Colon Cancer Detection Using Deep Learning Algorithm

Colon cancer is among the most prevalent and deadly forms of cancer globally, often progressing silently until reaching advanced stages. Early and accurate detection is critical for improving prognosis and survival rates. Traditional diagnostic methods such as colonoscopy and histopathological examination, while effective, are often invasive, time-consuming, and subject to human error. This study proposes a comprehensive machine learning-based framework for the automated detection and grading of colon cancer using multi-modal imaging data, including colonoscopy visuals, MRI scans, and histopathological slides. A combination of deep learning models, including Convolutional Neural Networks (CNN), Residual Networks (ResNet), and U-Net, alongside traditional machine learning algorithms like Support Vector Machines (SVM) and Random Forests (RF), is employed to perform classification and segmentation tasks. The dataset is sourced from The Cancer Imaging Archive (TCIA), Genomic Data Commons (GDC), and hospital records, ensuring diverse and annotated images representing various stages of colon cancer. The data undergoes rigorous pre-processing and augmentation to enhance quality and address class imbalances. The hybrid model achieves high accuracy, precision, recall, and F1-score, with superior performance in tumor segmentation using Dice Similarity Coefficient and Intersection over Union. Interpretability is enhanced using Grad-CAM and SHAP to visualize model decisions and feature importance. Evaluation results demonstrate that the proposed system not only achieves expert-level diagnostic accuracy but also significantly reduces processing time, offering potential for real-time clinical deployment.