Deep Learning Techniques for Colorectal Cancer Tissue Classification

The objective assessment of histological images is of paramount importance for the early diagnosis of colorectal cancer (CRC). However, the subjectivity of the evaluation interobserver variation and the traditional visual assessment is time-consuming and costly. On the other hand, the automatic recognition and analysis for digital pathology, the challenging remains due to the variability of the histological images containing more than one tissue type and characteristics of the histological images. In this work, we applied different deep learning techniques based on Convolutional Neural Networks (CNNs) to differentiate colorectal cancer from healthy tissues and benign lesions. Firstly, we trained the deep convolution neural network models and in the parts of network layers are extracting modified parameters to identify different tissue types that are abundant in histological images of CRC. Secondly, we used open dataset of histological images of human colorectal cancer including eight different types of tissue to assess the identification rate. In our results, the colorectal cancer tissue identification accuracy rates are significantly superior to the existing known methods. Therefore, the proposed approach could help doctors to enhance their diagnostic abilities and make better clinical decisions for patients.