Abstract: Abstract:In the practice of clinical endoscopy, the precise estimation of the lesion size is quite significant for diagnosis. In this paper, we propose a three-dimensional (3D) measurement method for binocular endoscopes based on deep learning, which can overcome the poor robustness of the traditional binocular matching algorithm in texture-less areas. A simulated binocular image dataset is created from the target 3D data obtained by a 3D scanner and the binocular camera is simulated by 3D rendering software to train a disparity estimation model for 3D measurement. The experimental results demonstrate that, compared with the traditional binocular matching algorithm, the proposed method improves the accuracy and disparity map generation speed by 48.9% and 90.5%, respectively. This can provide more accurate and reliable lesion size and improve the efficiency of endoscopic diagnosis.

基于深度学习的双目内窥镜三维测量方法

[1]Ahmad I, Levine JB, Anderson JC, 2016. Endoscopic measurement of colorectal polyps: how do we measure up? Gastroenterology, 150(3):769-771.

[2]Anderson BW, Smyrk TC, Anderson KS, et al., 2016. Endoscopic overestimation of colorectal polyp size. Gastrointest Endosc, 83(1):201-208.

[3]Cai HF, Wang R, Li Y, et al., 2018. Role of 3D reconstruction in the evaluation of patients with lower segment oesophageal cancer. J Thorac Dis, 10(7):3940-3947.

[4]Dimas G, Bianchi F, Iakovidis DK, et al., 2020. Endoscopic single-image size measurements. Meas Sci Technol, 31(7):074010.

[5]Furukawa R, Aoyama M, Hiura S, et al., 2014. Calibration of a 3D endoscopic system based on active stereo method for shape measurement of biological tissues and specimen. Proc 36^th Annual Int Conf of the IEEE Engineering in Medicine and Biology Society, p.4991-4994.

[6]Hirschmüller H, 2008. Stereo processing by semiglobal matching and mutual information. IEEE Trans Patt Anal Mach Intell, 30(2):328-341.

[7]Kendall A, Martirosyan H, Dasgupta S, et al., 2017. End-to-end learning of geometry and context for deep stereo regression. Proc IEEE Int Conf on Computer Vision, p.66-75.

[8]Khamis S, Fanello S, Rhemann C, et al., 2018. StereoNet: guided hierarchical refinement for real-time edge-aware depth prediction. https://arxiv.org/abs/1807.08865v1

[9]Nomura K, Kikuchi D, Kaise M, et al., 2019. Comparison of 3D endoscopy and conventional 2D endoscopy in gastric endoscopic submucosal dissection: an ex vivo animal study. Surg Endosc, 33(12):4164-4170.

[10]Ogino-Nishimura E, Nakagawa T, Sakamoto T, et al., 2015. Efficacy of three-dimensional endoscopy in endonasal surgery. Auris Nasus Larynx, 42(3):203-207.

[11]Omori J, Goto O, Higuchi K, et al., 2020. Three-dimensional flexible endoscopy can facilitate efficient and reliable endoscopic hand suturing: an ex-vivo study. Clin Endosc, 53(3):334-338.

[12]Sakata S, Mcivor F, Klein K, et al., 2018. Measurement of polyp size at colonoscopy: a proof-of-concept simulation study to address technology bias. Gut, 67(2):206-208.

[13]Shaw MJ, Shaukat A, 2016. Does polyp size scatter matter? Gastrointest Endosc, 83(1):209-211.

[14]Wang D, Liu H, Cheng X, 2018. A miniature binocular endoscope with local feature matching and stereo matching for 3D measurement and 3D reconstruction. Sensors, 18(7):2243.

[15]Wang XZ, Nie YF, Lu SP, et al., 2020. Deep convolutional network for stereo depth mapping in binocular endoscopy. IEEE Access, 8:73241-73249.

[16]Žbontar J, LeCun Y, 2014. Computing the stereo matching cost with a convolutional neural network. Proc IEEE Conf on Computer Vision and Pattern Recognition, p.1592-1599.

[17]Zhang FH, Prisacariu V, Yang RG, et al., 2019. GA-Net: guided aggregation net for end-to-end stereo matching. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.185-194.