Russia Registers AI System for Detecting Abnormalities During Colonoscopy
The system was developed by Eva Lab.
Roszdravnadzor has registered the Russian Polyptron.AI system developed by Eva Lab. It is designed to detect neoplasms during endoscopic diagnostics of the gastrointestinal tract.
The platform is based on artificial intelligence technologies. During diagnostics, the system scans the video stream from the endoscope. The algorithm automatically detects neoplasms, highlights them with an outline or frame, determines their size category (up to 10 mm or more than 11 mm) and displays the results directly on the doctor's screen. After the examination, the system saves the video recording with all markings and generates a report.
The developers note that the platform is based on the YOLO computer vision neural network algorithm. The model was trained on a dataset of 46,103 annotated images obtained from 1,873 unique colonoscopy examinations, including 27,392 abnormalities annotated by endoscopists.
