MAY 22, 2022
Artificial intelligence aids doctors during colonoscopy screening
MAR 08, 2022
CHICAGO — Colon cancer is highly treatable and often curable. But while early screening is the best way to catch the disease, studies indicate that 17% to 28% of colorectal polyps—the precancerous growths that can turn into cancer—are missed during colonoscopy.
According to the CDC, colorectal cancer almost always develops from precancerous polyps in the colon or rectum, but screening beginning at age 45 or earlier can help doctors locate and remove the growths before they turn into cancer.
“With colonoscopy, we know that we can significantly decrease the incidence of colorectal cancer,” said Dr. Irving Waxman, director of the digestive diseases division at Rush University Medical Center.
Although early screening has led to a drop in deaths in the last few decades, colon cancer remains the third leading cause of cancer-related deaths in the U.S., with more than 52,000 people expected to die from it this year.
During a 30-minute procedure, screeners view about 50,000 frames of images, making it easy for a small lesion to go unnoticed.
“One polyp could potentially be missed in one or two frames. So, if I sneeze or if I'm exhausted, I could potentially miss something,” said Waxman, who specializes in interventional gastroenterology with a focus on state-of-the-art endoscopic procedures.
He and his team are among the first in the nation to implement a new innovative technology that uses artificial intelligence as a second set of eyes during colon screenings.
“So, now, it's two visual participants trying to improve the detection of these lesions,” said Waxman.
Known as the GI Genius, the deep learning technology uses a database of 13 million images to detect mucosal abnormalities in real-time that might be missed by the naked eye.
“Based on all that library, it’s trying to identify anything that looks like the image database that has been fed into it,” said Waxman.
Like a video game, the system helps a physician navigate the colon to find those smaller potentially pre-cancerous growths by overlaying a green box over difficult to see polyps.
One study found that the AI-assisted colonoscopy was 78% more likely to detect polyps 6-9 mm in size and 26% more likely to detect polyps less than 5 mm, which are some of the most difficult to detect.
“It has a huge library to take from to help you identify polyps that you may miss otherwise,” said Waxman.
During one recent procedure, the AI detected a polyp overlooked by the doctor. It was subsequently removed for biopsy.
“That's exactly what this is supposed to do,” said Waxman. “This actually increases your detection rate by 13%, so you probably have a much more significant impact.”
Experts say decreasing the miss-rate of potentially cancerous polyps will undoubtedly save even more lives.