Using artificial intelligence to detect colon cancer, oesophageal cancer and inflammatory bowel disease
On Tuesday, November 21st 2023, Dr. Michael Byrne, Clinical Professor of Medicine in the Division of Gastroenterology, and Director of the Interventional Endoscopy Fellowship programme at Vancouver General Hospital/University of British Columbia, and CEO/founder of Satisfai Health, will speak to host Prof. Patricia Maguire about using artificial intelligence and allied technologies to deliver precision endoscopy and imaging to gastroenterology on our 15-minute Human Health Impact Technology webinar series (#HHITseries).
As CEO of a medical AI start-up, Dr Michael Byrne regularly finds himself trying to convince a roomful of fellow physicians that artificial intelligence tools will potentially make them better at detecting precancerous polyps in a patient’s colon.
“There is less pushback now than there used to be,” he says. “But you still get people saying, ‘You know what? I find polyps easily. I don’t need a beeping machine in the corner or something on a screen flashing and telling me to look there. There’s already enough distraction and noise in the room and I don’t need any more, thank you.’”
Dr Byrne’s response to this kind of challenge is “pretty black and white”, he says.
“If your cancerous polyp detection rate is 70%, no problem, you're right up there with the best performers on the planet, and I would send a family member to you for that procedure. However, if your so-called adenoma detection rate (ADR) is 35%, 30% or a good bit lower, which is incredibly common, and you think you don’t need help, you’ve got your head in the sand.”
ADR measures the average rate of precancerous polyps that a doctor identifies and removes during colonoscopy procedures.
“There are studies that show that the ADR can be increased by anything between 8-10% if you use a machine to help you,” says Dr Byrne. “That's pretty notable, because another study has shown that for every 1% increase in that ADR, you'll decrease colon cancer mortality by 3%.”
Colon cancer is the second leading cause of cancer-related deaths worldwide. In 2020, more than 1.9 million new cases were diagnosed and some 930,000 died from the disease.
Dr Byrne’s company, Satisfai Health, specialises in delivering AI and computer vision solutions in gastroenterology to tackle these alarming statistics. Also, Satisfai aims to improve outcomes for the estimated 10 million people globally who suffer with inflammatory bowel disease, and patients who are at risk for development of oesophageal cancer which is rapidly increasing in incidence.
“We aim to address all the major diseases across the entire GI tract, from cancers to inflammatory diseases like Crohn’s and colitis,” he says.
Satisfai’s solutions focus mostly on GI endoscopy, the procedure for which the inside of your body is examined using an endoscope, or long, thin flexible tube with a light source and camera at one end. The endoscope is put through the mouth or the colon to examine the GI tract, with the images relayed on a computer screen. Computer vision is a field of AI that enables computers to derive information from such images.
“The human eye obviously picks things up and you can do what’s called an optical biopsy, which is an instant virtual biopsy with your eyes. You might say, ‘Oh, I think that lesion on the screen is cancer or I think it’s not cancer’, whatever the case may be. The accuracy of the human eye in general is not that great for optical biopsy, so we rely on physical biopsy instead.”
Global experts in the field of gastroenterology can perform optical biopsies with a reasonable degree of accuracy.
“But that does not apply to most people,” says Dr Byrne. “For optical biopsy, we're asked to look for three things: colour, shape of the pits and the level of vascular change. By contrast, Satisfai’s tool is seeing at least one thousand things per polyp to make that differentiation between precancerous or not. That highlights the difference between the human eye and the computer eye. The machine sees things that the human eye cannot see, or even conceive of.”
Another more obvious advantage of machines is that unlike human beings, especially very busy clinicians, they don’t tire.
“The research suggests that the detection rate performance for GI disease goes down in the afternoon when humans get tired,” confirms Dr Byrne.
Now he is building a platform with an array of tools to support doctors throughout the endoscopy process.
“These tools finds polyps, perform a characterisation or an optical biopsy, assess the bowl preparation, and tell you when you’ve reached the end of the colon - a key quality metric. They also automatically time the withdrawal of the scope because if we are slower we find things better. Our tools can also guide doctors where to perform biopsies in suspected early cancer of the oesophagus, at a stage when it is curable. We also have solutions that can improve the accuracy of assessment of disease activity in Crohn’s and colitis, with implications for clinical trials and for day-to-day practice in terms of treatment decisions with expensive medications.”
He describes Satisfai’s AI detection and characterisation tools as being “an expert on your shoulder, a second opinion saying, ‘You should look there or biopsy there because I see something that you just missed.’”
Such AI tools are proving to be highly precise.
“The tools that groups like ours and others are developing are in the range of 90%+ accurate. They are as good as, if not better than, the human expert and the current gold standard. Conforming to the high standards that we've set, you're going to see a slew of solutions over the next few years in the gastroenterology space.”
Register to see Dr Michael Byrne on the UCD Human Health Impact and Technology Series.
Source: https://www.ucd.ie/discovery/newsandevents/storiesofdiscovery/mikebyrneaiforgi/
