CAD-assisted colonoscopy and advanced adenoma

October 10, 2023written by Sharon Reynolds

Artificial intelligence (AI) is suddenly everywhere. From ChatGPT to automated customer service “agents,” computers’ ability to perform increasingly complex tasks is transforming society.

AI is also rapidly making inroads into healthcare. One area that is widely tested is cancer screening, helping to interpret image-related tests such as mammography for breast cancer and colonoscopy for colorectal cancer. But are you ready to take on a leading role?

When it comes to screening for colorectal cancer, the answer appears to be “not yet,” according to two new studies.

One of the studies was the largest clinical trial ever to evaluate whether an AI-based technology called computer-aided detection (CAD) can improve colonoscopies. Studies have shown that with the use of CAD: Detection rate of advanced adenomas was not increasedAn experienced doctor will diagnose the tumor that is most likely to become cancerous.

Incorporating CAD into the screening procedure did increase the total number of adenomas detected, but that increase was driven by the detection of much smaller tumors that were least likely to become cancer.

The value of finding and removing these small tumors, also called polyps, remains controversial. But researchers led by Rodrigo Jobar, MD, PhD, of Spain’s General Hospital, wrote that removing them poses small but real risks, including damage to the colon.

Their study dated August 29th Annual report of internal medicine.

Another study in the same issue reported similar results from an analysis of 21 previous clinical trials that tested the addition of CAD to standard colonoscopy. Overall, the polyp detection rate improved with the addition of AI; This was caused by increased detection of small polypslarger, less threatening growth.

The AI-based CAD used in these studies was not as advanced as the AI ​​used to power new tools such as ChatGPT, explained Dr. Asad Umar, DVM at NCI. Cancer prevention departmentwas not involved in the new study.

The systems used in these trials are “still kind of primitive AI, not like what we’re seeing in 2022 or 2023,” Dr. Umar said. He continued that cutting-edge AI technology, if integrated into his CAD system, could help improve colorectal cancer screening.

“The next version is [of CAD] It’s likely to be much better,” he said.

Will AI improve the performance of experienced physicians?

During a colonoscopy, a long, thin tube called a colonoscope with a high-definition camera attached to the end is introduced through the anus, into the rectum, and throughout the colon. Physicians who perform the procedure do not work alone, explains Dennis Shun, MD, of Yale University, writing: Editorial accompanying new study.

“If I was pulling out the colonoscope during surgery and the nurse in the endoscopy room said, ‘I think I saw something,’ I would go back and look every time.” he says.

Over the past decade, AI-based computer systems have been designed to serve gastroenterologists and doctors performing colonoscopies alike.

These systems are based on software that scans the tissue inside the colon as it snakes through the colon. CAD software is “trained” on millions of images obtained during colonoscopies and can potentially recognize important changes that the human eye might miss. When the algorithm detects suspicious tissue, such as a polyp, the area lights up on the computer screen and makes a sound to alert the colonoscopy team.

Previous studies of these systems have provided conflicting evidence as to whether they improve endoscopist performance. Additionally, previous clinical trials were not large enough to determine whether there was a meaningful difference in detecting advanced adenomas.

Find small polyps that are not large.

To fill this data gap, Spanish researchers launched the CADILLAC trial in 2021. Researchers from six major colonoscopy centers enrolled more than 3,000 people who had blood in their stools on a fecal immunochemical test (FIT). People with a positive FIT test are recommended to have a follow-up colonoscopy because they are at increased risk for advanced adenomas and colorectal cancer.

Participants were randomly assigned to undergo either a standard colonoscopy (control group) or a colonoscopy using CAD. A total of 64 endoscopists from six centers participated in the study.

Overall, approximately 34% of participants in both groups were found to have advanced adenomas or colorectal cancer.

These results differ from those of a previous small study that found that adding CAD to surgery increased detection of advanced adenomas and colorectal cancer.

This difference may be due, at least in part, to the greater experience of the participating endoscopists and the correspondingly higher detection rate of advanced adenomas in the control group, the researchers wrote. .

“In lower situations, [adenoma detection rate] Alternatively, in a group of endoscopists considered to be ‘low detectors,’ computer-assisted detection systems may be even more effective,” the researchers wrote.

Computer-assisted detection has increased the number of small polyps less than 5 millimeters in diameter removed, even by experienced physicians.

Similar results were found in a second study conducted by a European research team that analyzed the results of 21 previous randomized clinical trials that added CAD to colonoscopy. The study, called a meta-analysis, included data from more than 18,000 people who took part in earlier trials.

Similar to the CADILLAC study, the meta-analysis showed that adding CAD increased detection of small polyps (in this case by more than 50%) but not advanced adenomas.

Polyps found during a colonoscopy are removed during surgery. So with the increased detection of small polyps, Dr. Umar explained, “we may be over-removing something that doesn’t necessarily need to be removed.” This is a phenomenon called overtreatment.

Whenever you remove a polyp, he continued, “you’re taking the mass out of the colon, so there’s a risk of perforation and bleeding.” “It’s rare, but it does happen.”

Future team member?

Dr. Umar explained that it was somewhat disappointing that CAD did not improve detection of advanced adenomas in these studies, as there is a need to further develop technologies that can speed up and streamline colonoscopy screening workflows. .

Currently, the number of people undergoing colonoscopies is far lower than what national guidelines recommend. “But we don’t have enough gastroenterologists to perform as many colonoscopies as we need,” he said. In some areas, such as some rural areas, a lack of trained professionals is already impacting people’s ability to get tested in a timely manner.

If AI-based systems can eventually speed up the colonoscopy process, it could help fill the workforce shortage in the field, Dr. Umar said.

He explained that in current randomized trials, adding computer-assisted detection does not slow down the process and is an important first step.

Umar added that the technology could eventually be useful in similar procedures, such as testing for early signs of esophageal cancer in people with a condition called Barrett’s esophagus.

Precancerous areas of the esophagus tend to be flat, making them more difficult to identify than tumors of the colon or rectum. “But it’s the same problem everywhere [in the GI tract]: How do we identify lesions that develop into cancer?” he said.

For now, AI systems may play a role in training much-needed new endoscopists, but their inexperience will naturally lead to them missing more important polyps. Dr. Umar explained.

A truly revolutionary advance is a computer system that can provide some real-time certainty about whether a polyp identified during a colonoscopy is potentially dangerous, Dr. Shun said. . Currently, tissue removed during surgery must be sent to a pathologist for analysis, which can take several days.

“I have an algorithm on my wish list that detects all my polyps and tells me if they are cancerous or not. [in real time]. …then you can spend more time with your patients and advise them on what needs to be done [next]” said Dr. Shun.

While such systems won’t be introduced into clinics anytime soon, companies that make colonoscopy machines are already starting to incorporate AI directly into their systems, Dr. Shun added.

“the [eventually] Endoscopes will be directly mounted. I think that will be the future,” he said.

“people [tend to be] They worry that AI will take their jobs,” Dr. Shun added. “But I think of AI as a member of the team, as a complement. [eventually free up time] “It’s about having humans take care of humans,” he said.