Toward an AI-embedded capsule endoscopy system for medical healthcare

Automated computer-aided techniques for detecting colorectal polyps in capsule endoscope images present a promising avenue for enhancing the post-processing phase of endoscopy, thereby assisting healthcare professionals with increased accuracy. A notable challenge lies in the absence of precise localization information, hindering physicians' ability to precisely identify the location of detected polyps for subsequent procedures such as screening or removal. In this research, we present an innovative electronic system for capsule endoscope localization, along with an AI-powered recognition model, aimed at integrating into an active locomotion system for the automatic detection and localization of colorectal polyps. The active capsule endoscope system consists of a magnetic capsule endoscope (MCE) and a magnetic actuation device capable of controlling the MCE's movement in five degrees of freedom. A deep learning model is employed for automatic detection of colorectal polyps with high precision. The localization device incorporates a 2D array of magnetic sensors and acquisition electronic circuits to measure the magnetic signal emitted by the MCE. The capsule's precise position in five degrees of freedom is determined using an optimization algorithm. Upon polyp detection, the MCE’s location is recorded alongside the captured image of the detected polyp, facilitating historical analysis and aiding in subsequent examinations. While this study primarily focuses on extracting information related to colorectal polyps, the proposed method has the potential to be expanded for detecting abnormalities throughout the entire digestive tract.