Liver Contour Extraction Using Modified Snake with Morphological Multiscale Gradients

Liver segmentation from CT image is a difficult task due to abdomen apparatus complexity. Snakes, or active contours, are extensively used in medical image segmentation. The automatic generation of initial snake curves and improving snake performance in case of blur edges are still open challenges for liver segmentations. In this paper, texture classification and morphology filter are employed to generate initial Snake points. Then, a novel external force, the gradient map derived from morphological multiscale gradients, is proposed to enhance blur edge information in modified snake algorithm. Experiment shows that the proposed algorithm achieves accurate liver boundaries in different sliced abdomen CT images.