Active Contours with Weighted External Forces for Medical Image Segmentation

Parametric active contours have been widely used for image segmentation. However, high noise levels and weak edges are the most acute issues that hinder their performance, particularly in medical images. In order to overcome these issues, we propose an external force that weights the gradient vector flow (GVF) field and balloon forces according to local image features. We also propose a mechanism to automatically terminate the contour’s deformation. Evaluation results on real MRI and CT slices show that the proposed approach attains higher segmentation accuracy than snakes using traditional external forces, while allowing initialization using a limited number of selected points.