Quantitative analysis of regional endocardial geometry dynamics from 4D cardiac CT images: endocardial tracking based on the iterative closest point with an integrated scale estimation.

Regional cardiac function analysis is important for the diagnosis and treatment planning of ischemic heart disease, but has not been sufficiently developed in the field of computed tomography (CT). Therefore, we propose a 3D endocardial tracking framework for cardiac CT using local point cloud registration based on the iterative closest point with an integrated scale estimation algorithm. We also introduce regional function descriptors that express the curvature and stretching of the endocardium: Surface distortion (E) and Scaling rate (S). For a region-to-region comparison, we propose endocardial segmentation according to coronary perfusion territories defined by the Voronoi partition based on coronary distribution. Our study of 65 endocardial segments in ten subjects showed that global endocardial deformation has a positive relationship with the stroke volume index (r = 0.896 and 0.829 in (E) over bar and (S) over bar, respectively) and ejection fraction (r = 0.804 and 0.835), and a positive relationship with the brain natriuretic peptide level (r = 0.690 and 0.776). A positive relationship between segmental E and S (r = 0.845), a higher value of E in ischemic segments (p = 0.021) that are determined by fractional flow reserve estimated from coronary CT data, and a higher value of S in the left circumflex artery territory (p < 0.05) were also observed. The required radiation dose was 5.0 +/- 0.7 mSv and the computation time was 7.2 +/- 1.1 min. The result suggests that proposed endocardial deformation analysis using CT can be conducted on site and in time for the acute setting, and may be useful for the diagnosis of cardiac dysfunction or myocardial ischemia.