Image Segmentation of Gastrointestinal Polyps in the Human Gastrointestinal Tract Using Machine Learning

Overview

Currently, colorectal cancer is one of the most common cancers in the world, with it contributing to over 10% of all cancers. With colorectal cancer being a significant contributor to the worldwide mortality rate, it is imperative to find effective methods for treatment and detection. Early disease detection has a huge impact on survival from colorectal cancer, and polyp detection is therefore important. Although colonoscopies have seemingly been perceived to search and find a large amount of potentially cancerous polyps, there are still many issues with thorough detection. Several studies have shown that many polyps are often overlooked during colonoscopies, with 38.69% of patients having at least one polyp missed during colonoscopy, along with a general polyp miss rate of 17.24%. Increasing the detection of polyps has been shown to decrease the risk of colorectal cancer. Thus, automatic detection of more polyps at an early stage can play a crucial role in improving both prevention of and survival from colorectal cancer. This convolutional neural network model, trained on an annotated medical dataset(KVAIR-SEG), is able to accurately detect gastrointestinal polyps with a state-of-the-art accuracy of 96.8%. This machine learning model built using the fast.ai library will immensely improve the polyp detection accuracy of colonoscopies in the future.

Competition history

• ISEF 2024 Translational Medical Science · Entry TMED050 · Los Angeles, California, United States

Resources

• ISEF abstract page ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.