A Novel Method for Measuring Myocardial Performance Index using Non-contact Ballistocardiogram System

Abstract

Aim: The aim was to validate the Systolic Time Intervals (STI) measured by Ballistocardiography (BCG) with STI derived from simultaneously performed Transthoracic Echocardiogram (TTE) and attempt to create an AI algorithm that automatically calculates Tei Index from BCG tracings. Study design:  Cross-sectional study. Place and Duration of Study: Department of Cardiology and Department of Electrophysiology of Sri Jayadeva Institute of Cardiovascular Sciences & Research, Bangalore, India, between January 2020 and January 2021. Methodology: Two hundred seventy-four patients with clinically indicated TTE were enrolled in the study, average age was 52. Simultaneous recordings on BCG and TTE were done. 150 patients had clinically usable TTE images for accurate calculations. STI was calculated independently by operators experienced in TTE and BCG. Results were compared using Pearson’s R. A proprietary AI algorithm for automatically calculating the MPI, was trained over a subset of patients. Its accuracy in detecting STI was compared to that of TTE and manually calculated STI from BCG. Results: There was a strong positive correlation (r=0.766, P<0.00, 99%CI [0.691,0.824]) between the TTE and BCG derived MPI values. The result was validated over predetermined subgroups of subjects with reduced EF (EF<50) and subjects with normal EF (EF>=50). The AI algorithm had correlation of 0.54(p<0.01) with the MPI calculated by TTE and 0.34(P<0.10) with the manually calculated MPI on the BCG. Conclusion: BCG derived manual and automated MPI correlates well with TTE derived MPI in a variety of EF fraction subgroups. Automated calculation algorithms for MPI derived from BCG remain a work under progress.