An Effective Segmentation Method for MRI Images Based on TV-L1 and GVF Model

Abstract

Liver magnetic resonance imaging (MRI) is of vital importance for computer-aided diagnosis and it is equally important in liver surgery planning. In this paper, accurate contours of the liver in MRI images automatically for subsequent adaptive radiation therapy can be extracted by the work. It is consisted of three components. Firstly, noise and artifacts are removed from the MRI image by an edge-preserving filtering using total variation with L1 norm (TV-L1). Secondly, the wavelet parameters are calculated at different levels of scale, and then the initial contour of the liver is obtained at the appropriate scale. And finally the precise liver structure is extracted by the gradient vector flow (GVF) model converging to the initial contour. The accuracy of the segmentation results are verified by comparing with the manually ones. For clinical cases, the numerical results illustrates enough accuracy and robustness for medical environments. And it also has a reasonable computational cost.