SGH and A*STAR Move AI Antibiotic Testing Toward Commercial Scale

Singapore General Hospital (SGH) and A*STAR’s Diagnostics Development Hub (DxD Hub) are moving the Individualised Antibiotic Combination Test (iACT) from a specialised hospital service toward a scalable diagnostic product. The goal is to make antibiotic selection more targeted for difficult-to-treat infections while creating a locally developed technology that can be deployed in more laboratories and, eventually, overseas.

What iACT does

Drug-resistant infections can leave clinicians choosing among medicines that may be ineffective, toxic or difficult to combine. iACT is an in-vitro service designed to test the patient’s own bacterial isolate against antibiotic combinations rather than assessing each drug only in isolation. Its earlier development focused on extensively drug-resistant Gram-negative bacteria and evaluated combinations at clinically relevant concentrations.

The platform is described as containing 180 proprietary antibiotic combinations. A predictive software layer interprets growth-inhibition results and helps clinicians rank plausible regimens for the particular organism. That gives infectious-disease and microbiology teams an additional evidence source alongside conventional susceptibility testing, clinical judgement and the patient’s condition.

The intended benefit is more rational, individualised treatment—not an automatic prescription. The test does not remove the need for specialist oversight, and its usefulness can vary by pathogen, resistance pattern, turnaround time and clinical setting.

Why the test matters for antimicrobial stewardship

The clinical rationale is especially important when standard options are limited. Identifying a combination that works against a particular isolate could help clinicians avoid or reduce reliance on last-resort antibiotics such as polymyxins or colistin, which are associated with substantial kidney-toxicity concerns.

That is a potential stewardship benefit, not a guarantee for every patient. The strongest case for iACT will depend on whether testing changes treatment decisions quickly enough, improves outcomes in real-world settings and does so at a cost hospitals can support.

Early economic evidence is encouraging but narrower than a universal clinical endorsement. A 2026 analysis concluded that adopting iACT for carbapenem-resistant Pseudomonas aeruginosa infections in Singapore was likely to be cost-effective. That finding does not by itself establish clinical benefit across every pathogen, hospital or healthcare system.

Why SGH and A*STAR want a dry-format kit

The current iACT format is broth-based and requires frozen handling. That creates practical obstacles for manufacturing, inventory, transport and use in laboratories that cannot easily maintain demanding cold-chain conditions.

The SGH–DxD Hub collaboration is intended to convert the test into a stable dry-format kit. A dry product could simplify shipping and storage, extend practical shelf life and make preparation more consistent across laboratories. Those characteristics are important if iACT is to move beyond a bespoke service at SGH and become a repeatable diagnostic product.

The format change is therefore more than a packaging exercise. It is a scale-up step connecting a clinician-developed method to manufacturing, quality validation, regulatory review and routine laboratory workflows.

Commercialisation plans and remaining hurdles

The organisations signed a three-year memorandum of understanding to productise and commercialise iACT. SGH had used the test 84 times in the three years before the announcement, and the stated direction is to expand access to additional Singapore hospitals and laboratories before pursuing overseas deployment.

The commercial proposition combines two elements:

• a proprietary panel of antibiotic combinations;
• software that helps interpret laboratory results and support treatment selection.

That combination could give iACT a differentiated position in precision infectious-disease diagnostics. However, commercial success will depend on more than technical performance. Developers and healthcare systems will also need to establish a workable test price, turnaround time, manufacturing process, regulatory pathway, clinical workflow and reimbursement model. Local resistance patterns and treatment practices will matter as the test moves between countries.

The wider healthcare-AI programme

The iACT agreement was announced alongside other initiatives that show how Singapore is linking clinical deployment with research partnerships, governance and regional expansion.

PENSIEVE-AI: lower-burden cognitive screening

PENSIEVE-AI is a digital pre-dementia screening tool developed by SGH with GovTech, Duke-NUS Medical School and the National University of Singapore. It analyses drawing tasks completed on a touchscreen device to identify signs associated with early cognitive problems. A reported study involved 1,758 community-dwelling adults, and the tool is designed to complete the assessment in under five minutes.

Its role is screening, not diagnosis. The appeal is that a relatively brief, accessible assessment could help identify people who may need further clinical evaluation, particularly as Singapore responds to population ageing.

Bhutan: localising chest-X-ray AI

SingHealth and Bhutan’s Gyalpozhing College of Information Technology agreed to adapt an AI-assisted chest-radiograph model based on MerMED-FM using Bhutanese patient data. The project is intended to support detection of lung infection and cancer in settings where access to specialist radiology expertise can be limited.

The plan includes responsible-AI guidelines, capability building and deployment in three Bhutanese hospitals by 2027.

Localisation is central to the project. A medical-AI model trained or developed in one healthcare environment should not simply be assumed to perform identically in another: populations, disease prevalence, imaging equipment, clinical workflows and regulatory expectations can differ. Testing and adapting the model with local data is therefore part of making deployment safer and more useful.

What Singapore’s approach reveals

Taken together, the announcements suggest a practical health-and-industry model rather than an AI strategy built around technology alone.

First, the projects begin with defined clinical problems: antimicrobial resistance, early cognitive decline, limited imaging expertise and chronic-disease risk. Second, they connect public hospitals with research agencies, universities and government technology partners. Third, they treat deployment requirements—product formats, local validation, governance and clinical oversight—as part of innovation rather than as an afterthought.

For iACT, that means converting a promising hospital-based testing service into a manufacturable kit. For the Bhutan project, it means adapting an imaging model to the local environment instead of exporting it unchanged. The common thread is translation: turning clinical research into tools that can be used responsibly at scale, while also creating intellectual property, products and regional partnerships.

The opportunity is substantial, but the evidence should be read carefully. iACT’s cost-effectiveness result is specific to a defined infection and setting, while broader deployment still requires validation and regulatory work. Singapore’s strategy is not simply to announce medical AI; it is to build the infrastructure needed to test, govern, commercialise and use it in practice.