Colorectal Tumor Classification

Early detection of colorectal cancer through colonoscopy is a standard procedure that relies on visual inspection of lesions using Narrow Band Imaging (NBI) zoom-videoendoscopes. This study presents a computer-aided recognition system that classifies NBI images of colorectal tumors into three types based on NBI magnification findings, using a local feature-based method with a bag-of-visual-words (BoW) approach and Support Vector Machine (SVM) classifiers. The proposed system achieved a high recognition rate of 96% in cross-validation and 93% in an external test dataset, demonstrating its effectiveness in classifying colorectal tumor images.