Development of a software algorithm for comprehensive right ventricular strain analysis employing meshes derived from three-dimensional speckle-tracking echocardiography

Abstract

Introduction Although right ventricular (RV) function assessment is important in cardiac surgical patients, dedicated software solutions to measure RV strain on three-dimensional (3D) speckle-tracking echocardiography-based (STE) meshes are lacking. Hence, a novel tool to do so was developed and tested for its feasibility in a clinical setting. Methods RV meshes were generated using dedicated 3D-STE software (4D RV-Function 2.0, Tomtec). A software algorithm was programmed in the Visualization Toolkit (www.vtk.org) to measure systolic peak strains of four regional longitudinal and circumferential contours on the mesh surface (see figure). Longitudinal strains were tracked using fixed-angle pathways (“fixed”) and compared with frame-to-frame angle definition (“variable”). The four circumferential strain layers were defined by strict long-axis segmentation in each frame (“constant”) and compared to “pinning” the planes to points along the anterior and posterior edge (“tilted”). For longitudinal strains, different strategies for the segmentation into basal and apical segments were tested. Global strain values were calculated by averaging the corresponding regional strains. Results 22 patients undergoing mitral valve surgery (MVS) received intraoperative 3D transesophageal echocardiography (TEE) before and after cardiopulmonary bypass. STE yielded 44 meshes for inclusion. 44 percutaneous mitral valve repair (PMVR) patients received 3D-TEE before and after PMVR and 88 meshes were generated. Altogether, 132 total meshes were included. Complete strain analysis was feasible in 126 meshes (95%). Visual validation of each mesh carried out by experienced investigators revealed that “fixed” longitudinal strain calculation resulted in more plausible contour tracking and shortening patterns compared to “variable” calculation, whereas peak longitudinal strain values did not differ (-0.04%, p=0.12). Baso-apical segmentation was best achieved by fixing the segment border in planes defined by neighboring mesh points in the end-diastolic frame. Regarding circumferential strains, “tilted” tracking was visually superior to “constant” segmentation in maintaining functionally matching surface units in the circumferential planes along the heart cycle. “Tilted” vs. “constant” tracking resulted in significantly different global strain values (1.5%, p Discussion RV strain analysis on meshes derived from transesophageal 3D-STE using a custom-made algorithm is feasible. In patients undergoing cardiac surgery, this approach might able be to detect subtle changes of RV function more sensitively than established parameters.