AI models tuned to S’pore patients’ clinical data being built under new national drive

SINGAPORE – Healthcare artificial intelligence models tailored specifically for Singaporean patients and medical practices are being developed under a new national initiative.

Announced on July 9, the initiative called Singapore Medical Foundation AI Model (SIMFONI) aims to help clinicians better diagnose conditions such as diabetes, high cholesterol and eye diseases by addressing a gap in the data used for AI training.

The problem is that most AI foundation models used in healthcare today are trained on data from Western populations, which can limit the accuracy and relevance in Singapore’s clinical settings.

“In other words, they haven’t gone to our local medical school. SIMFONI would have gone to our local medical school,” Health Minister Ong Ye Kung said, in announcing the new initiative at the NCS Impact 2026 conference.

Ong said that the AI models can offer possible diagnosis, treatment pathways and next steps, but doctors still make the final call.

The plan is for the AI models to be deployed throughout the public healthcare system when it is ready, he added, without disclosing the timeline.

SIMFONI is a programme under the Consortium for Clinical Research and Innovation, Singapore (Cris), which brings together several research programmes of the Ministry of Health.

Executive director of SIMFONI, Professor Robert Morris, said that the AI models are still being selected. They need to be able to interpret medical images, understand clinical records and support clinical reasoning.

Many of the AI models available also differ in performance, safety behaviour and how readily they can be adapted to Singapore’s clinical context.

“Each candidate goes through a rigorous evaluation and selection process. (They are) tested on established medical benchmarks, assessed against Singapore’s clinical guidelines, and validated on local data,” said Morris.

De-identified clinical data from Singapore’s public healthcare system that reflects the local population will be used.

One of the AI models will be trained to help doctors manage diabetes, high blood pressure and high cholesterol – which together is one of the largest chronic disease burdens on Singapore’s primary-care systems.

For instance, Asians tend to develop diabetes at a lower body mass index compared with Westerners, and genetic risk factors also differ.

So, the AI system will need to identify patients who may need early treatment, and check Singapore’s national clinical guidelines to ensure recommendations follow current standards of care. Then, the system will assess the patient’s medical history, current condition and other relevant data to suggest treatment options or complications that the doctor may need to look at.

Another focus area is eye diseases such as cataracts, retinal diseases and glaucoma. New multimodal AI systems that can understand text, images and audio will be designed to process conversations with patients, eye images and medical records into clinical notes to support doctors’ decisions.

As the eye provides valuable insights into other conditions – including diabetes, and cardiovascular and neurological disorders – the work lays the foundation to expand into other medical specialities.

Ong also outlined three criteria needed to deploy AI well in healthcare: A strong digital operating environment, good quality data, and a sound policy and organisational structure.

“Good use of AI is not just about having the best AI tools. There need to be prerequisites. Otherwise, it is like having a state-of-the-art home appliance, say, a very smart TV. But so what? You have no electrical socket,” said Ong.

Taking on board the advice, MOH is in the final phases of replacing numerous isolated IT systems to ensure that systems covering the entire public healthcare sector are integrated for data sharing.

The “Herculean task” has been done over many years, and includes adopting a common electronic medical record system by 2028. It will link all three public healthcare clusters – National University Health System, NHG Health and Singapore Health Services – together, said Ong.

Other improvements to the digital operating environment include the adoption of national IT systems for core functions such as billing, pharmacy operations, patient referrals across community settings and supply chain management.

Singapore also needs to generate good quality data.

Ong pointed out that just as organisations have finance and human resource departments to look after people and money, there needs to be departments looking after data for research, analysis and operational improvement.

One major milestone is the National Electronic Health Record (NEHR) system, where key health information of patients will be shared across public and private healthcare providers.

All healthcare providers will have to contribute to NEHR by early 2027.

Another milestone is the national data analytics platform TRUST that gives public researchers and selected industry partners secure access to nearly 50 anonymised health and health-related datasets.

Ong pointed out that healthcare is a well-regulated sector that carefully considers how to proceed with Al.

“We ask ourselves: What are the problems we face? What are the areas of improvement? How does AI technology help to overcome them?” he said.

For instance, a lot of thought is put into deciding how to automate medical note-taking and clinical coding, predict outpatient attendance to improve hospital operations, help radiologists interpret scans, and identify people at high risk of developing chronic diseases for early intervention. 

Al diagnostic tools are capable of detecting even the slightest anomaly. But it will be designed to highlight only cases where there are established ways to diagnose, treat or follow up on.

Urging the audience to learn to work and coexist with Al, Ong said: “We cannot charge ahead, driven solely by commercial considerations, even as recursive AI systems gain self-reinforcing intelligence, agency and influence. Otherwise, the machines just seem wiser than their makers.”

He added: “We must decide deliberately where to embrace AI, where to rein it in, and where human judgement and effort must prevail.”