In conducting clinical studies with several hospitals in Australia and Singapore, the See-Mode team gained access to thousands of anonymised scans of plaque build-ups in carotid arteries, including plaques that went on to cause a stroke, and those that have not proven harmful.
See-Mode has trained a machine learning algorithm to recognise the difference between dangerous and benign plaques on an ultrasound image.
“This is something that a human being, no matter how experienced, just cannot see. In one of our clinical studies the human expert accuracy in classifying these dangerous plaques was less than 50 per cent,” Dr Mohammadzadeh said.
Web application
An ultrasound to the human eye is a “black-and-white, noisy image”, however artificial intelligence is able to establish relationships between colour intensity, and the way the pixels in a plaque are connected, which are “hard for human beings to recognise, let alone analyse.”
A web application for classifying plaques in ultrasounds will be See-Mode’s first product. After validation by clinical studies, the start-up is now seeking Food & Drug Administration approval to begin screens on real patients.
A future product will analyse computed tomography (CT) scans or magnetic resonance images (MRIs) to detect blood flows in the neck and brain that risk causing a haemorrhage, using three-dimensional reconstructions to simulate flows from patients that went on to suffer a stroke, as well as healthy flows, and training an algorithm to spot the difference.
While everyone can reduce their risk of stroke by leading a healthier lifestyle, Dr Mohammadzadeh hoped See-Mode would allow clinicians to spot warning signs of an imminent stroke and prevent it, perhaps by prescribing blood-thinning medication, or in extreme cases surgery to scrape plaques out of arteries.
“Currently the decision-making process around whether to refer someone for surgery is based on clinical studies that are 20 or 30 years old,” he said.
“It is horrifying to think doctors are guessing on medical images, when algorithms built on top of the latest clinical evidence allows them to make decisions much more accurately and systematically.”
See-Mode has received funding of $1 million from Australia’s Blackbird Ventures, and Singapore’s SGInnovate and Cocoon Capital, to take it through clinical trials and commercial release of the plaque-classifying product, which Dr Mohammadzadeh hoped would occur before the end of 2019.
from Just News Viral https://ift.tt/2IB7by3
0 Comments