A Medical Liquid Varifocal Endoscope for Abdominal Cavity and Its Parallax Estimation Algorithm Compatible With WBANs

Abstract

Due to the rapid development of science and technology over recent years, such as Virtual Reality (VR) and low power wireless transmission, electronic image endoscope has gained an important role in the early disease diagnosis as one of the most significant medical sensors. In this paper, we introduced a varifocal medical endoscope with a wide field of view for VR-assisted diagnosis. Inspired by the artificial compound eye, a multi-channel probe was designed to establish wide FOV imaging. A voltage was applied for the realization of the varifocal lens so as to reduce the gap between membrane and substrate, and a filled conductive was squeezed so as to change the curvature of the membrane. Besides, we proposed an improved SIFT algorithm by introducing a secondary projection, while the algorithm’s feasibility was quantitatively verified. Moreover, an SSIM index of 0.945, K-Blur index of 0.8927, and PSNR value of 29.271 were obtained over a test set of images, which were used to verify the algorithm’s effectiveness. At the same time, we proposed the parallax estimation algorithm of binocular endoscopic images based on HS-ResNet, thus achieving the three-dimensional display of endoscopic image parallax. Mean SSIM index of 0.945 and Mean PSNR value of 29.271 were obtained over a test set of images, which proved that the algorithm had good robustness. Zoom Medical Endoscope 3D reconstruction algorithm of endoscopic images can obtain real-time pathological information and transmit it to the medical terminal via Wireless Body Area Network (WBAN).