Abstract
Measuring living heart anatomy using three-dimensional (3D) images remains challenging. To address this, a method called isosurface measurement on volume-rendered images (IMVR) has been developed. This study aims to validate IMVR in quantitative measurement by comparing it with curved planar reformation (CPR). Five 3D-printed human cardiac models created from computed tomography (CT) images were optically scanned, and selected features were measured for reference. The models were CT-scanned, and the datasets were processed for IMVR and CPR measurements. Overall, 157 anatomical features (105 in the aortic root, 52 in the coronary artery) were measured three times by two observers for each method, and the agreement with the reference values was assessed using the Bland–Altman analysis. In the aortic root measurement, the lower and upper 95% limits of agreement (LOAs, mm) for IMVR were (− 3.1, 2.4) and (− 1.3, 0.9), whereas those for CPR were (− 5.9, 5.2) and (− 5.9, 6.3). In the coronary artery measurement, the LOAs for IMVR were (− 2.6, 2.2) and (− 1.2, 0.8), while those for CPR were (− 9.2, 8.6) and (− 9.5, 8.5). For both methods, the intra-class coefficient indicated high intra- and inter-observer reliability. Thus, IMVR demonstrated greater accuracy than CPR and facilitated 3D measurements of complex cardiovascular features.