Case Studies

DeepGI: Real-Time AI-Assisted Endoscopy for Colorectal Cancer Prevention

Background and Necessity: Colorectal cancer is the world’s third most common cancer and the second leading cause of cancer-related deaths, with 1.9 million new cases and 900,000 deaths globally in 2020. In Thailand, it is also the third most prevalent cancer, with over 15,000 new cases annually and more than 5,400 deaths per year. Colonoscopy is the gold standard for screening, but its effectiveness depends heavily on physician expertise, with a miss rate of up to 22%. Limited numbers of endoscopists further exacerbate the challenge. Against this backdrop, DeepGI was developed as a collaborative innovation between Chulalongkorn University’s Faculty of Engineering and Faculty of Medicine, together with industry partners, to harness artificial intelligence for real-time polyp detection and cancer prevention. The primary objective is to enhance adenoma detection rates (ADR) and ultimately reduce preventable colorectal cancer cases.

Results and Wider Benefits: DeepGI has reached TRL-7, with multi-hospital testing across Thailand and randomized controlled trials involving over 1,200 patients at five centers. The project has produced numerous high-impact publications in both engineering and medical journals, highlighting its interdisciplinary strength and global visibility. Clinical validation showed significant gains: adenoma detection rate (ADR) rose by 17% (38%→55%), proximal ADR by 16% (23%→39%), and advanced ADR by 3.5% (6%→9.5%). The system also distinguished neoplastic from hyperplastic polyps with 79% accuracy and 85% sensitivity. These results confirm DeepGI’s capacity to improve diagnostic quality, reduce missed cancers, and save lives. Societal benefits include lower treatment costs through prevention, equitable support for physicians of varying expertise, and stronger public confidence in screening. Economically, DeepGI is more affordable than commercial alternatives, expanding access in resource-limited settings and positioning Thailand at the forefront of medical AI innovation.

Future Directions and Impact: DeepGI continues to advance AI research in gastrointestinal health, extending to gastroscopy for detecting gastric intestinal metaplasia (GIM) and cholangioscopy for bile duct abnormalities, both demonstrated through live broadcasts from Thailand to Korea at IDEN 2024 and 2025. The system is being adapted for multi-organ detection, further expanding its clinical applications. Beyond healthcare, DeepGI stands as a model of interdisciplinary STEM innovation, inspiring students and the wider public by demonstrating how AI and engineering can drive early cancer detection and timely treatment—ultimately saving lives and enhancing long-term quality of life.

Conclusion: DeepGI exemplifies STEM research with far-reaching public benefit. By addressing a major health burden with a real-world, clinically validated AI solution, it demonstrates how higher education research can save lives, reduce inequalities, and inspire the imagination of society.

By

  • Faculty of Engineering, Chulalongkorn University
  • Faculty of Medicine, Chulalongkorn University
PDPA Icon

We use cookies to enhance your browsing experience (Settings)

Privacy Preferences

You can choose your cookie settings by enabling/disabling cookies for each category as needed, except for necessary cookies.

Accept All
Manage Consent Preferences
  • Necessary cookies
    Always Active

    Necessary cookies are essential for the functioning of the website, allowing you to use and browse the site normally. You cannot disable these cookies in our website's system.

  • Analytical Cookies for Enhancing User Experience

    These cookies are used to collect information about website usage, such as the number of visitors, popular web pages, and browsing behavior, which helps the website owner improve the user experience.

  • Functional Cookies for Remembering User Settings

    These cookies help enhance the website's functionality by remembering user settings such as username, language, region, or customizations.

Save