Computational mesh as a descriptor of left ventricular shape for clinical diagnosis

Shape and size of the left ventricle are cardiac biomarkers used in clinical routine practice. They are typically assessed by partial metrics including volume, length, diameter or wall thickness. The aim of this work is to illustrate the potential of an alternative shape analysis methodology based on a comprehensive description of the anatomy using a computational atlas. 40 cardiovascular magnetic resonance scans of young women defined the cohort data set. A stack of 7 to 8 slices from end diastolic frames of dynamic MRI studies were analysed by manual segmentation and automatic personalization of high order computational meshes. The most significant modes of variation of shape of this population were identified by principal component analysis. Statistical significant differences in shape were found in women with higher cardiovascular risk profiles (P<;0.05, Hotelling T2 test). The analysis revealed differences in the position of the apex in the left to right direction, which had not been captured by standard clinical parameters. These results show computational statistical atlases may offer the potential to improve stratification of cardiac diseases.