DefinePK

Abstract

Computed Tomography (CT) images have been widely used for diagnosis of liver disease and volume measurement for liver surgery or transplantation. The approach is presented with respect to liver segmentation, but it can be easily extended to any other soft tissue by setting appropriately the values of the parameters for the splitting and merging algorithm and for the region growing refinement step. Sparse coding with data-adapted dictionaries has been successfully employed in several image recovery and vision problems. A novel, automated segmentation technique for detecting affected region in liver was proposed in this paper. In the new approach, we constructed ensemble kernel matrices using the pixel intensities and their spatial locations, and obtained kernel dictionaries for sparse coding pixels in a non-linear feature space. The resulting sparse codes were used to train an Extreme Learning Machine (ELM) classifier that determines if a pixel in the image belongs to an affected region. From the experimental results using ten test datasets distributed for the competition, it was confirmed that our method kernel sparse coding based liver segmentation performs better than previous methods or models.