Feature-Based Transfer Learning as a Means for Colorectal Cancer Diagnosis

Colorectal cancer remains a significant global health concern, necessitating accurate and efficient diagnostic tools for early detection. Transfer learning, specifically feature-based transfer learning, has shown promise in enhancing the performance of medical image analysis tasks. In this study, we investigate the effectiveness of feature-based transfer learning, leveraging the InceptionV3 architecture, for the early detection of colorectal cancer. A dataset of 800 histological images containing eight classes of different tissue types was utilized in a ratio of 70:15:15 for training, validation and testing. The pre-trained InceptionV3 convolutional neural network (CNN) model was used to extract discriminative features from the images. Three classifiers were used, namely k-Nearest Neighbors (kNN), Support Vector Machine (SVM) and Logistic Regression (LR), to classify images based on the extracted features. The findings from the present study suggest that the InceptionV3 + LR pipeline was able to discern the classes well.