At the National University Hospital of Singapore, doctors are testing machine learning to automate the time-consuming process of characterising a thyroid lump.
Ngiam Kee Yuan, Group Chief Technology Officer of the National University Health System, has a vision to integrate AI into Singapore’s healthcare system. That will help hospitals move away from “largely reactive medicine to proactive, predictive medicine,” he says.
GovInsider spoke to him to find out how NUHS is trialling AI tools and collaborating with computer scientists to achieve that vision.
A sandbox for AI development
NUHS has been able to trial multiple AI projects thanks to a platform known as Discovery AI. It collates and aggregates massive amounts of patient data such as medical history, lifestyle habits and history of admission in hospitals.
Researchers can use this data to test their AI models in a “safe and secure way”, says Ngiam.
The data is already anonymised, so researchers don’t have to worry about privacy issues. “Almost all of our AI projects are run off the Discovery AI’s data sets,” he adds.
Clinicians and computer scientists have used data from Discovery AI to create an appendicitis diagnosis machine in the Accident and Emergency Department. When doctors input clinical observations for patients with stomach pain, the algorithm reads the text and provides a diagnosis with 90 per cent accuracy.
NUHS is currently working on a “production layer” known as Endeavour AI, says Ngiam. This allows all the AI tools on the Discovery AI platform to be produced and integrated into the healthcare system.
“Now we’re going to run it as if it’s a medical device, it’s a tool that we use on a day to day basis. So it’s not research anymore, it’s production,” he adds.
For successful AI development for healthcare, collaboration between clinicians and computer scientists is essential, says Ngiam.
If computer scientists handle the project alone, they’d make it “fantastic in a technical way” – but it may not be relevant to the clinician, he adds. On the other hand, a project managed only by clinicians will be oversimplified because “they don’t have the deep technical knowledge to synthesise data”.
Ngiam has collaborated closely with the National University of Singapore’s School of Computing on AI projects, including a model to predict the progression of kidney disease. “This will change the way we, the kidney doctors, give medication to these patients to prevent them from deteriorating their kidney function.”
NUHS also organises a datathon with NUS and the Massachusetts Institute of Technology. Data scientists and clinicians are given problem statements, data, and “two days to crunch it”, says Ngiam.
The datathon is a “trigger” for new ideas that may become actual clinical projects, he adds, and a platform for data scientists and clinicians to interact. “It’s important to put people together so that they can work in an interdisciplinary way, and you must facilitate that process.”
The future of telehealth and bots
Covid-19 has done healthcare “a massive favour” in accelerating the use of telemedicine. “Now that we have people more willing to use teleconsults, they might be willing to do more tele-other-kind-of-things,” says Ngiam.
For instance, it could improve the impact of pharmacy. Patients typically consume around 50 per cent of medication that is prescribed – hindering the effectiveness of medical care, says Ngiam. Hospitals can now look at adopting telepharmacy to stay in touch with patients to check that they’re taking the right medicines regularly.
In Singapore, the proportion of older adults with multiple chronic diseases nearly doubled from 2009 to 2017. To “flatten that curve”, hospitals need to change patient behaviours – but it can’t in a 30 minute consultation, says Ngiam.
NUHS is trialling a chatbot to follow-up on patients after consultations, he adds. The chatbot will help doctors monitor and advise patients on how they can carry out healthy behaviours at home.
While bots won’t replace doctors “anytime soon”, Ngiam believes they can be integrated into healthcare to help doctors stay in touch with patients and make care delivery more seamless.
As the cliche goes, prevention is better than cure. AI’s ability to aggregate and analyse massive data sets will be highly valuable as healthcare adapts to tackle new challenges.
Image by NUHS