AI Initiative aims to use existing data to revolutionise ICU care
A research initiative aiming to leverage Artificial Intelligence’s capabilities to detect early signs of patient deterioration, especially in those undergoing mechanical ventilation, is receiving significant backing from the Wesley Research Institute.
At the helm of this innovative project are Dr James Winearls and CCRG’s Professor John Fraser, who are working with leaders in Machine Learning (ML) and AI to create an early warning system using AI and ML assessment of every second of data from life support machines.
“The hundreds of machines and monitors in the ICU continuously giving us vast amounts of data but we only really use a fraction of it. The doctors will go home at night – sometimes, but the machines are there 24/7 – they don’t have days off and they don’t get sick. We need to harness the power of all this data.
“This project aims to create a state-of-the-art early warning system that can pick up problems before they develop by harnessing the power of extraordinary amounts of data available in the ICU, and in doing so improve the care and recovery of our patients,” said Professor Fraser.
Highlighting the volume of data, Professor Fraser highlighted the incredible number of alarms that staff and patients face daily.
“In one month of analysis, we saw 600,000 alarms – that’s 600,000, beeps and bongs!”
The relentless noise not only adds stress to an already tense environment but also increases the risk of overlooking genuine medical alerts.
Professor Fraser and Dr Winearls’ project aims to use AI and ML to filter through the noise, identifying significant changes in patient conditions that require immediate attention.
“AI has the remarkable ability to identify slight changes,” said Professor Fraser.
“There’s a huge amount of data coming at us as doctors, allied health and nursing stuff. But the question is, how well are we using it? Computers can look at subtle, subtle changes and see them much earlier than we can.
“So, for example, a ventilator may show subtle signs that the flow of oxygen has changed if there is a build up of sputum in the breathing tube, but unless a clinician recognises the pattern, they might miss the early signs.
“We want to teach the computer to learn tiny, very early changes, because if we don’t spot that, until it’s completely blocked, a patient can die.
“So, this is an early warning system, using data using waveforms that’s already there. And all we’re doing is teaching the existing computers and the existing software to notice a pattern way before our human eye can. And that makes your intensive care experience safer.”
Professor Fraser said he and Dr Winearls were undertaking ‘real-life research’ that would have direct impact on patients.
“We don’t want this to be a manuscript that we get published and someone says we’re clever. Tomorrow’s treatment grows from today’s research. So that’s what we aim to do - translate, translate, translate.”
Dr James Winearls