In the Journals

AI colonoscopy system may detect clues physicians ‘not tuned in to recognize’

Using an artificial intelligence system for automatic polyp detection during colonoscopy helped increase adenoma detection, particularly smaller, diminutive polyps, according to study results published in Gut.

Tyler M. Berzin, MD, of Beth Israel Deaconess Medical Center, told Healio Gastroenterology and Liver Disease that the AI system (Shanghai Wision AI) can train on large image sets and its deep learning system determines a variety of visual features that help it identify when a polyp is on the screen.

“What's interesting about a deep learning system like this is that it may be recognizing features/colors/patterns that are different than the visual cues an endoscopist uses,” he said. “In other words, the AI system may recognize polyp features or clues that physicians are not tuned in to recognize at all.”

Researchers tested the system in an open, non-blinded trial in which they randomly assigned consecutive patients to undergo diagnostic colonoscopy with (n = 522) or without (n = 536) assistance of real-time automatic polyp detection. The primary outcome of the trial was adenoma detection rate.

Investigators found that the AI system significantly increased ADR (29.1% vs. 20.3%; P < .001), as well as the mean number of adenomas per patient (0.53 vs. 0.31; P < .001).

The better ADR was due, in large part, to an increase in the detection of diminutive adenomas found (185 vs. 102; P < .001). Additionally, the AI helped increase the number of hyperplastic polyps found (114 vs. 52; P < .001).

“Finding small adenomas may help identify patients who are at higher risk for adenoma formation and who need more careful surveillance,” Berzin said in an interview. “In the future, I will be particularly interested to see if AI polyp-detection systems can improve detection of sessile polyps in the right colon, which can be especially difficult to identify.” – by Alex Young

Disclosures: Berzin reports that he has received research support and is a paid consultant for Shanghai Wision AI. He reports being a paid consultant for Fujifilm Medical Systems USA. Please see the study for all other relevant financial disclosures.

Using an artificial intelligence system for automatic polyp detection during colonoscopy helped increase adenoma detection, particularly smaller, diminutive polyps, according to study results published in Gut.

Tyler M. Berzin, MD, of Beth Israel Deaconess Medical Center, told Healio Gastroenterology and Liver Disease that the AI system (Shanghai Wision AI) can train on large image sets and its deep learning system determines a variety of visual features that help it identify when a polyp is on the screen.

“What's interesting about a deep learning system like this is that it may be recognizing features/colors/patterns that are different than the visual cues an endoscopist uses,” he said. “In other words, the AI system may recognize polyp features or clues that physicians are not tuned in to recognize at all.”

Researchers tested the system in an open, non-blinded trial in which they randomly assigned consecutive patients to undergo diagnostic colonoscopy with (n = 522) or without (n = 536) assistance of real-time automatic polyp detection. The primary outcome of the trial was adenoma detection rate.

Investigators found that the AI system significantly increased ADR (29.1% vs. 20.3%; P < .001), as well as the mean number of adenomas per patient (0.53 vs. 0.31; P < .001).

The better ADR was due, in large part, to an increase in the detection of diminutive adenomas found (185 vs. 102; P < .001). Additionally, the AI helped increase the number of hyperplastic polyps found (114 vs. 52; P < .001).

“Finding small adenomas may help identify patients who are at higher risk for adenoma formation and who need more careful surveillance,” Berzin said in an interview. “In the future, I will be particularly interested to see if AI polyp-detection systems can improve detection of sessile polyps in the right colon, which can be especially difficult to identify.” – by Alex Young

Disclosures: Berzin reports that he has received research support and is a paid consultant for Shanghai Wision AI. He reports being a paid consultant for Fujifilm Medical Systems USA. Please see the study for all other relevant financial disclosures.