Title:Volumetric parcellation of the right ventricle for regional geometric and functional assessment

Abstract:In clinical practice, assessment of right ventricle (RV) is primarily done through its global volume, given it is a standardised measurement, and has a good reproducibility in 3D modalities such as MRI and 3D echocardiography. However, many illness produce regionalchanges and therefore a local analysis could provide a better tool for understanding and diagnosis of illnesses. Current regional clinical measurements are 2D linear dimensions, and suffer of low reproducibility due to the difficulty to identify landmarks in the RV, specially in echocardiographic images due to its noise and artefacts. We proposed an automatic method for parcellating the RV cavity and compute regional volumes and ejection fractions in three regions: apex, inlet and outflow. We tested the reproducibility in 3D echocardiographic images. We also present a synthetic mesh-deformation method to generate a groundtruth dataset for validating the ability of the method to capture different types of remodelling. Results showed an acceptable intra-observer reproduciblity (<10%) but a higher inter-observer(>10%). The synthetic dataset allowed to identify that the method was capable of assessing global dilatations, and local dilatations in the circumferential direction but not longitudinal elongations