When doctors and AI collaborate, patients benefit

The implementation of AI into the medical field has the potential for great improvements — as well as great harm. IHH Healthcare Singapore has started the cautious foray down this road using AI to enhance their diagnostics.

Photo: Anna Shvets
Share this article

In healthcare, time is most certainly of the essence. Nowhere is this more evident than in diagnostics, where accuracy and speed can dramatically alter a patient’s prognosis. To enhance both, IHH Healthcare Singapore has begun integrating artificial intelligence into its hospitals — which include Mount Elizabeth and Gleneagles — and considerably improved accuracy of diagnoses, including that of cancer.

One of the most pressing challenges radiologists face today is the sheer volume and complexity of modern imaging. A single low-dose computed tomography (CT) scan can produce between 400 to 600 images. Each of these must be reviewed manually, where radiologists search through grainy lines and spots to identify abnormalities that may be as small as 4 millimetres in size. 

In this context, these tools can flag minute anomalies that a human eye might miss, and do so at a pace that exceeds that of humans. For diseases such as lung cancer, where early detection is key to effective treatment, this kind of support can be life-changing.

Sharper eyes, better detection

The technology is incorporated in different ways for its every function. Take for instance, Parkway Radiology, where the tool Lunit INSIGHT MMG is implemented for mammography, SenseCare Chest CT is used for chest CT, and Annalise.ai, which can be used for chest X-rays. 

With Lunit INSIGHT MMG, deep learning assists radiologists in identifying abnormalities, improving detection sensitivity by up to 5% in dense breast tissue and 12% in fatty tissue. 

While these figures may sound modest, it matters significantly in Singapore, where 1 in 13 women will be diagnosed in their lifetime, and where an estimated 10% to 30% of tumours go undetected in standard mammography. By applying deep learning to scan analyses, the system supports radiologists in identifying potential tumours more consistently, offering women a better chance of realising that they have cancer early.

Managing the surge

The value of AI becomes even more apparent when scaled across growing diagnostic demand. Since the COVID-19 pandemic, IHH Healthcare Singapore have seen a dramatic rise in CT screenings for lung and chest conditions — from 1,483 in 2019 to over 4,000 in 2024. 

Deploying SenseCare Chest CT and Annalise.ai have aided in managing this influx. SenseCare Chest CT has led to a 31% reduction in report generation time and increased nodule detection rates from between 64% and 66%, to between 75% and 81% within a year.

Similarly, Annalise.ai has improved chest x-ray reporting times by 13% and in some cases, increased detection accuracy by up to 12%. These tools help radiologists zero in on potential abnormalities faster, enabling them to prioritise urgent cases and improve overall reporting efficiency.

Deliberate integrations

Due to the nature of the medical industry, where the wellbeing and lives of individuals are at stake, IHH Healthcare Singapore have taken a cautious and deliberate approach as they integrate AI. For example, Parkway Radiology pilots each system in a dedicated reporting centre by senior radiologists. Only after stringent performance thresholds were met, was the tech deployed.

Each AI was also implemented in deliberately specific ways, with some tools are embedded directly into scanning protocols, and others work in tandem with reporting systems to flag issues and compare scans with previous studies.

“Despite concerns globally about AI introducing bias or being used as a substitute for clinical judgment, Parkway Radiology’s approach underscores the importance of human oversight,” said Dr Peter Goh, Senior Consultant Vascular and Interventional Radiologist, and Group Medical Director, Parkway Radiology.

 “AI tools are seen as safety nets and productivity boosters, not autonomous diagnosticians. The final call remains with the radiologist, who must interpret, verify, and contextualise every finding. The systems are used to check the adequacy of reports and suggest what might otherwise go unnoticed — but never to override human expertise.”

Bias and the need for accountability

This includes the matter of bias. The introduction of AI always inevitably raises concerns about bias — both in data and in outcomes. However, medical diagnosis itself always involves a degree of subjectivity, whether it is patients report their symptoms or doctors interpreting these based on objective criteria. The present of a doctor becomes pertinent to identify when an abnormality is actually significant after being identified.

“We are using AI to help detect abnormalities and to suggest possible diagnoses,” explained Goh. “However, any bias that AI replicates should ideally mirror the bias informed by our own clinical experience as radiologists.”

In other words, even as AI tools aid with productivity and provide safety nets to minimise the chances of abnormalities going undetected, the doctor presence is still non-negotiable. The radiologist is responsible to assess the relevance and significance of what the AI surfaces and to ensure that no decision is made without human oversight.

This means that above all, there is always someone responsible. And that matters when so much is at stake.

The road ahead

The presence of AI in the healthcare field will only continue to grow, with IHH Healthcare Singapore actively looking toward AI to aid with other parts of the body. In one case, Parkway Radiology is now actively exploring AI’s use beyond detection, particularly in treatment planning for interventional radiology, where medical imaging is used to perform minimally invasive procedures. Within this research, AI-assisted robots may be implemented.

In the face of increasing diagnostic loads and the never ending quest for perfect health, AI has the potential to not only improve efficiencies, but also sensitivities — though only ever under the discerning eye and involved hand of a doctor at the helm. 

Share this article