A Modified Framework for Multislice Image Fusion for High Contrast Liver Cancer Detection

The exact boundary extraction of liver using abdominal computed tomography (CT) images continues to be the greatest challenge in the computer-assisted diagnosis of liver tumour, as the organ boundary is very weak. In this paper, an efficient algorithm is proposed to enhance the edge information of the CT images prior to segmentation, by means of multislice image fusion and anisotropic diffusion filtering in non-subsampled contourlet (NSCT) domain. The two adjacent slices of CT images are decomposed using NSCT, and the fusion of low- and high-frequency coefficients is obtained by means of phase congruency and sum-modified Laplacian operators, respectively. The major highlight of this work is that, prior to high-frequency fusion; the high-frequency coefficients of both images are processed using anisotropic diffusion to strengthen the edge information. Finally, the edge-enhanced image is obtained by NSCT reconstruction. The performance metrics show that the information pertaining to the edges is found to be precise in the fused image. The proposed logic, when applied to the real-time contrast-enhanced triple-phase CT image, has proven to be highly effective. All the image samples used in this work for test purpose were obtained from Jawaharlal Institute of Postgraduate Medical Education Research (JIPMER), a Medical Research Institute and Hospital at Puducherry, India.