Using AI to better treat aortic stenosis

“We take images of patients who have aortic valve disease, along with additional data based on their clinical state, their demographics, their blood tests, and other information we can capture from hospital databases, and we put that into sophisticated machine learning algorithms to determine risk prediction,” says A/Prof Namasivayam.

“Our lab is looking at things like disease progression in people who have early stage disease – that is how their disease progresses over time and how that leads to remodelling of the heart that's potentially dangerous in the future.”

While there are effective therapies available for the treatment of AS - such as valve replacement, either surgically or through transcatheter approaches - one of the greatest challenges of the condition is knowing the optimal timing for these therapies.

“The current approaches to determining how severe a person's AS is are limited,” says A/Prof Namasivayam.

“Even if we use the gold standard approaches, there are a large group of patients - some studies have shown up to 50 percent - where the true severity of the disease is not clearly met on initial testing. And so that means two things happen: either we miss people and they get treated too late, or they're treated too early because we think it's more severe than it really is. This leads to bad outcomes in both directions.”

For those patients where valve replacement is being considered as a treatment option, A/Prof Namasivayam and his team have been working on an algorithm that can predict survival post replacement.

“This will be an incredibly useful tool because if we can identify the group of patients who are at greater risk of dying post replacement, we can potentially identify the factors that are causing this discrepancy and address them,” says A/Prof Namasivayam.

A/Prof Namasivayam and his team are also working on using machine learning to review imaging for AS patients in greater detail.

“We are using image-based computer vision, which is the same sort of system that self-driving cars use to recognise their environment, to read CT scans to a level of detail that we are not physically able to see,” says A/Prof Namasivayam.

“This technology reads into the deep encoding of each frame of the thousands of frames that we take per patient. Which is really exciting because it’s superior to what a human could do.”

A/Prof Namasivayam and his team set up the Institute’s Heart Valve Disease and Artificial Intelligence Lab in 2021. With early data from their research now starting to come through, they are hoping this knowledge can be put to use in clinical practice over the coming years.

“I'm quite interested in making practical solutions that clinicians can use rather than just putting out research, which is important, but as a cardiologist I also want to make something that's usable to improve outcomes for the patient,” says A/Prof Namasivayam.

“I work closely with patients who suffer from AS and I’ve seen the impact it can have on their quality of life. If we can treat these patients sooner and more effectively, we can potentially avoid adverse events and help people live longer and feel better – which is always our goal.”

While A/Prof Namasivayam can see both the positives and negatives of an AI-focused future – including the potential to address or exasperate health inequalities – his philosophy with AI is that it’s better to be the driver than the passenger.

“I think it’s important that we don’t just go along for the ride, we need to try to control this new technology - set the future, set the tone and participate. And I think we are in a unique position at the Institute to become a centre of excellence in this field over the coming years,” says A/Prof Namasivayam.