Aspiring Radiologists Should Consider Interventional Field as AI Advances, Says Singapore-Based PCC Official
Key Takeaways
- •Dr. Ang Peng Tiam of Parkway Cancer Centre advised medical trainees to specialize in interventional radiology because AI is expected to take over diagnostic image reading.
- •In a lung cancer case at the center, a radiologist identified three nodules while AI-assisted analysis detected twenty, creating a clinical dilemma over whether surgery was still advisable.
- •Dr. Ang said diagnostic radiology and pathology are the two medical fields most affected by AI because both rely on pattern recognition in images.
- •The U.S. FDA has cleared hundreds of AI-enabled medical devices, many of them in radiology.
- •Despite AI's advances, Dr. Ang stated the technology cannot replace humans in healthcare, which requires trust, care, and hope.
Aspiring radiologists would be wise to specialize in interventional radiology rather than diagnostic radiology, as artificial intelligence (AI) may soon take over the task of interpreting medical images, according to the co-founder of Singapore-based Parkway Cancer Centre (PCC).
"If anyone wants to do radiology, you should not do diagnostic radiology. You must be an interventional radiologist," Dr. Ang Peng Tiam, medical director and senior consultant of PCC, said during a media launch on Friday.
"(This means) that you must be the person who does the angiogram, to block off the vessels, the one who does the biopsy, and not the one who just reads the X-rays," he said.
"Because that reading in the future will all be done by computer and AI. So this is a whole learning process," he added.
Interventional radiology is the subspecialty in which physicians perform minimally invasive, image-guided procedures — such as angiograms, biopsies, and tumor-feeding vessel embolization — hands-on, in contrast to diagnostic radiologists, whose primary role is interpreting scans and reporting findings.
Mr. Ang's remarks stem from the center's experience with a lung cancer patient, in which a radiologist initially identified only three nodules in the lungs, while subsequent AI-assisted analysis revealed a total of 20 nodules.
That discrepancy created a clinical dilemma, forcing doctors to weigh whether to still recommend surgery based on the traditional human reading of three nodules — which academic literature suggests could offer a 25% chance of a cure — or reconsider treatment in light of the 20 nodules detected by AI, he explained.
Because historical medical research has been based entirely on human visual recognition rather than computer counts, physicians face a complex learning curve in determining how to interpret and act on machine-detected findings.
"So that's why when you look at the whole field of medicine, there are two areas which have changed and will be changed in a very big way: one is diagnostic radiology, and two is pathology," Mr. Ang said.
Both fields center on pattern recognition in images — scans in radiology, tissue slides in pathology — which is precisely the class of task where machine learning tools have advanced fastest in recent years, with regulators such as the U.S. Food and Drug Administration having cleared hundreds of AI-enabled medical devices, many of them in radiology.
Although AI in medicine is an emerging force, he said the technology can never replace humans, because healthcare also entails "the element of trust, care, and hope" — qualities he noted that only humans can provide.
Mr. Ang's statement comes as the AI in healthcare market is projected to be valued at $28.2 billion in 2026 and grow to $146.3 billion by 2031, representing an annual growth rate of 39%, according to Wissen Research, a global consulting firm. — Edg Adrian A. Eva
