Classification of cancer cells based on morphological features from segmented multispectral bio-images

In this paper a new approach aiming to detect and classify colon cancer cells is presented. Our detection approach was derived from the "Snake" method but using a progressive division of the dimensions of the image to achieve faster segmentation. Classification of different cell types was based on nine morphological parameters and on probabilistic neural network. Three types of cells were used to assess the efficiency of our segmentation and classifications models, including Benign Hyperplasia (BH), Intraepithelial Neoplasia (IN) that is a precursor state for cancer, and Carcinoma (Ca) that corresponds to abnormal tissue proliferation (cancer). Results showed that segmentation of microscopic images using this technique was of higher efficiency than the conventional snake method. The time consumed during segmentation was decreased to more than 50%. The efficiency of this method resides in its ability to segment Ca type cells that was difficult through other segmentation procedures. In classification only three morphologic parameters (area, Xor convex and solidity) were found to be effective to discriminate between the three types of cells. The results obtained using several images show the efficacy of the method.