Enhancing Diagnostic Precision for Lung and Colon Cancers Using an Advanced Xception Deep Learning Model on Histopathological Imagery

Lung and colon cancers are among the most prevalent and deadly cancers globally, with patient outcomes heavily dependent on early and accurate diagnosis. The current standard, histopathological analysis conducted by pathologists, is a meticulous process susceptible to subjective interpretation, observer fatigue, and significant time delays. These challenges can lead to diagnostic inconsistencies, particularly in regions with a high volume of cases or limited access to specialist pathologists. This research addresses these critical limitations by proposing a novel, automated diagnostic framework based on a sophisticated deep-learning architecture. We adapted the Xception convolutional neural network, renowned for its efficiency and accuracy, to classify histopathological images from the comprehensive LC25000 dataset. This dataset contains 25,000 high-resolution images spanning five critical tissue categories: colon adenocarcinoma, benign colon tissue, lung adenocarcinoma, lung squamous cell carcinoma, and benign lung tissue. Our customized Xception model demonstrated exceptional performance, achieving near-perfect metrics with an accuracy, precision, and recall of 99.96% and an F1-score of 99.92%. These results signify a robust ability to distinguish between malignant and benign tissues with remarkable reliability. The study concludes that this AI-driven framework holds immense potential to serve as a powerful decision-support tool for pathologists. By augmenting human expertise, it can accelerate diagnostic workflows, enhance diagnostic consistency, reduce the burden on healthcare systems, and ultimately contribute to earlier intervention and improved survival rates for cancer patients.