Automatic Initialization of Level Set Methods in Discontinuity Regions of Liver in CT Scans

Background: This research presents a novel automatic approach to initialization of contour curve problem for the level set segmentation method inside liver discontinuity regions in CT 2D images. This approach allows different types of level set methods to achieve accurate results in segmenting liver regions. Methods: The proposed approach includes three steps. First, a pixel-level texture feature extraction of an abdominal CT scan using Gray Level Co-occurrence Matrices (GLCM) is performed. Second, Principal Component Analysis (PCA) classifier is used to classify the pixel texture features in abdominal CT scans to get the discontinuity regions (multiple lobes) of the liver. Finally, an initial contour curve inside discontinuous regions (lobes) of liver is generated and level set segmentation is performed. Results & Conclusion: Experimental results on abdominal CT scans demonstrate the significant performance of our approach.