Automatic Coronary Centerline Extraction Using Gradient Vector Flow Field and Fast Marching Method From CT Images

In current medical imaging, coronary artery stenosis quantification requires fast and accurate coronary centerline computation. This paper develops a new framework for extracting coronary centerlines from 3-D segmented coronary arteries models. The approach utilizes the gradient vector flow (GVF) filed-based speed image of the vessel model and implements a wavefront propagation technique for centerline branch tracking. The approach was validated over 17 3-D synthetic vessel models. The results showed a good agreement between the proposed method and ground truth centerline in synthetic vessel models with an average distance of 0.25 mm and overlap measure of 96.0%, given the CT scans with a resolution of about 0.3 mm × 0.3mm × 0.4mm. Second, the proposed method was further tested in six clinical coronary arteries models reconstructed from computed tomography coronary angiography in human patients and found to be applicable in both left coronary arteries and right coronary arteries with an average processing time of 16 minutes per case. In conclusion, the proposed GVF field and the fast marching-based method should have more routine clinical applicability.