Polyp Detection Using U-Net Deep Learning Method

Polyp Detection Using U-Net Deep Learning Method

Authors- Mr. S. Sinimoxon Lee, Professor Arpita Das

Abstract-Polyp detection in colonoscopy images plays a crucial role in the early diagnosis and prevention of colorectal cancer. This project implements a U-Net model to segment polyp regions from colonoscopy frames using the CVC-ClinicDB dataset, which includes annotated ground truth masks. The U-Net architecture, with its encoder-decoder structure and skip connections, is particularly well-suited for medical image segmentation, enabling precise identification of polyp regions. The model’s performance is evaluated using Intersection over Union (IoU) to measure segmentation accuracy, alongside precision-recall curves to assess detection reliability. Training progress is monitored through the visualization of training and validation loss curves, as well as accuracy curves, ensuring the model’s effectiveness and generalization. The results demonstrate that the U-Net model can significantly improve the accuracy of polyp detection, contributing to more reliable colorectal cancer screening.

DOI: /10.61463/ijset.vol.12.issue5.299