Enhancing Colorectal Cancer Histological Image Classification Using Transfer Learning and ResNet50 CNN Model

Abstract

Medical image analysis is crucial in healthcare research. The convolutional neural network (CNN) has great potential in improving the precision and speed of medical diagnosis. In medical diagnostics, CNNs have displayed promising results, indicating their capability to enhance the accuracy and efficiency of the diagnostic process, accurately classifying complex medical images remains challenging. Colorectal cancer, a significant cause of global mortality, emphasizes the need for early detection and diagnosis to ensure successful treatment. We develop a new method combining transfer learning and a ResNet50 CNN model with the Adam optimizer to increase the accuracy in the classification of the histopathology images of colorectal cancer. The experimental results demonstrated outstanding performance with an accuracy of 99.99% in training and an accuracy of 99.77% in validation which were excellent performance on widely recognized evaluation metrics. In conclusion, the proposed method surpasses other related studies using CNN models for histopathology image classification. It provides a practical solution to further improve the classification performance of colorectal cancer histopathology images. The study result shows the efficacy of transfer learning in the analysis of medical images. Moreover, the proposed approach outperforms existing methods in medical image analysis, underscoring its potential to empower medical professionals in enhancing diagnostic capabilities and making more informed clinical decisions for patients.