Early Detection of Colon Cancer Using Deep Learning Techniques

Colon cancer is the third most prevalent kind of cancer worldwide. A colonoscopy is the global standard for colorectal screening and treatment because it enables the elimination of polyps before they transform into cancerous tumours. Due to the wide variety of polyp types in terms of shape, texture, size, and color, as well as the presence of numerous mimic polyps during colonoscopy, automatic detection of colonic polyps remains an issue that needs more work. Deep learning architectures enhance polyp detection by extracting polyp characteristics. In order to train deep neural networks to recognise polyps in the sparsely available polyp photos, we looked into image enhancement techniques. In our research, we relied solely on public datasets that were supplemented with additional data to yield 3, 972 positive polyp images. Our model is made up of three stages that are preprocessing, polyp detection, and data analysis. In the preprocessing, techniques for image enhancement like rotation and flipping were applied. This model heavily relied on the YOLOv4 detection tool. Finally, the error analysis section aimed to compare the results of our detector with the ground truth for all polyp images. We accomplished 94% precision, 90% recall, and an F1 score of 92%.