GI tract anomaly detection from endoscopy images

GI tract anomaly detection from endoscopy images

Teams: Bishesh Khanal, Sharib Ali, and Shruti Shrestha

Motivation: 

The aim of this research is to build comprehensive techniques for detection, localization and segmentation of anomalies present in GI organs (not a single organ) and validate it on publicly available benchmark dataset.  

Research Questions: 

  1. Can the state-of-the-art deep learning model give better result on the endoscopy dataset? 
  2. What changes in this deep learning model could be done to further increase the accuracy? 
     

Brief Description: 

Polyps generally occur as a protrusion of the mucosa looking like a bumpy structure. The wide variation in shape, size, intensity of polyps, and specular reflection in colonoscopy images can make polyps very difficult to detect by endoscopists that can have a severe impact on CRC patients and often are contributor to higher mortality rate. Colonoscopy is preferred for detecting and removing the colorectal polyps, which are the predecessors of Colorectal Cancers. Currently, we are focusing on building novel architecture for segmentation of polyps from endoscopy images to bring out better results than the state-of-the-art techniques. Ultimately, our aim is to build a tool for supporting endoscopists to better detect the anomalies during colonoscopy. 

Research Themes: Computational Endoscopy, Surgery & Pathology (CESP), Transforming Global health with AI (TOGAI)
Project Category: Computer Vision, Machine Learning, Medical Imaging