In order to assess regional diastolic function of the left ventricle (LV), we use LV cine-CT images to build finite element models. To quantitatively evaluate the accuracy of our geometric reconstruction technique used in building these models, we introduce a new measure of shape-similarity, and compare and contrast the results obtained from the new measure to the results obtained from traditional shape-similarity measures. All of these measures are used to compare the endocardial and epicardial LV contours obtained from cine-CT images of canine hearts with those obtained from video images of the same hearts. Our results show that our imaging procedure accurately reproduces shape, and further suggest that the new descriptor has the sensitivity and resolution required to distinguish between images separated by as little as 3 mm.