2019/05/22 by Nils Gessert, Gessert, Nils, Torben Priegnitz +16
Engineering · Medicine · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Intraocular Surgery and Lenses #Optical Coherence Tomography Applications #Soft Robotics and Applications #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1905.09282
openalex publication_date 2019/05/22 · openalex created_date 2022/07/13 · openalex updated_date 2026/07/28
Purpose. Precise placement of needles is a challenge in a number of clinical\napplications such as brachytherapy or biopsy. Forces acting at the needle cause\ntissue deformation and needle deflection which in turn may lead to misplacement\nor injury. Hence, a number of approaches to estimate the forces at the needle\nhave been proposed. Yet, integrating sensors into the needle tip is challenging\nand a careful calibration is required to obtain good force estimates.\n Methods. We describe a fiber-optical needle tip force sensor design using a\nsingle OCT fiber for measurement. The fiber images the deformation of an epoxy\nlayer placed below the needle tip which results in a stream of 1D depth\nprofiles. We study different deep learning approaches to facilitate calibration\nbetween this spatio-temporal image data and the related forces. In particular,\nwe propose a novel convGRU-CNN architecture for simultaneous spatial and\ntemporal data processing.\n Results. The needle can be adapted to different operating ranges by changing\nthe stiffness of the epoxy layer. Likewise, calibration can be adapted by\ntraining the deep learning models. Our novel convGRU-CNN architecture results\nin the lowest mean absolute error of 1.59 +- 1.3 mN and a cross-correlation\ncoefficient of 0.9997, and clearly outperforms the other methods. Ex vivo\nexperiments in human prostate tissue demonstrate the needle's application.\n Conclusions. Our OCT-based fiber-optical sensor presents a viable alternative\nfor needle tip force estimation. The results indicate that the rich\nspatio-temporal information included in the stream of images showing the\ndeformation throughout the epoxy layer can be effectively used by deep learning\nmodels. Particularly, we demonstrate that the convGRU-CNN architecture performs\nfavorably, making it a promising approach for other spatio-temporal learning\nproblems.\n