Researchers from the University of Occupational and Environmental Health in Japan have developed the Autonomous Colonoscope Robot System (ACRS), an AI-powered robotic platform capable of performing fully automated colonoscope insertion under controlled laboratory conditions. Built on the Endoscopic Operation Robot (EOR) version 4, the system was trained using insertion data from an expert endoscopist and combines computer vision with robotic control to autonomously navigate a colonoscopy training model.
In testing, the ACRS successfully completed fully autonomous (Level 4) insertion in 62 of 72 evaluable procedures (86.1%), with an average cecal intubation time of 2.9 minutes. Although slower than an expert endoscopist (1.4 minutes), the robot’s performance was comparable to that of gastroenterology trainees, demonstrating that AI can reproduce many expert insertion techniques. Most instances requiring human intervention occurred in the technically challenging sigmoid colon, highlighting where further refinement is needed.