Object contour extraction in medical images by fast adaptive B-Snake

Detection of edge contours to define the object class and segmenting it from the background(s) class is a major challenge in medical image processing. In this paper, we introduce a fast adaptive B-Spline Snake algorithm that overcomes the limitations of previous B-Snakes by avoiding computationally expensive optimization stages. Furthermore, we present novel strategies for adaptive knot insertion to fulfill the specific requirements of the medical application accordingly. Experimental results on pulmonary CT images and brain MR images demonstrated that our method is superior both in accuracy and convergence speed over previous B-Spline Snake algorithms. The proposed algorithm is also robust to missing the boundaries.