Colonoscopy with intelligence

AI Instantly diagnoses colon cancer with 100% accuracy%

Sergey Kolenov, Hi-tech+

The new technique, borrowed from ophthalmology, will allow to detect this dangerous disease with absolute efficiency. Researchers suggest using it in conjunction with traditional colonoscopy.

Colonoscopy remains the most common method of diagnosing colon and rectal cancer. The technique allows you to examine the mucous membrane with a camera and take tissue samples for analysis. Unfortunately, colonoscopy has a number of limitations. It often misses small malignant formations at an early stage, and also does not allow to detect changes in the deep layers of tissue.

Researchers from Washington University in St. Louis decided to improve the methodology with the help of artificial intelligence. According to the press release of Machine learning, imaging technique may boost colon cancer diagnosis, the team took as a basis optical coherence tomography (OCT), which is usually used in ophthalmology to obtain retinal images.

OCT allows you to detect the difference in light refraction between healthy and affected tissues at a depth of several millimeters. This ensures high sensitivity to precancerous and cancerous changes. The RetinaNet neural network was used to analyze the images obtained using OCT. To train her, the team fed her 26,000 images of tissues, both healthy and malignant.

Testing has shown that the algorithm detects colon and rectal cancer with 100% accuracy. The analysis takes only four seconds. The authors of the development believe that it can be used in conjunction with colonoscopy. They are already planning to create a catheter for use in parallel with an endoscope.

Portal "Eternal youth" http://vechnayamolodost.ru