Singapore AI heart scan tool heads for wider hospital testing

Researchers seek clinical approval after the model achieved more than 90% accuracy in studies.

Singapore is moving an artificial intelligence (AI) tool that measures heart damage from MRI scans towards clinical use, with researchers planning wider hospital testing before seeking regulatory approval.

CARDIA-GM, developed at the National Heart Centre Singapore (NHCS), achieved more than 90% accuracy for measurements of heart-attack-induced heart muscle scarring and microvascular obstruction (MVO) volumes, Adjunct Clinical Associate Professor Tan Ru San, a senior consultant at NHCS's Department of Cardiology, told Healthcare Asia.

NHCS plans to validate the technology at other hospitals in Singapore before pursuing clinical licensing.

National University Hospital, Changi General Hospital, and Tan Tock Seng Hospital are amongst the potential sites, whilst hospitals in China, Malaysia, and Thailand have also expressed interest, said Assoc Professor Zhong Liang, principal investigator and senior clinician-innovator at the National Heart Research Institute Singapore, NHCS.

“The next 12 to 18 months will focus on testing the technology across a larger and more diverse patient population,” Liang said in a Zoom call with Tan.

CARDIA-GM analyses heart muscle scar and MVO volumes in cardiac magnetic resonance imaging (MRI) scans in less than a minute per patient, compared with about 20 to 30 minutes required for manual analysis by experienced researchers, Tan said.

Liang said the shorter processing time translates into an 80% to 90% reduction in analysis time.

“Our goal is not to say that AI is better than our specialists," Liang said. He described the technology as a “second pair of eyes” and a “copilot” for cardiologists, with doctors retaining responsibility for final clinical decisions.

Tan said detailed measurement of microvascular obstruction is often impractical in routine clinical settings because it requires significant analysis time and experienced specialists.

The measurements may help identify patients at higher risk of complications. Patients with microvascular obstruction exceeding 1.4% of total heart muscle volume face a greater risk of death during follow-up, whilst patients with scarring exceeding 20% have mortality rates three times higher than those below that threshold, Liang said.

Pratibha Thammanabhatla, practice head for medical devices at GlobalData Plc, said investment interest in AI imaging continues to grow as healthcare systems face rising rates of cardiac disease and shortages of radiologists and cardiologists.

Government grants and research institutions, however, remain the main funding sources for many projects in the region, she said.

“Singapore's CARDIA-GM remains grant-funded and has no disclosed full commercialisation path so far,” Thammanabhatla said in an emailed reply to questions.

She said the broader Asian market for AI cardiac diagnostics is expected to expand, although success depends on more than technical performance.

“Singapore therefore has an opportunity to compete through quality and credibility rather than scale,” Thammanabhatla said.

Companies that secure regulatory approvals, integrate their products into hospital workflows, and demonstrate measurable clinical benefits are likely to have the strongest position in the market, she added.