Segmentation and analysis of liver cancer pathological color images based on artificial neural networks

Liver cancer is one of those sneaky conditions that can disappoint a physician before the diagnosis is finally made. Thus far, the only definitive test for liver cancer is needle biopsy. In this paper, we present an unsupervised approach using Hopfield neural network for the segmentation of color images of liver tissues prepared and stained by standard staining method. We formulate the segmentation problem as a minimization of an energy function synonymous to that of Hopfield neural network for the optimization, with the addition of some conditions to reach a status close to the global minimum in a prespecified time of convergence. Then we extract the nuclei and their corresponding cytoplasm regions which are used as a base for formulating the diagnostic rules of a computer aided diagnosis system for liver cancer. In computer, each liver color image is represented in the R-G-B, H-S-V and H-L-S color spaces and the segmentation results are comparatively presented with discussion and physician comments. Most of the data base of liver color images that we have collected have been successfully segmented with the exception of some images which were not stained carefully.