Colon cancer detection using whole slide histopathological images

Introduction: Colon cancer is one of the major reasons of cancer-related deaths, whereas both the traditional and the current methods are complex to conduct. Materials and Methods: This paper presents a simple and effective approach for colon cancer detection by classifying cancer and non-cancer colon images based on computer assisted diagnosis. A classifier is developed using SVM, which has excellent performance in practice. Eighteen simple features, including grayscale mean, gray-scale variance and 16 texture features extracted by Gray-Level Co-occurrence Matrix (GLCM) method, are chosen as the feature set. Results: In order to evaluate the accuracy of our classification, we calculate precision, recall and F-measure of different classifiers produced by using different feature combinations. And 3-fold cross-validation is applied. Three indicators precision, recall and F-measure are used to describe the performance of our system. Experiment results show that: when all features are used, the mean value of precision, recall and F-measure are 96.67% 83.33% 89.51% respectively. Discussion: These results demonstrate the great advantage of the method on colonic histopathology images' classification. The simple and efficient method will have great contributions on colon cancer detection.