A multi-task deep neural network for segmentation and landmark detection in cardiac computerized tomography

Multimodal bioimaging is increasingly recognized for its potential to integrate multiple types of information. This is particularly relevant in interventional cardiology, where structural imaging may be fused with complementary data, such as metabolic or electrophysiological data. Automating the preprocessing steps required for image alignment and registration is crucial to accelerate procedures in clinical settings. This study explores the feasibility of using a multi-task deep neural network for the automatic segmentation of the left ventricle from cardiac computerized tomography scans and the prediction of a landmark position required for image alignment. The model, based on a 3D UNet architecture, simultaneously segments the left ventricle and localizes its apex. It was trained and tested on the segmented images of the Multi-Modality Whole Heart Segmentation dataset, where the apex position was manually annotated by an expert. The network achieved an average Dice score of 0.91 and an average Euclidean distance of 11 mm for the segmentation and the landmark detection, respectively.