Liver MRI segmentation with edge-preserved intensity inhomogeneity correction

The accurate liver segmentation for MRI is challenging because of the intensity inhomogeneity. However, most existing intensity inhomogeneity correction sometimes leads to detail smoothing. In this paper, a novel model is proposed for liver segmentation based on the level set method with edge-preserved intensity inhomogeneity correction (EPIICLS). EPIICLS corrects the intensity inhomogeneity with the minimization of the local entropy. And the edge-preserving filter is applied to compensate the detail smoothing by cooperating with the original image. The intensity inhomogeneity correction works on the internal energy of the level set method, which efficiently makes the level set function automatically approximate the signed distance function. And the edge preservation works on the external energy function, which precisely drives the zero-level curve toward the liver boundaries. Experiment results show that the proposed model leads to better segmentation results.