Colonoscopy AI Saves Lives

While many people know to get their colon screened when they turn forty-five, most aren't aware that the rates at which polyps are missed varies among gastroenterologists from six to forty-one percent. That's not good since precancerous polyps are what lead to colon cancer, the third deadliest cancer in the US. But a new study shows when artificial intelligence is used in colonoscopies, it cuts the number of polyps missed by a whopping thirty percent. That's significant when you consider that a screening without AI already cuts a person's risk of dying from colon cancer by sixty percent.

The study involved two hundred patients and compared the standard white light colonoscopy with one using deep-learning based computer-aided detection system. One group got the standard colonoscopy first immediately followed by the computer aided colonoscopy while the other group had the testing order reversed. The same doctor performed both. Patients who underwent the computer aided colonoscopy first had a miss rate of about twenty percent while those who first underwent standard white light colonoscopy had a miss rate of thirty four percent.

AI was profoundly better at missing fewer polyps and it's expected to be even better as the machine learning improves. Deep learning artificial intelligence is just beginning to be applied to diagnostic testing in medicine but its role and importance can't be overstated.