Cardiac computed tomography-derived myocardial mass at risk using the Voronoi-based segmentation algorithm: A histological validation study

Abstract

BackgroundMyocardial mass at risk (MMAR) is an important predictor of adverse cardiac events in patients with ischemic heart disease. This study aims to validate the accuracy of MMAR calculated from cardiac computed tomography (CCT) data using the Voronoi-based segmentation algorithm in comparison with actual MMAR measured on ex-vivo swine hearts prepared by injecting a dye into the coronary arteries. MethodsFifteen extracted swine hearts had India ink injected into one of the major coronary arteries. Subsequently, all coronary arteries manually injected with methylcellulose-based iohexiol-370 were imaged by 16-row CT. The ventricles were cross-sectioned perpendicularly to the long axis of the left ventricle (LV). The stained area and the total LV area of individual slices were measured, and actual MMAR was calculated as the ratio of the LV volume with the disc-summation method. CT-based MMAR of each coronary artery was calculated automatically with the Voronoi-based segmentation algorithm. The results were compared using Pearson's correlation coefficient. ResultsThe median value of CT-based MMAR was 50.8% for the left anterior descending artery (LAD), 36.6% for the left circumflex artery (LCX), and 23.0% for the right coronary artery (RCA). Actual MMAR was 49.8% for LAD, 32.2% for LCX, and 25.9% for RCA. CT-based MMAR was significantly related to actual MMAR (r = 0.92, p = 0.02 for LAD; r = 0.96, p = 0.009 for LCX; r = 0.96, p = 0.009 for RCA). ConclusionCT-based MMAR obtained by Voronoi-based segmentation algorithm reliably estimates actual MMAR measured on ex-vivo swine hearts.