Whereas AI can supply immense potential to enhance affected person outcomes and streamline medical processes, system flaws can pose dangers to ethics, safety and knowledge accuracy.
Dr Gao Yujia, assistant group CTO of the Nationwide College Well being System, urged warning in AI implementation in healthcare within the keynote session, “Unveiling the Darkish Facet of AI,” at HIMSS24 APAC.
“Whereas AI has undoubtedly enhanced affected person care, it is essential to acknowledge the potential pitfalls,” he emphasised.
A main concern is the danger of bias in AI algorithms. Dr Gao cited an incident throughout COVID-19 the place an AI algorithm, skilled on biased knowledge, disproportionately down-prioritised folks of color who had the identical circumstances as different sufferers.
It’s important, Dr Gao pressured, that AI techniques are skilled on various and consultant datasets.
“Just like a ‘rubbish in, rubbish out’ situation, AI can solely be as correct as the info that’s fed to it. Some AI algorithms fail to carry out as anticipated as a result of they weren’t skilled on related knowledge or had been overfitted to particular datasets.”
Past knowledge high quality, Dr Gao additionally raised issues in regards to the moral implications of AI in healthcare.
“We use AI to assist us organise our ideas, however that very same intent can be utilized maliciously. For instance, AI-generated deepfakes have been used to impersonate healthcare employees for ransomware assaults. Fraudulent medical analysis can also be a risk due to how convincing AI-generated content material could be.”
Requested if all is “doom and gloom” for AI in healthcare, Dr Gao disagreed.
“A extra immersive and instantaneous digital world is inevitable, however what protects us [from AI risks] is sustaining a steadiness between human company and AI’s utility.”
Based on Dr Gao, moral frameworks, transparency and human oversight are key elements in harnessing the ability of AI whereas mitigating its potential harms.
He suggested healthcare professionals to neither worry AI nor be fearless of AI.
“We should always take a step again to know, admire, perceive and plan the best way to combine AI into healthcare in our society.”