GVF and CV Model-Based Pulmonary Artery Edge Detection of CTPA Image Sequence

Since the gradient vector flow (GVF)model is not suitable for multi-objects edges detection and cannot adapt to the change of the object's geometric topology, and the Chan-Vese (CV)model is easy to result in false detection, this paper proposes a new method for detecting the edges of the pulmonary artery for the computed tomographic pulmonary angiography (CTPA)image sequences which combines the advantages of GVF model and CV model. Firstly, the initial contour is driven by the GVF field. After getting the converged contour curve, the image inside the curve is extracted; Secondly, CV model-based edge detection is performed on the segmented image to solve the problem of multi-objects detection. Experiments show that the proposed method can effectively solve the problem of pulmonary artery edges detection in CTPA images. Applying the algorithm to targets tracking for image sequence, the method can detect the edges of each target separately, and obtain the independent contours, when the pulmonary artery is split into two.