Researchers in Singapore and South Korea have developed new AI-driven systems to detect and predict gastrointestinal cancers, offering more accurate diagnostics and potentially earlier interventions. In Singapore, a team from Singapore General Hospital (SGH) and ASTAR Institute of Molecular and Cell Biology (ASTAR IMCB) created the Tumour Immune Microenvironment Spatial (TIMES) Score, a predictive tool for assessing the recurrence risk of hepatocellular carcinoma, a common type of liver cancer. By analysing the spatial distribution of natural killer cells and five specific genes within liver tumour tissues, TIMES can predict recurrence risk with 82% accuracy, outperforming current staging methods. Validated using tissue samples from 231 liver cancer patients across five hospitals, the system will undergo further validation at SGH and the National Cancer Centre Singapore later this year. TIMES is currently accessible via a free web portal for research, with plans to integrate it into clinical workflows and develop a diagnostic test kit. Dr Joe Yeong, the study’s principal investigator, said that with up to 70% of Singapore’s liver cancer patients experiencing recurrence within five years, the system could enable earlier intervention, while senior research officer and study co-author Denise Goh added that identifying high-risk patients allows for proactive adjustments to treatment and monitoring strategies.
In South Korea, Seoul National University (SNU) researchers developed ColonOOD, a computer-aided diagnosis (CAD) system for detecting colonic polyps with the added capability of confidence scoring. Trained and validated on around 3,400 colonoscopy datasets from four local hospitals and two public sources, ColonOOD can localise polyps, classify them with confidence scores, and detect out-of-distribution (OOD) polyps—something existing systems cannot do. While current CAD tools typically differentiate only between adenomas and hyperplastic polyps, ColonOOD can identify the location, classify types from colonoscopy images, and detect rare, minor polyps by learning their distribution. In validation studies published in Expert Systems with Applications, the system achieved up to 79.7% accuracy in classifying all colonic polyps and up to 75.5% accuracy for minor polyps. Study co-lead Dr Dong-heon Lee said future work will focus on verifying usability in prospective and multi-institutional trials.
These developments reflect a wider regional embrace of AI in gastrointestinal healthcare. A recent survey of digestive system specialists in Asia revealed strong clinician trust in AI for diagnosing, removing, and characterising colorectal polyps. Several hospitals in Singapore, including Farrer Park Hospital, Sengkang General Hospital, and the upcoming National University Centre for Digestive Health, are already implementing or planning multiple AI systems for detection, diagnosis, and quality control of cancerous gastrointestinal lesions. In Thailand, Chulalongkorn University has also developed an AI-based tool for identifying colonic polyps, highlighting the momentum of AI-powered innovation in digestive cancer care across Asia.
Click here to read the original news story.