Design and Characterisation of Stereo Endoscope for Polyp Size Measurement

Stereo imaging in endoscopy is an advancement in medical imaging that enhances depth perception while viewing anatomical structures. Depth perception helps improve surgical navigation and diagnostic accuracy and enables polyp size measurement. Stereo endoscope systems use dual cameras to capture images from different viewpoints. Depth estimation from these images involves rectifying the images and using triangulation technique based on baseline width, disparity, and focal length. Designing stereo endoscopes presents challenges due to constraints like optimising baseline width for enhanced depth measurement accuracy, yet balancing this with maintaining an adequate field of view overlap. This is particularly critical in confined spaces such as the distal tip, where the diameter can be as small as 13 mm. In this paper, we show our strategies for addressing these design constraints and developing a stereo endoscope with a low error rate when measuring objects like polyps. We also propose an algorithm that uses the features extracted using the VGG16 model to find the corresponding point to calculate the disparity with high accuracy. We conducted extensive validation experiments to test our system and compared our proposed algorithm with the widely used Semi Global Matching (SGM) algorithm. Our designed system could synchronously acquire and process the stereo images at 60 frames per second. With a baseline width of 6.5 mm and a field of view overlapping above 80% at 30 mm depth, we could estimate the depth and measure a polyp with an error rate of less than 5%, indicating the high accuracy and reliability of our system design.