How One CMO is Using AI to Advance Colorectal Cancer Screening

Here's how AI tools are moving the needle in detecting colon cancer.

Artificial intelligence (AI) tools are improving the effectiveness of colonoscopies, according to the CMO of GI Alliance.

Colorectal cancer is the fourth most common form of cancer among men and women in the United States, according to the Centers for Disease Control and Prevention. It is the fourth leading cause of cancer deaths, the CDC says.

Colonoscopies are the most effective method to detect polyps in the colon that can progress into colon cancer, explains J. Casey Chapman, MD, CMO of GI Alliance, which features gastroenterology groups across the country.

"Colonoscopy is the best test we have—it is the gold standard for colorectal cancer screening as well as polyp detection," Chapman says. "There is nothing else that even comes close."

The main limitation for colonoscopies is related to the gastroenterologist workforce, according to Chapman.

"One of the limitations is that there are a lot of people who need an average-risk colonoscopy without enough people to do them," Chapman says.

With the shortage of gastroenterologists, it is imperative that colonoscopies are utilized appropriately and are done effectively, Chapman explains.

"No. 1, we have got to be sure that when we do a colonoscopy, it is sensitive to pick up polyps and colon cancer when they are present, and it is specific, meaning when colon cancer is there, we find it," Chapman says.

Polyp detection is crucial, according to Chapman.

"If you find a polyp, you want to make sure you bring that patient back for another colonoscopy at the correct interval," Chapman says. "If you don't find a polyp, you are very confident that the patient does not need to come back on a shortened interval."

J. Casey Chapman, MD, is CMO of GI Alliance. Photo courtesy of GI Alliance.

Adoption of AI tools

AI tools are making a difference in the detection of polyps in the colon, according to Chapman.

"AI has helped to start a bit of a renaissance of closing the gap for human error," Chapman says. "A gastroenterologist can miss a colon polyp. If that happens, there is an increased risk of interval colon cancer, which means you have cancer between the initial colonoscopy and the recommended second colonoscopy."

Several companies have developed computer-assisted devices that help detect colon polyps during a colonoscopy, Chapman explains.

"Basically, it is like having an extra set of eyes in the colonoscopy procedure to make sure there is fact-checking," Chapman says. "Where you see something, the AI will see it. Where you do not see something, the AI may show you there is something present."

The strength of AI in colonoscopies is going to progress as it teaches itself or as gastroenterologists teach the machine learning algorithms, according to Chapman.

"What we must do as human beings and the operator is make sure that what AI says is a colon polyp is a colon polyp, and it is getting better and better," Chapman says. "We also need to make sure when AI says there is no colon polyp there is actually no colon polyp. In other words, we still must make the ultimate decision, but AI does help to alert us that there could be danger."

The application of AI tools in colonoscopies is just beginning and exciting advances are on the horizon, Chapman explains.

"We are going to be able to utilize not only two-dimensional polyp detection as it exists today, but also move into three-dimensional spatial computing," Chapman says. "Not only are you going to see the polyp, but you are also going to be able to see it at a high image rate and in a three-dimensional fashion."

While AI has the potential to revolutionize colonoscopies, it is unlikely that AI will ever be able to conduct the procedure independently of a gastroenterologist, according to Chapman.

"The hard part would be patient buy-in. Patients probably would not want a machine doing their colonoscopy," Chapman says. "In addition, it takes about 10,000 colonoscopies to be really good at it because every colon is unique. If you are going to train AI to do colonoscopies, it is going to take hundreds of thousands of colonoscopies to train it."

Replacing the human touch in colonoscopies would be extremely difficult, Chapman explains.

"It is impossible to teach a machine trust and empathy," Chapman says. "When you are dealing with a person who might have colorectal cancer or colon polyps, the most important part of the entire process is the human-to-human relationship."