A machine learning algorithm developed at the famous Mayo Clinic in the USA proves to be superior to doctors in detecting precancerous polyps in the large intestine.
Most polyps (lumps in the inner wall of the colon) are benign, but some develop over time into cancer of the colon or rectum.
Colorectal cancer is currently the second deadliest cancer in the world with about 1,9 million cases and nearly one million deaths by 2020, according to the World Health Organization. Cancer or precancerous lesions are detected by colonoscopy, a test in which a camera catheter examines the rectum and bowel, usually under light anesthesia.
Read also: Artificial intelligence predicts if and when one will die from opposition
The artificial intelligence (AI) system halves the cases when the doctor fails to detect existing precancerous polyps, say researchers from the Mayo Clinic in the review Gastroenterology.
Polyps in the large intestine on colonoscopy (Stephen Holland, MD, Naperville Gastroenterology)
Each of the 230 volunteers in the study underwent two colonoscopies on the same day, one in which AI was used and one without.
Previous studies have estimated that doctors fail to detect 25% of precancerous polyps. In this study, the rate was 32,5% for conventional colonoscopy.
But AI reduced it to 15%, as it located more small and flat precancerous polyps, either near the rectum or deeper in the intestine.
In addition, the percentage of false negative results, ie the cases where the test was negative while there were alterations, decreased from 29,6% to 6,8%.
"Colorectal cancer can be prevented almost completely with screening tests," said Michael Wallace, who led the study.
"The use of artificial intelligence to detect polyps in the colon is a welcome and encouraging news for patients and their families.
in.gr