Application of global and regional myocardial deformation using cardiovascular magnetic resonance: an assessment of feature tracking in vivo and using numerical simulation

Abstract

Cardiovascular diseases are responsible for approximately a third of all death worldwide, with hypertension being a major risk factor for many of those. Hypertension can lead to left ventricle hypertrophy and diastolic and systolic dysfunction. Myocardial deformation parameters have been shown to have high sensitivity at the early stage of contractile dysfunction. They can be derived from myocardial tagging, considered to be the goldstandard method, or from routinely acquired cine images using feature tracking (FT) techniques. This work aimed to validate FT as a post processing technique. Three FT software packages were used to measure strain parameters in healthy subjects and hypertensive patients in order to assess agreement. Intra- and inter-observer reproducibility was also investigated. The CVI42 software was found to have the best reproducibility. Good agreement across the three software packages and both groups was also observed for circumferential strain calculated from mid-ventricle short axis and longitudinal strain parameters. CVI42 was also compared to the reference tagging analysis by applying both techniques to a healthy and hypertensive patient cohort. Although tagging could discriminate between the two populations (longitudinal strain), no statistically significant differences were found by CVI42. The final validation step was to generate simulation models mimicking simplified cardiac views to compare the experimental results against a true gold-standard for which strain values are known. Two commercial FT software packages were used to analyze the simulated cine images with increasing complexity levels. Both showed inaccurate tracking and high errors compared to analytical values. This indicated that more realistic and complex numerical models should be investigated. Although FT is a relatively new and promising technique, the results demonstrated that it still requires going through standardization to better understand inter-vendor variability.