An Efficient and Robust Method for Automatically Identifying the Left Ventricular Boundary in Cine Magnetic Resonance Images

摘要

Efficient and robust identification of the left ventricular borders remains a challenging problem in cardiology. In this paper, we proposed an automatic method to segment the left ventricles and then identify their borders robustly. The proposed method is named as “ABDC” because it utilizes the strengths of four techniques: Automatic threshold selection; Boundary extraction, Deformation flow tracking, and Convex shape modeling. We compared the proposed method with the PDE optical flow method on 1660 images which are obtained from ten complete short-axis cine MRI datasets (five normal subjects and five patients). As it turned out, the proposed method is more efficient and robust than the benchmark in segmenting LV borders.