New AI-driven systems for detecting and predicting gastrointestinal cancers have been developed through research projects in Singapore and South Korea.
WHY IT MATTERS
A research team from Singapore General Hospital (SGH) and A*STAR Institute of Molecular and Cell Biology (A*STAR IMCB) built a scoring system for predicting the recurrence of hepatocellular carcinoma, a common type of liver cancer.
Called the Tumour Immune Microenvironment Spatial (TIMES) Score, it analyses the spatial distribution of natural killer cells and five specific genes within liver tumour tissues to predict cancer recurrence risk.
In a study published in Nature, it has been shown to accurately predict the risk of liver cancer recurrence by 82%, which SGH said "[outperforms] existing staging methods."
The scoring system was validated using tissue samples from 231 liver cancer patients in five hospitals. Further validation studies will be conducted at SGH and the National Cancer Centre Singapore later this year.
"In Singapore, up to 70% of liver cancer patients experience recurrence within five years. TIMES offers a significant advancement in predicting these outcomes, enabling clinicians to intervene at the earliest possible stage," explained Dr Joe Yeong, the study's principal investigator from A*STAR IMCB and SGH.
"By identifying patients at higher risk of relapse, we can proactively alter treatment strategies and monitoring, potentially saving more lives," added Denise Goh, the study's first co-author and a senior research officer at A*STAR IMCB.
The research team has made TIMES accessible via a free web portal for research use. Plans are in place to integrate the AI-driven scoring system into routine clinical workflows; the team is also in talks with diagnostic partners to develop it into a diagnostic test kit.
Meanwhile, researchers from Seoul National University (SNU) in South Korea developed a computer-aided diagnosis (CAD) system for detecting colonic polyps that can also provide confidence scoring.
The system, called ColonOOD, has three functions: polyp localisation, polyp classification with confidence scoring, and out-of-distribution (OOD) polyp detection.
The system was trained and validated using approximately 3,400 colonoscopy data from four local hospitals and two public datasets.
ColonOOD builds on existing CAD systems by introducing a module for detecting and confirming OOD polyps and providing confidence levels following polyp classification. Existing systems, the researchers claimed, can only distinguish between two polyp types – adenomas and hyperplastic.
It automatically identifies the location and classifies the type of polyps from colonoscopy images. It also learns the distribution of major polyps and is capable of detecting the distribution of rare, minor polyps.
Based on a validation study, which findings have been published in Elsevier's Expert Systems with Applications, ColonOOD classified all colonic polyps with up to 79.7% accuracy and confirmed minor polyps with up to 75.5% accuracy.
Study co-lead and SNU Hospital professor Dr Dong-heon Lee said they will verify the system's usability in prospective and multi-institutional studies in the future.
THE LARGER TREND
A recent survey of digestive system specialists in Asia found that most of them accept and trust the use of AI in diagnosing, removing, and characterising colorectal polyps. Clinician-led AI initiatives in the region, such as the projects mentioned above, further demonstrate this confidence.
Another example is the upcoming National University Centre for Digestive Health at the National University Hospital in Singapore, which will be the first centre in the country to implement three AI systems for the detection, diagnosis, and quality control of cancerous gastrointestinal lesions.
Farrer Park Hospital, a private hospital in Singapore, already offers an AI-assisted colorectal screening service. Fellow Singaporean hospital, Sengkang General Hospital, also provides AI-assisted colonoscopy for colorectal cancer screening.
Outside Singapore, Chulalongkorn University in Thailand has also developed an AI-based system for spotting colonic polyps.
