A comparative analysis of global shape analysis methods for the assessment of the human right ventricle

Abstract

Two contemporary statistical shape analysis workflows are presented and compared with respect to analysis of the human right ventricle (RV). The methods examined include an approach that directly applies proper orthogonal decomposition to harmonically mapped surfaces (DM-POD) and an approach that expands the harmonically mapped surfaces onto spherical harmonic functions prior to further analysis (SPHARM). The structure of both workflows is elaborated upon and compared, particularly regarding the details of several key sub-steps, including the shape parameterisation, alignment and statistical decomposition. The performance is evaluated for the components of each framework at the various analysis stages, as well as for the output of the complete workflows in terms of the potential to assess right ventricular function through application to a set of RV endocardial surfaces with varying levels of pulmonary hypertension. In addition, DM-POD and SPHARM are examined with respect to different methods of utilising the available phases captured throughout a single cardiac cycle for the patient set. The DM-POD workflow is quantitatively shown to provide more anatomically consistent representations of the RV, while in general, the features produced by the two workflows are shown to be distinctly different. Furthermore, both the workflow components and the components (i.e. phases) of the cardiac cycle utilised for the shape analysis are quantitatively shown to significantly affect the pattern analysis of the patient set.