Enhanced Extraction of Activation Time and Contractility From Myocardial Strain Data Using Parameter Space Features and Computational Simulations

Abstract

A computational model enables the extraction of two critical myocardial tissue properties: activation time (AT) and contractility (Con) from recorded cardiac strains. However, interference between these parameters reduces the precision and accuracy of the extraction process. This study investigates whether leveraging features in the parameter space can enhance parameter extraction. We utilized a computational model to simulate sarcomere mechanics, creating a parameter space grid of 41 × 41 AT and Con pairs. Each pair generated a simulated strain pattern, and by scanning the grid, we identified cohorts of similar strain patterns for each simulation. These cohorts were represented as binary images—synthetic fingerprints—where the position and shape of each blob indicated extraction uniqueness. We also generated a measurement fingerprint for a strain pattern from a patient with left bundle branch block and compared it to the synthetic fingerprints to calculate a proximity map based on their similarity. This approach allowed us to extract AT and Con using both the measurement fingerprint and the proximity map, corresponding to simple optimization and enhanced parameter extraction methods, respectively. Each synthetic fingerprint consisted of a single connected blob whose size and shape varied characteristically within the parameter space. The AT values extracted from the measurement fingerprint and the proximity map ranged from −59 to 19 ms and from −16 to 14 ms, respectively, while Con values ranged from 48% to 110% and from 85% to 110%, respectively. This study demonstrates that similarity in simulations leads to an asymmetric distribution of parameter values in the parameter space. By using a proximity map, this distortion is considered, significantly improving the accuracy of parameter extraction.