Multi-Resolution 3D Convolutional Neural Networks for Automatic Coronary Centerline Extraction in Cardiac CT Angiography Scans

Zohaib Salahuddin, Matthias Lenga, Hannes Nickisch

Abstract: We propose a deep learning-based automatic coronary artery tree centerline tracker (AuCoTrack) extending the vessel work of [1]. A multi-resolution 3D Convolutional Neural Network (CNN) is employed to simultaneously predict movement directions and detect bifurcations. Moreover, an automated artery endpoint detector is used to prevent premature termination of the tracking process. On Coronary Computed Tomography Angiography (CCTA or coronary CTA) scans annotated by clinical experts, an average sensitivity of 87.1% and clinically relevant overlap of 89.1% could be obtained. In addition, the MICCAI 2008 Coronary Artery Tracking Challenge (CAT08) training and test datasets were used to benchmark the algorithm and to assess its generalization capabilities. On CAT08, an average overlap of 93.6% and a clinically relevant overlap of 96.4% were achieved.