Forum: Ensure AI tools are tailored to meet Singapore’s healthcare needs
Sign up now: Get ST's newsletters delivered to your inbox
Follow topic:
I refer to the article “ AI to help doctors predict disease risk in next phase of Healthier SG: Ong Ye Kung
However, as a practising gastroenterologist who has seen many new AI products emerging on the scene, I have some concerns.
First, many AI tools are developed using data from outside Singapore, and may not be directly applicable to the local population. For example, diagnostic AI trained on images of skin lesions from Caucasian patients may not be suitable for Asian skin tones.
Before relying on such tools, it is essential to evaluate and validate them using local data to ensure their accuracy and relevance.
Second, many AI tools often perform the same function, but few head-to-head comparisons are available to assess which is most effective or cost-efficient. In my own experience with three endoscopy AI tools, their performance was similar, making it hard to identify a clear front runner.
Third, AI tools are costly and can drive up healthcare expenses. For example, one such tool used to improve polyp detection raises endoscopy fees by about 15 per cent. Hospitals must weigh such purchases carefully, as the cost ultimately falls on patients.
No AI product is perfect, and developers typically do not take responsibility for errors. Ultimately, it is the doctor who must assess the AI’s recommendations and bear full responsibility for the final decision.
We need locally made, locally validated systems. If they are subsidised or supported with government grants during their development phase, they should be made available at a lower cost for our healthcare sector.
Desmond Wai (Dr)
