EP-4738377-A1 - PERSONALIZED STRUCTURAL AND FUNCTIONAL CARDIAC DIGITAL TWIN FOR ASSESSMENT OF CARDIOPULMONARY ENDURANCE OF ATHLETE

Abstract

A method and system that builds a regression model from a personalized structural and functional Cardiac Digital Twin (CDT) for assessment of cardiopulmonary endurance of an athlete is disclosed. The personalized Cardiac Digital Twin (CDT), which replicates echo like functionality under dynamic conditions integrates subject specific kinematics data real time acquired to run personalized CDT and generate intrinsic metrices to evaluate performance in different phases of exercise or endurance activity. Most of existing works are focused on computing mere metrices for entire activity as whole. However, without judicial combination of these metrices obtained in different phases, no meaningful inference can be drawn on performance evaluation.

Inventors

• MAZUMDER, OISHEE
• SINHA, ANIRUDDHA
• MUKHERJEE, Ayan
• CHANDEL, Vivek
• BHATTACHARYA, SAKYAJIT
• Ahmed, Nasimuddin
• Khandelwal, Sundeep
• GHOSE, Avik

Assignees

• Tata Consultancy Services Limited

Dates

Publication Date

20260506

Application Date

20251021

Claims (15)

1. A processor implemented method (200), the method comprising: time synchronizing (202), via one or more hardware processors, sensor data acquired during an endurance activity performed by each subject among a plurality of subjects, wherein the sensor data represents a plurality of data types comprising electrocardiogram (ECG) data, accelerometer data providing speed, Gravity data, Global Positioning System (GPS) data and Barometer data acquired from a plurality of sensors worn by each subject, and wherein the plurality of subjects are a mix of a professional athlete, a mid-level athlete and an amateur athlete; segmenting (204), via the one or more hardware processors, each of the plurality of data types into a plurality of segments, wherein a first step comprises segmenting by identification of i) an initial resting or warmup segment, ii) an Intense Activity (IA) segment and iii) a recovery segment post the IA segment based on Heart Rate (HR) variation and associated Metabolic Equivalent Task (MET); and wherein a second step comprises segmenting the IA segment into i) an initial ramp-up, ii) a cruise, iii) an occasional dip in speed, and iv) an occasional increase in speed and/or heartrate; running (206), via the one or more hardware processors, a personalized Cardiac Digital Twin (CDT) model, built for each subject, on corresponding segments of each of the plurality data types to extract a plurality of sets of cardiopulmonary dynamics, wherein a plurality of sets of cardiopulmonary features are derived from the plurality of sets of cardiopulmonary dynamics for each subject, and wherein a distribution of the sets of cardiopulmonary features is processed via a feature transformation technique to obtain a transformed cardiopulmonary feature vector for each subject; extracting (208), via the one or more hardware processors, a set of kinematic features from one or more of the of the plurality of data types acquired during the endurance activities, wherein a distribution of the set of kinematic features is processed via a feature transformation technique to obtain a transformed kinematic feature vector for each subject; generating (210), via the one or more hardware processors, an annotated feature matrix comprising a plurality of features vectors representing the plurality of subjects via the transformed kinematic feature vector concatenated with the transformed cardiopulmonary feature vector, wherein each feature vector among a plurality of feature vectors of the annotated feature matrix is annotated with a proficiency score of each of the subject for the endurance activity; and creating (212), via the one or more hardware processors, a plurality of trained data regression models using the annotated feature matrix for predicting the proficiency score for the professional athlete, the mid-level athlete and the amateur athlete.
2. The processor implemented method as claimed in claim 1, wherein during inference personalized guidance and training plan for future runs of a test subject is generated based on the predicted proficiency score, a personalized CDT of the subject, and a set of kinematic and cardiopulmonary features extracted for the test subject.
3. The processor implemented method as claimed in claim 2, wherein the personalized guidance and training plan generation comprises determining a difference of the kinematic features and cardiopulmonary features for the test subject from the professional athlete and the mid-level athlete depending upon the predicted proficiency score of the test subject to identify a plurality of measures to be focused upon for improvement with reference the a mid-level athlete later progressing towards the professional athlete or an amateur athlete progressing towards mid-level.
4. The processor implemented method as claimed in claim 1, wherein the personalized CDT model is built using i) a plurality of cardiac structural parameters obtained from MRI and Echo test of each subject, ii) a plurality of subject-specific baseline clinical parameters, and iii) body physique and heart associated metadata of each subject.
5. The processor implemented method as claimed in claim 1, wherein the set of cardiopulmonary features comprise metabolic equivalent of task (MET), (av) arteriovenous, (per) perfusion, pva: pressure volume area, mep: mean power, VE: ventricular efficiency, Heart rate (HR), Energy ejected (EE), Stroke work (SW), mep: mean power, VE: ventricular efficiency, ESP end systolic pressure, EDV: end diastolic volume, ESPVR: end systolic pressure volume ratio, EDPVR: end diastolic pressure volume ratio, Mean power (Pmean), Cardiac output (CO), Stroke volume (SV), Ejection Fraction (EF), and Mean arterial pressure (MAP).
6. The processor implemented method as claimed in claim 1, wherein the set of kinematics features comprise Work Intensity (WI), Running VO2 (VO2run), Average running efficiency (REavg), REconomy, Average heart rate (HRavg), Average breathing rate (BRavg), Session time (ST), Calorie (Cal), maximum speed(Smax), Average Cadence(Cadavg), Average MET (METavg) , and Total distance (TD).
7. A system (100) comprising: a memory (102) storing instructions; one or more Input/Output (I/O) interfaces (106); and one or more hardware processors (104) coupled to the memory (102) via the one or more I/O interfaces (106), wherein the one or more hardware processors (104) are configured by the instructions to: time synchronize sensor data acquired during an endurance activity performed by each subject among a plurality of subjects, wherein the sensor data represents a plurality of data types comprising electrocardiogram (ECG) data, accelerometer data providing speed, Gravity data, Global Positioning System (GPS) data and Barometer data acquired from a plurality of sensors worn by each subject, and wherein the plurality of subjects are a mix of a professional athlete, a mid-level athlete and an amateur athlete; segment each of the plurality of data types into a plurality of segments, wherein a first step comprises segmenting by identification of i) an initial resting or warmup segment, ii) an Intense Activity (IA) segment and iii) a recovery segment post the IA segment based on Heart Rate (HR) variation and associated Metabolic Equivalent Task (MET); and wherein a second step comprises segmenting the IA segment into i) an initial ramp-up, ii) a cruise, iii) an occasional dip in speed, and iv) an occasional increase in speed and/or heartrate; run a personalized Cardiac Digital Twin (CDT) model, built for each subject, on corresponding segments of each of the plurality data types to extract a plurality of sets of cardiopulmonary dynamics, wherein a plurality of sets of cardiopulmonary features are derived from the plurality of sets of cardiopulmonary dynamics for each subject, and wherein a distribution of the sets of cardiopulmonary features is processed via a feature transformation technique to obtain a transformed cardiopulmonary feature vector for each subject; extract a set of kinematic features from one or more of the of the plurality of data types acquired during the endurance activities, wherein a distribution of the set of kinematic features is processed via a feature transformation technique to obtain a transformed kinematic feature vector for each subject; generate an annotated feature matrix comprising a plurality of features vectors representing the plurality of subjects via the transformed kinematic feature vector concatenated with the transformed cardiopulmonary feature vector, wherein each feature vector among a plurality of feature vectors of the annotated feature matrix is annotated with a proficiency score of each of the subject for the endurance activity; and create trained data regression models using the annotated feature matrix for predicting the proficiency score for the professional athlete, the mid-level athlete and the amateur athlete.
8. The system as claimed in claim 7, wherein during inference personalized guidance and training plan for future runs of a test subject is generated based on the predicted proficiency score, a personalized CDT of the subject, and a set of kinematic and cardiopulmonary features extracted for the test subject.
9. The system as claimed in claim 8, wherein the personalized guidance and training plan generation comprises determining a difference of the kinematic features and cardiopulmonary features for the test subject from the professional athlete and the mid-level athlete depending upon the predicted proficiency score of the test subject to identify a plurality of measures to be focused upon for improvement with reference the a mid-level athlete later progressing towards the professional athlete or an amateur athlete progressing towards mid-level.
10. The system as claimed in claim 7, wherein the personalized CDT model is built using i) a plurality of cardiac structural parameters obtained from MRI and Echo test of each subject, ii) a plurality of subject-specific baseline clinical parameters, and iii) body physique and heart associated metadata of each subject.
11. The system as claimed in claim 7, wherein the set of cardiopulmonary features comprise metabolic equivalent of task (MET), (av) arteriovenous, (per) perfusion, pva: pressure volume area, mep: mean power, VE: ventricular efficiency, Heart rate (HR), Energy ejected (EE), Stroke work (SW), mep: mean power, VE: ventricular efficiency, ESP end systolic pressure, EDV: end diastolic volume, ESPVR: end systolic pressure volume ratio, EDPVR: end diastolic pressure volume ratio, Mean power (Pmean), Cardiac output (CO), Stroke volume (SV), Ejection Fraction (EF), and Mean arterial pressure (MAP).
12. The system as claimed in claim 7, wherein the set of kinematics features comprise Work Intensity (WI), Running VO2 (VO2run), Average running efficiency (REavg), REconomy, Average heart rate (HRavg), Average breathing rate (BRavg), Session time (ST), Calorie (Cal), maximum speed (Smax), Average Cadence (Cadavg), Average MET (METavg), and Total distance (TD).
13. One or more non-transitory machine-readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause: time synchronizing sensor data acquired during an endurance activity performed by each subject among a plurality of subjects, wherein the sensor data represents a plurality of data types comprising electrocardiogram (ECG) data, accelerometer data providing speed, Gravity data, Global Positioning System (GPS) data and Barometer data acquired from a plurality of sensors worn by each subject, and wherein the plurality of subjects are a mix of a professional athlete, a mid-level athlete and an amateur athlete; segmenting each of the plurality of data types into a plurality of segments, wherein a first step comprises segmenting by identification of i) an initial resting or warmup segment, ii) an Intense Activity (IA) segment and iii) a recovery segment post the IA segment based on Heart Rate (HR) variation and associated Metabolic Equivalent Task (MET); and wherein a second step comprises segmenting the IA segment into i) an initial ramp-up, ii) a cruise, iii) an occasional dip in speed, and iv) an occasional increase in speed and/or heartrate; running a personalized Cardiac Digital Twin (CDT) model, built for each subject, on corresponding segments of each of the plurality data types to extract a plurality of sets of cardiopulmonary dynamics, wherein a plurality of sets of cardiopulmonary features are derived from the plurality of sets of cardiopulmonary dynamics for each subject, and wherein a distribution of the sets of cardiopulmonary features is processed via a feature transformation technique to obtain a transformed cardiopulmonary feature vector for each subject; extracting a set of kinematic features from one or more of the of the plurality of data types acquired during the endurance activities, wherein a distribution of the set of kinematic features is processed via a feature transformation technique to obtain a transformed kinematic feature vector for each subject; generating an annotated feature matrix comprising a plurality of features vectors representing the plurality of subjects via the transformed kinematic feature vector concatenated with the transformed cardiopulmonary feature vector, wherein each feature vector among a plurality of feature vectors of the annotated feature matrix is annotated with a proficiency score of each of the subject for the endurance activity; and creating a plurality of trained data regression models using the annotated feature matrix for predicting the proficiency score for the professional athlete, the mid-level athlete and the amateur athlete.
14. The one or more non-transitory machine-readable information storage mediums of claim 13, wherein during inference personalized guidance and training plan for future runs of a test subject is generated based on the predicted proficiency score, a personalized CDT of the subject, and a set of kinematic and cardiopulmonary features extracted for the test subject.
15. The one or more non-transitory machine-readable information storage mediums of claim 14, wherein the personalized guidance and training plan generation comprises determining a difference of the kinematic features and cardiopulmonary features for the test subject from the professional athlete and the mid-level athlete depending upon the predicted proficiency score of the test subject to identify a plurality of measures to be focused upon for improvement with reference the a mid-level athlete later progressing towards the professional athlete or an amateur athlete progressing towards mid-level.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS AND PRIORITY The present application claims priority to Indian application no. 202421082908, filed on October 29, 2024. TECHNICAL FIELD The embodiments herein generally relate to the field of machine learning and predictive analytics and, more particularly, to a method and system for personalized structural and functional Cardiac Digital Twin (CDT) for assessment of cardiopulmonary endurance of an athlete. BACKGROUND Health digital twins are essentially digital replicas of human organs, like heart, liver, etc. emulating its functional properties that can be used in for individualized prediction of different treatment outcomes with the goal to virtually select the most promising strategy. Modelling human heart or creating a Cardiac Digital Twin (CDT) of the heart can revolutionize cardiac healthcare in precision medicine and therapy management domain. Such models can also be envisaged for other applications that requires predictive analysis, and high endurance athletic cardiac remodeling is a perfect example where these models can provide groundbreaking insights and discoveries into various parameters effecting the cardiac health and athletic performance. Utilization of CDT has been primarily used in medical domain and its application in athletic training or stress activities for enhanced predictive analytics is open area for research. SUMMARY Embodiments of the present disclosure present technological improvements as solutions to one or more of the above-mentioned technical problems recognized by the inventors in conventional systems. For example, in one embodiment, a method for personalized structural and functional Cardiac Digital Twin (CDT) for assessment of cardiopulmonary endurance of an athlete is provided. The method includes time synchronizing sensor data acquired during an endurance activity performed by each subject among a plurality of subjects, wherein the sensor data represents a plurality of data types comprising ECG data, accelerometer data providing speed, Gravity data, GPS data and Barometer data acquired from a plurality of sensors worn by each subject, and wherein the plurality of subjects are a mix of a professional athlete, a mid-level athlete and an amateur athlete. Further, the method includes segmenting each of the plurality of data types into a plurality of segments. A first step comprises segmenting by identification of i) an initial resting or warmup segment, ii) an Intense Activity (IA) segment and iii) a recovery segment post the IA segment based on Heart Rate (HR) variation and associated Metabolic Equivalent Task (MET). A second step comprises segmenting the IA segment into i) an initial ramp-up, ii) a cruise, iii) an occasional dip in speed, and iv) an occasional increase in speed and/or heartrate. Further, the method includes running a personalized Cardiac Digital Twin (CDT) model, built for each subject, on corresponding segments of each of the plurality data types to extract a plurality of sets of cardiopulmonary dynamics. A plurality of sets of cardiopulmonary features are derived from the plurality of sets of cardiopulmonary dynamics for each subject, and wherein a distribution of the sets of cardiopulmonary features is processed via a feature transformation technique to obtain a transformed cardiopulmonary feature vector for each subject; Furthermore, the method includes extracting a set of kinematic features from one or more of the of the plurality of data types acquired during the endurance activities. A distribution of the set of kinematic features is processed via a feature transformation technique to obtain a transformed kinematic feature vector for each subject. Further, the method includes generating an annotated feature matrix comprising a plurality of features vectors representing the plurality of subjects via the transformed kinematic feature vector concatenated with the transformed cardiopulmonary feature vector, wherein each feature vector among a plurality of feature vectors of the annotated feature matrix is annotated with a proficiency score of each of the subject for the endurance activity. Furthermore, the method includes creating trained data regression models using the annotated feature matrix for predicting the proficiency score for the professional athlete, the mid-level athlete and the amateur athlete. During inference personalized guidance and training plan for future runs of a test subject is generated based on the predicted proficiency score, a personalized CDT of the subject, and a set of kinematic and cardiopulmonary features extracted for the test subject. In another aspect, a system for personalized structural and functional Cardiac Digital Twin (CDT) for assessment of cardiopulmonary endurance of an athlete is provided. The system comprises a memory storing instructions; one or more Input/Output (I/O) interfaces; and one or more hardware processors coupled to the memory via the one or more I/O interfa