Colon Cancer Detection in Biopsy Images for Indian Population at Different Magnification Factors Using Texture Features

This paper presents the identification of colon histology images as normal and malicious ones for the determination of colon cancer. Such classification is very difficult as it takes into account numerous characteristics such as color, several biological interpretable factors. Also this process is subjective and may have some observational variations. In this paper the colon biopsy images are classified as normal and malicious ones based on the texture of the images for the Indian population. Various texture features such as GLCM, LBP, HOG, Gabor, GLRLM, Histogram are extracted from the images with different magnification factors such as 10X, 20X and 40X. Based on these features extracted at different magnification factor, they are trained using classifiers SVM with linear kernel, Naive Bayes and Perceptron. The system has been experimented on 24 normal and 35 malignant biopsy images of colon that were acquired fom Aster Medcity, Kochi. SVM and Perceptron is giving higher accuracy for magnification 10X and 20X.