with proprietary algorithms the personalized features related to at least the calculated ground reaction forces, the personalized biomechanical data and combinations thereof, on the three-dimensional space axis and the pertinent shear planes of each axis and identify and store the distribution of the optimal fit per feature, respectively iv) demarcate a distribution of the optimal fit per feature into more than one equal segments, wherein each segment defines a mammal’s personalized muscle performing zone v) train a predictive model by using as input the processed personalized features of step iv) vi) input real-time biomechanical data into the predictive model of step v) to obtain real-time personalized muscle performance by machine learning scoring method.

The term “muscle performance” typically refers to the capacity of the muscle to do work. The muscle performance is related to strength, power and endurance. There are multiple factors which can affect a mammal’s muscles performance such as age, biological age, muscle injuries and many more. Injuries from straining a muscle are common and can minimize the mammal’s quality of life. Calculating the minimum and maximum load per muscle unit, is necessary for testing a mammal’s muscle performance. Testing a mammal’s muscle performance comprises a stress-strain curve which is a graphical way to show the reaction of a material, such as a muscle, when a load is applied, to evaluate the current state and predict the future state of each muscle. The stress-strain curve corresponds to load per unit vs elongation. Within the meaning of the invention the muscle stress, corresponds to the maximum force per muscle unit, when muscles start failing.

The term “biomechanical data” within the scope of this invention relates to at least the linear acceleration, the angular velocity, the speed and the instantaneous acceleration impulse of the mammal. The “biomechanical data” is data which are collected from a wearable electronic device designed to be worn on the mammal’s body, such as vests or cameras. Preferably, the wearable electronic device can be a vest, designed to enclose at most the upper torso of a mammal. Said wearable electronic device is non-restraining to the normal routine of the wearer. The wearable electronic device such as vest, comprise communication means for transmission of the biomechanical data to the console/platform which can be stored in a cloud server. The data are directed through an application programming interface (API) to the invention’s cloud infrastructure for further processing.

The invention’s cloud infrastructure is a cross platform software and/or a web application. The user can find stored data, history training data of each mammal, extract graphical representations which can assist in decision making upon the mammal’s muscle performance.

The term “magnetic resonance imaging data” typically refers to a form of pictures of the anatomy, the physiological processes of the body and especially the muscles and their connections to the bones. The magnetic resonance imaging can detect abnormalities such as abnormal muscle volume, abnormal muscle signal, mass lesion, and abnormal anatomy and is a diagnostic technique which is non-invasive and does not involve exposing the patient under study to potentially harmful radiation. The use of magnetic resonance imaging (MRI) apparatus is needed, but it is only available indoors. The abnormalities cannot be detected by a clinical examination and represent areas with asymptotic pathologies.

Like biomechanical data, the magnetic resonance imaging data is uploaded by the user to a specified remote repository, embedded to the application of the present invention.

It is within the scope of this invention to predicting how the muscle reacts to real-world forces by using the muscle structural analysis. The healthy muscle is represented by a reverse engineering procedure by taking all micro pathologies as healthy tissue areas. The abnormalities are visualized as white spots in the MRI.

Each muscle is analyzed into a large number (thousands to hundreds of thousands) of finite elements. By solving mathematical equations to train the artificial intelligence model the behavior of each element is predicted. By adding up all individual behaviors of each element to calculate the load per muscle unit hence the behavior of each muscle. Preferably, the muscle structural analysis is obtained by processing magnetic resonance imaging data of the mammal into 3-dimensional muscle models using a slicer software, import the 3-dimensional muscle models into pre-processing computer aided engineering software to obtain a muscle’s mesh and import the resulting muscle mesh into a finite element analysis software and using optionally, stored experimental data of the material properties of muscle to final obtain a complete muscle structural analysis. Preferably, the system software of the two artificial intelligence models is used to input personalized biomechanical data, preferably derived from vests or camera and to output calculated load per muscle unit.

The magnetic resonance imaging data for the mammal’s muscular system software is processed by a software to create a 3D geometry of a muscle. Those geometries are stored in the cloud. Followed by processing of the stored 3D geometry of a muscle by another software to obtain the geometry’s computational mesh being stored optionally, in a separate file. Furthermore, the geometry’s computational mesh of the muscle is processed by another software to obtain the structural analysis using the mammal’s muscle viscoelastic behavior. Optionally, the mammal’s muscle viscoelastic behavior is determined, by stored experimental data of the material properties of muscle. Preferably, the output of the calculated forces by the first artificial network combined with the structural analysis is used for training a second artificial neural network model. More preferably, the system software of the two artificial neural network models is used to input personalized biomechanical data, preferably derived from vests or camera and to output calculated load per muscle unit. The calculated load per muscle unit is necessary for testing muscle performance and the results are personalized for the mammal.

The outputs are 3D models of the muscle and visualized graphical plots of stress - strain curves. In this way it is viable to assess a part of the human body, which is under the specific strain for a specific time period, in a personalized manner.

The term “application database” typically refers to a database that is controlled and accessed by a single application, wherein the single application is the invention’s application. The term Artificial Intelligence (Al) model, preferably an artificial neural network (ANN) comprises of an input layer, a number of hidden layers and an output layer; each layer further comprises of a different number of artificial neurons and each artificial neuron has at least one input, within the scope of this invention.

The term “training” of an artificial neural network model comprises each artificial neuron has at least one input with an associated weight and bias which is adjusted while this artificial neural network is trained by using the observed data as input and through backpropagation changing the weights and bias using real time measured data input. The first step, in training, involves input data into a artificial intelligence system software and evaluate their accuracy. The second step, validating, evaluates how well the trained a artificial intelligence model performs on previously unseen data. Finally, testing is done to find out if the final artificial intelligence model makes accurate predictions with new data that it has never seen before.

A system software comprising trained a artificial intelligence model is used within the scope of this invention, wherein training involves input biomechanical data, selected from at least the linear acceleration, the angular velocity, the speed and the instantaneous acceleration impulse of the mammal and measured in force platforms ground reaction forces. The artificial intelligence model is evaluated and tested for the accuracy of the calculated ground reaction forces. The system software can further input muscle structural analysis and calculated ground forces to output the load per muscle unit. The artificial intelligence model can be independently evaluated and tested for accuracy. The trained artificial intelligence model is stored in a computing repository.

It is within the scope of this invention, a system software for use in a computer- implemented method of personalized testing of a mammal’s muscle performance of the invention according to the claims, obtainable by training of artificial intelligence model using as input measured biomechanical data, ground reaction forces and optionally, a muscle structural analysis and output calculated ground reaction forces and calculate optionally, the load per muscle unit. Preferably, a system software for use in a computer- implemented method of personalized testing of a mammal’s muscle performance of the invention according to the claims, obtainable by training of artificial intelligence model using as input measured biomechanical data, ground reaction forces and a muscle structural analysis and output calculated ground reaction forces and calculate the load per muscle unit.

It is within the scope of this invention, a artificial neural network model for use in a computer-implemented method of personalized testing of a mammal’s muscle performance of the invention according to the claims, obtainable by training using as input measured biomechanical data, ground reaction forces and output calculated ground reaction forces. It is within the scope of this invention, an artificial intelligence model for use in a computer-implemented method of personalized testing of a mammal’s muscle performance of the invention according to any of the claims, obtainable by training using as input a muscle structural analysis and calculated ground reaction forces to output the load per muscle unit.

Within the meaning of this invention processing with proprietary algorithms a series of the personalized features related to at least the ground reaction forces, biomechanical data, load per muscle unit and combinations thereof, on the three-dimensional space axis and the pertinent shear planes of each axis and identify and store the distribution of the optimal fit per feature, respectively. The optimal fit can be based on the minimum Kolmogorov-Smirnov test statistic, which is a non-parametric test statistic that quantifies the distance between 2 continuous distributions (within the scope of this invention, the empirical distribution function of the sample and the cumulative distribution function of each reference distribution). The calculated measurement of at least the ground reaction forces, biomechanical data, load per muscle unit and combinations thereof is fitted with numerous different distributions, to find the one with the optimum fit. Preferably, the processing of the personalized features with proprietary algorithms and their demarcation into the optimal fit are executed and stored on the invention’s cloud-based servers.

The term “personalized performance zones” per feature typically refers to the result of demarcating a distribution of the optimal fit per feature into more than one segments. The optimum distribution’s inverse Cumulative Distribution Function (CDF) are calculated for every feature. Preferably, the segments are six. Each feature is demarcated in 6 equal segments, to reduce the bias of the results. When the mammal’s muscle performance is in a high personalized performance (upper) zone (i.e. 5, 6) the user to make effective and adequate decisions. The invention’s personalized performing zones calculation does not demarcate each feature in fixed (constant) zones. When demarcating each feature in fixed (constant) zones, especially, for lower muscle performance, the workload is disproportionately distributed in the lower zones which leads to the erroneous conclusion that mammal’s muscle under-performs.

FIG. 1 is a system diagram illustrating an exemplary operating environment for the various embodiments disclosed herein. Embodiments may be implemented on a commercial MRI system and biomechanical data obtained by commercial vest and/or cameras. FIG. 1 illustrates training an artificial intelligence model, preferable an artificial neutral network, using as input biomechanical data comprising at least the linear acceleration, the angular velocity, the speed and the instantaneous acceleration impulse of the mammal obtained by commercial vests and/or camera and ground reaction forces obtainable by a force plats. Ground reactions forces can be measured once of. The output of the artificial intelligence model are the calculated ground reaction forces. Preferably the artificial intelligence model is an artificial neutral network model. The ground reaction forces obtained in lab environment are used to train the artificial neutral network model and are obtained for this purpose. The calculated ground reaction forces are further used to the training of another artificial intelligence model. Said artificial intelligence model uses as input processed magnetic resonance imaging data and the calculated ground forces obtained by the first artificial intelligence model. The magnetic resonance imaging data are processed by a computer-aided design and drafting/simulation software. The derived data represent the mammal’s structural analysis and the muscles 3D model, which are used for training the artificial intelligence model. Biomechanical data are again fed into the system software comprising the trained Al models, to calculate the load per muscle unit as illustrated in FIG. 2. The obtained load per muscle, the calculated ground reaction forces and the personalized biomechanical data are being processed with proprietary algorithms into the optimal fit. The demarcation of the optimal fit into segments, is followed. Furthermore, the personized muscle performance zone is obtained as shown in FIG.2. The personalized performing zone and personalized features can be graphically displayed. The personalized performing zones include and are graphically illustrated, with reference to the biomechanical data and the ground reaction forces. In an embodiment furthermore the load per muscle unit is graphically displayed. As per the embodiment of FIG.2 the predictive model is trained using at least once, as input the personized muscle performing zone, enclosing all stored data regarding to ground reaction forces and biomechanical data, preferably more than once, more preferably at least 5 consequent set of data or days of training. In the embodiment of FIG. 3 real-time biomechanical data are used as input in the predictive model to calculate real time personalized muscle performance. The application software makes the information accessible to the end user.

FIG. 4 shows the algorithmic flowchart of an embodiment of the invention, wherein data from force plates (3) and wearable tracking devices (2) are used to train the Al models. Further data from wearable tracking devices (2) are processed and used as input in the trained Al models. The outcome corresponds to the workload metrics which are the calculated ground forces and to a personalized database of the mammal. The magnetic resonance imaging data (4) can be processed by a serious of software comprising slicer software, a pre-processing computer aided engineering software finite analysis software and software system for biomechanical modeling, simulation and analysis, using the obtained workload metrics to obtain the myoskeletal metrics from the muscle structural analysis of each mammal. Both metrics/data are stored and processed into the invention’s cloud infrastructure being available to the end user by various devices (5).

The interconnectivity of an embodiment of the invention is illustrated in FIG. 5. As shown in FIG.5 biomechanical data are received from conventional wearable tracking devices (2) through third party console/platform (6) and are stored to a cloud server. An application programming interface translates the data from the conventional applications of the wearable devices (2) into data suitable for the embodiment’s application. The biomechanical raw data, the ground reaction forces raw data and the magnetic resonance imaging raw data are processed according to the invention in the invention’s cloud infrastructure using proprietary algorithm’s processing. The output calculated data are stored in the invention’s cloud infrastructure and are available to the end user through cross platform or web software for various devices (5).

EXAMPLES EXAMPLE 1

Training an artificial neural network model

The force plates were embedded in the laboratory floor. The mammals participating wore a vest consisting of GPS sensors and a gyroscope, while performing a series of exercises. These sensors recorded a multi-dimensional input which corresponds to the linear acceleration on the three axes of motion (x,y,z), gyroscope measures one for each axis (x,y,z), the speed and the instantaneous acceleration impulse. The output consists of the three-dimensional force components on the three spatial axes. Meanwhile the mammals were performing the exercises on the force plates that were measuring the ground reaction forces.

Measurements from the GPS sensors were used as input data to the artificial neural network and the measured ground reaction forces. The calculated ground reaction forces consisted of the model’s outputs. Linear interpolation was applied on the data recorded by the GPS sensors for the input data to match the 10-fold lower sampling frequency of the force plate output data.

Experimental protocol

1 healthy professional athlete voluntarily participated in this research. The athlete were instructed to perform a series of exercises. Those exercises derive from 10 simple movements, which were selected from the NSCA Basics of Strength and Conditioning Manual (Sands, A. W., Wurth, J. J., & Hewit, K. J. (2012). The National Strength and Conditioning Association’s (NSCA) Basics Of Strength And Conditioning Manual), based on the assumption that all complex movements are the algebraic summary of those simple ones. To ensure consistency of the measurements, when jumps were included, the exercise was performed on the center of the force plate.

The exercises were the following:

Exercise 1 : The human, while standing on one foot, pushed-off at maximal capacity in order to achieve a maximal-height vertical jump. He landed on the same foot, took a step forward onto the second force plate with the other foot and then immediately maneuvered to the other side. The exercise was executed two times, once for each leg.

Exercise 2: The human, while standing on both feet, pushed-off at maximal capacity in order to achieve a maximal-height vertical jump. He landed on both feet and then took a step forward on the second force plate with one foot and immediately maneuvered to the other side direction. The exercise was executed two times for each direction.

Exercise 3: The human performed a rapid acceleration up to the center of the second force plate, where they abruptly made a sidestep cutting maneuver with one leg stopping followed by sidestepping. The exercise was executed twice for each side.

Exercise 4: The human performed maximal acceleration efforts over the corridor of X m length where the force plates were embedded.

Exercise 5: The human performed maximal acceleration efforts until they reached the second force plate, where they abruptly tried to decelerate as fast as possible.

The artificial neural network used consist of computational units that are interconnected to each other and distributed in layers. A feedforward network with backpropagation was used. The connections between units of continuous layers are weights. Input data is presented to the units of the first layer. From this layer the data is forwarded to the units in the hidden layer while the data is multiplied by the weight factor. In each receiving unit, the weighted data of all incoming connections is summed, and a bias term is added. Then the summed input plus the bias term is processed by an activation function. This is the output value and is then forwarded to the units of the next layer. By adjusting the weights, the network can be trained. It can find relationships between input and output patterns, which in this study is the relationship between GPS sensor data and ground reaction forces. To train a neural network we provide data from both ground reaction forces measurements and GPS sensors, for several conditions.

After the training is complete, the artificial neural network can use one specific pattern. The artificial neural network can generalize 'knowledge' obtained during training for a selected set of situations, to new situations for which it has not been trained. EXAMPLE 2

Testing Muscle performance

Following the teaching of this invention, we observe a tendon abnormality in the hamstring of the right leg, in the region of the ischial curvature of right thigh. FIG.6 is the modeling of the injury to the right bicep femoris muscle origin, as detected by the MRI. The right is the MRI with the initial injury and the left is after six weeks with the muscle under healing procedure. We compare the mechanical properties of the muscle under the above conditions with the full healthy model, using the trained artificial neural network and predictive models.

Region of injury and plane detail

By comparing of healthy and two conditions of injured model, as calculated by the invention’s method (initial injury on the mid and healed on the bottom) of FIG.7 the diagrams of FIG. 8 and FIG 9 show that:

In this comparison, the plane of injury the stresses along z axes are significant higher on the Initial injured model in comparison with the healthy one. Furthermore, on healed model the total stresses on that plane have decreased, as expected.

Also, there is a significant stress concentration in the injured area as, the same geometric point has a stress value of 0.28383 MPa in the healthy model, while on the model with initial injury it has a value of 0.64691 MPa and on the healed model has a value of 0,33162 MPa.

Locally, at the specific point on the plane we have a better view about the athlete’s condition as we notice a significant increase of 128% of the stresses between the healthy model and the one with the initial injury. Nevertheless, there is a decrease of 48,73% of stresses on the same point between the model with the initial injury and the healed one. Finally, the healed model has only an increase of 16,83% of stresses in comparison with the healthy one. In overall, this increase of stress along with the fact that this injury is near to the origin of the muscle could led us to estimate that the initial injury is highly possible to affect the athlete's performance.

Also, the healed model is close to the healthy one so the athlete possible return to action soon.

EXAMPLE 3

Testing Muscle performance by Performance zone

The acute muscle injuries of the lower limb (incl. the area of thigh, knee, lower leg etc), excluding any contact/collision ones, was studied in human subject, preferably a professional athlete. The exerted forces (linear and angular acceleration) and demarcated to personalized zones, while an optimized mathematical equation is used, were calculated. Also, leveraging the personalized zones, the time interval each player performs in the upper zone 6 is calculated, giving extra meaningful insights. The subject was evaluated for 20 days. In FIG. 10 linear and angular acceleration across the 3 axes acceleration, goes out of the calculated upper limit (1 .3) on 2022-08-17 (1 day prior to the injury incident). The same is shown in FIG. 11 , which is a box and whisker plot. A box and whisker plot is defined as a graphical method of displaying variation in a set of data, which provides additional detail than a histogram. In addition, it allows multiple sets of data to be displayed in the same graph. In our case, we evaluate each metric for different time windows. FIG. 11 , shows that the latest 5 days prior to injury (1 ^st bar), the distribution shows high values, close to the upper limit of 1 .3. Particularly, 50% and 75% of the values are above 1 .3 and 1 .4, respectively - a clear indication that an injury may happen. Then, it is the aggregated time interval during which the mammal performs at the highest (personalized) zone 6. In particular, the mammal shows a remarkable maximum duration right at the day before the injury happened as in FIG.12.

Finally, we evaluate the aggregated (net) accelerations (sum of X, Y and Z). FIG. 13 shows that during the latest 5 days prior to injury (1^st bar), the aggregated (net) acceleration’s distribution shows high values, where 75% of the values (Q3) significantly exceed the respective one of the last-1 Odays-prior-window (3^rd bar); a clear indication that an injury may happen.

Claims

1. A computer-implemented method of personalized testing of a mammal’s muscle performance, the method comprising: i) input personalized features selected from biomechanical data wherein the biomechanical data comprising at least the linear acceleration, the angular velocity, the speed and the instantaneous acceleration impulse of the mammal into a system software comprising a trained artificial intelligence model stored in a computing repository which calculates the ground reaction forces and calculate the load per muscle unit ; ii) optionally, repeat step i); iii) process with proprietary algorithms personalized features selected from at least the ground reaction forces, the biomechanical data, the load per muscle unit and the combinations thereof, on the three-dimensional space axis and the pertinent shear planes of each axis and identify and store the distribution of the optimal fit per feature, respectively; iv) demarcate a distribution of the optimal fit per personalized feature into more than one segments, wherein each segment is a mammal’s personalized muscle performing zone; v) train a predictive model by using as input the mammal’s personalized muscle performing zone of each feature from step iv) vi) optionally, graphically simulate the muscle and muscle loads; vii) input real-time biomechanical data into the predictive model of step v) to obtain real-time personalized muscle performance features by machine learning scoring method;

2. A computer-implemented method personalized testing of a mammal’s muscle performance according to proceeding claims, wherein the segments defining a mammal’s personalized performing zones per feature are six.

3. A computer-implemented method personalized testing of a mammal’s muscle performance according to claim 1 , wherein the biomechanical data are obtained by a wearable sensor and/or a camera.

4. A computer-implemented method personalized testing of a mammal’s muscle performance according to claim 3, which automatically receives the data collected from the wearable sensor and/or a camera either through an external application programming interface or by a custom procedure it incorporates and stores it to a cloud storage.

5. A computer-implemented method personalized testing of a mammal’s muscle performance according to claim 1 , wherein the proprietary algorithms are executed on a cloud-based server.

6. A computer-implemented method of personalized testing of a mammal’s muscle performance, according to claim 1 , wherein a trained artificial intelligence model is at least one trained artificial neural network model.

7. A computer-implemented method of personalized testing of a mammal’s muscle performance, according to claim 6, comprising a trained artificial neural network model obtainable by input biomechanical data, ground reaction forces and output calculated ground reaction forces;

8. A computer-implemented method of personalized testing of a mammal’s muscle performance, according to claim 6, wherein the trained artificial intelligence model is obtainable by input biomechanical data, ground reaction forces and a muscle structural analysis and output calculated ground reaction forces and load per muscle unit;

9. A computer-implemented method according to claim 8, wherein the muscle structural analysis is obtained by processing magnetic resonance imaging data of the mammal into 3-dimensional muscle models using a slicer software, import the 3-dimensional muscle models into pre-processing computer aided engineering software to obtain a muscle’s mesh and import the obtained muscle mesh into a finite element analysis software with optionally, stored experimental data of the material properties of muscle.

10. A computer-implemented method of personalized testing of a mammal’s muscle performance according to claim 8, wherein training of an artificial neural network model is through backpropagation

11. A computer-implemented method personalized testing of a mammal’s muscle performance according to proceeding claims, wherein the mammal is selected from a human, a dog and a horse.

12. A computer-implemented method of personalized testing of a mammal’s muscle performance according to processing claims, characterized in being a cross platform software and/or a web application.

13. A system software for use in a computer-implemented method of personalized testing of a mammal’s muscle performance according to proceeding claims, obtainable by training of artificial intelligence model using as input biomechanical data, ground reaction forces and optionally, a muscle structural analysis and output calculated ground reaction forces and load per muscle unit.

14. A system software for use in a computer-implemented method of personalized testing of a mammal’s muscle performance according to claim 14, obtainable by training of artificial intelligence model using as input biomechanical data selected from at least the linear acceleration, the angular velocity, the speed and the instantaneous acceleration impulse of the mammal.

15. A system software for use in a computer-implemented method of personalized testing of a mammal’s muscle performance according to claim 13-14, obtainable by training of artificial intelligence model using as input ground reaction forces measured on force platforms.