CHICAGO — Artificial intelligence is infiltrating our world and now Chicago doctors say they’re using it to detect the earliest signs of colon cancer.

In routine colonoscopies at Northwestern Medicine, gastroenterologist Dr Rajesh Keswani navigates the organ looking for suspicious spots.

Some are can be very subtle and difficult to catch.

During a recent colonoscopy, an area caught the attention of the artificial intelligence system that was in use during the procedure.

It was an adenoma.

“It’s a classic colon polyp that can turn into cancer over time,” Keswani said.

A trained human eye makes the ultimate call.

Looking at the results of more than 4800 colonoscopies, Northwestern Medicine gastroenterologists detected and removed 13% more polyps – the precursor to colon cancer – when they used AI during screenings.

“It’s there always helping us,” Keswani said. “Some people liken it to being another colleague in the room or a mentor telling you where you might need to look and pay more attention.”

The hospital system started using the assistive technology full-time two weeks ago.

“If you’re doing this all day everyday you have to be vigilant the whole time and not miss things,” Keswani  said. “That’s why it’s helpful to have some support from AI.”

But is there a risk of over diagnosis and unnecessary removal each time the AI platform highlights an area?

“AI is really good, but it’s not perfect just like humans are not perfect,” Keswani said. “We as humans need to continue to be active participants of these procedures. And when we see a polyp, if the AI says ‘This is a polyp,’ I need to make an intelligent decision, ‘Is this a pre-cancerous polyp or not?’ Some of these polyps are just benign polyps that have no potential for cancer, and then I won’t do anything with them.”

Keswani says the technology will make an even bigger impact in underserved areas.

“For every one percent increase in detection of colon polyps, your patients are three percent less likely to develop colon cancer in the future,” he said. “There’s really no limit to where AI can help us in these regards.”

Current U.S. versions of AI don’t distinguish between cancerous and non-cancerous polyps. That skill is being developed for future generations of the technology.

For Medical Watch stories and headlines delivered to your inbox, sign up for the Medical Watch newsletter here.