LONDON, Jan 18 — Scanning a heart for coronary artery disease.
Doctors misdiagnose the results of this procedure — on average — every one in five cases — but not here.
This scan will be analysed using artificial intelligence — with remarkable results.
“Cardiologists on average get it right about eighty per cent of the time. Some people do very well, they do better than that. Some people do less than better than that. The same person who does very well one day may not do quite so well the next day. I think what we can achieve with this system is we can standardise and get it accurate every time rather than relying on that day-to-day variability,” professor of cardiovascular medicine at the University of Oxford Paul Leeson said.
A University of Oxford spin-off, the Ultromics system breaks down the scan into tens of thousands of tiny measurements to produce a result in seconds.
“What we thought would be sort of 10 to 20 measurements which is currently what the clinician looks for in an image, turned into eighty thousand measurements. So we were able to extract so many more measurements than we ever thought possible really,” Ross Upton, CEO and founder of Ultromics said.
The system learned from Professor Leeson’s archive of scans and patient outcomes — and keeps learning.
“The consequences of getting the diagnosis wrong is that you either send the patient through to an invasive surgical procedure that they don’t need or, which is even worse, you send the patients home with coronary artery disease and those patients you send home will have a heart attack out of hospital when they should have gone through and had surgery there and then and that improvement in accuracy can help stop this thing so it’s critical for the patient,” Upton said.
Results of a trial underway at six British hospitals will be published later this year — and are expected to radically improved diagnosis rates. — Reuters
