Autonomous Endoscope Navigation Base on Image Contour Data and Pixel Ratio

The sustainability and quality of colonoscopy procedures face substantial challenges due to physician burnout and less-than-ideal ergonomics. To mitigate these issues, our study proposes an autonomous navigation of endoscopes based on image contour data and pixel ratio. The navigation strategy has been thoughtfully divided into vertical (up-down) direction and horizontal (left-right) direction. In the vertical navigation, we first subdivide the original image into quadrants from the center. Subsequently, we design and implement a swift image preprocessing and contour extraction algorithm. The ultimate direction of vertical motion is established based on the ratio of black pixels to total pixels in each quadrant. In contrast, for the horizontal direction, we begin by rapidly extracting the intestine's contour using a specialized image preprocessing and contour extraction algorithm. Once this step is accomplished, we identify the open contour and accommodate a circle to align with this open contour. The direction of left-right turning is determined by assessing the relative position of the circle's center to the image's center. Moreover, there are instances where a clear contour cannot be discerned. For these cases, we have developed a multi-control method that enables the endoscope to capture clear contours. Our multistage adaptive control method was rigorously evaluated and validated using an intestinal model featuring a range of curved configurations. The results illustrate that our approach successfully managed turns in all directions during the intestinal model process. The accuracy of steering decisions, verified from the images of the intestinal model, consistently exceeded 90%. In conclusion, our research provides new ideas for efficient and reliable endoscopic autonomous navigation.