Scientists develop AI system that can predict liver cancer recurrence

AI tool predicts liver cancer relapse with 82% accuracy, tackling a disease causing third-highest cancer-related deaths globally

In a major medical breakthrough, scientists in Singapore have developed an AI-powered diagnostic tool capable of accurately predicting the recurrence of liver cancer — specifically hepatocellular carcinoma (HCC), one of the deadliest cancers globally.

 

The Tumour Immune Microenvironment Spatial (TIMES) score, an innovative diagnostic system, has been developed through a joint effort by researchers at the Agency for Science, Technology and Research (A*STAR)’s Institute of Molecular and Cell Biology (IMCB) and Singapore General Hospital (SGH), according to a press release from SGH.

 

The TIMES model, which was recently featured on the cover of the renowned journal Nature, is being seen as a game-changer in personalised cancer diagnostics and early intervention. 

 

What is hepatocellular carcinoma?

Hepatocellular carcinoma (HCC) is the most common type of liver cancer, often linked to chronic liver diseases such as cirrhosis. It remains the third leading cause of cancer-related deaths worldwide, and recurrence rates are alarmingly high.

 

In Singapore alone, around 70 per cent of liver cancer patients experience a relapse within five years of treatment, making early detection of the recurrence vital for improving survival rates.

How the TIMES system works

The TIMES score uses advanced machine learning and spatial biology to assess the likelihood of liver cancer returning after surgery. By integrating multiplex immunofluorescence imaging, spatial transcriptomics, and proteomics data, the model uses the XGBoost machine learning algorithm to detect molecular patterns within tumour tissue—patterns that traditional diagnostic methods cannot identify. 

 

Specifically, it evaluates the distribution of natural killer (NK) cells and the expression of five key genes inside the tumour microenvironment. This combination allows the AI to determine a patient’s risk of recurrence with approximately 82 per cent accuracy, outperforming existing clinical tools. 

Potential of the TIMES system

Early and accurate prediction of relapse means that doctors can tailor follow-ups and treatment plans more effectively. This would increase the chances of long-term survival. 

 

According to Dr Joe Yeong, Principal Investigator at A*STAR IMCB and SGH’s Department of Anatomical Pathology, the TIMES system represents a big leap in the ability to anticipate cancer relapse and initiate timely intervention.

 

The study also identified a biomarker called SPON2, produced by NK cells. SPON2 has been found to be associated with the risk of recurrence. Studies have further revealed that SPON2 and NK cells enhance anti-tumour activity by improving migration towards cancer cells and activating CD8 and T-cells. This finding could also pave the way for improved AI-guided immunotherapy.

 

Denise Goh, co-first author and senior research officer at A*STAR IMCB, explained, “TIMES turns standard pathology slides into predictive diagnostic tools. Not only does the AI algorithm improve prognostic precision, but it also enables clinicians to revise treatment and monitoring plans proactively — potentially saving lives.” 

Validated and ready for wider use

The TIMES model was tested using tumour samples from 231 patients across five hospitals, demonstrating its reliability across diverse datasets. To encourage global collaboration, the team has also launched a free online portal that allows medical professionals to upload tissue images and get AI-generated recurrence risk assessments.

 

The underlying software framework has been patented, and further validation trials are scheduled at SGH and the National Cancer Centre Singapore later this year. The research team is currently working with diagnostic partners to standardise the system and transform it into a clinically approved diagnostic kit for routine hospital use.

 

SGH, Singapore’s largest tertiary healthcare institution and a globally recognised academic medical centre, played a key role in this project and will continue to support its clinical rollout. If successful, the TIMES score could become a key breakthrough for future cancer care.