Development of a fiber-based shape sensor for navigating flexible medical tools

Robot-assisted minimally invasive surgical procedure (RAMIS) is a subfield of minimally invasive surgeries with enhanced manual dexterity, manipulability, and intraoperative image guidance. In typical robotic surgeries, it is common to use rigid instruments with functional articulating tips. However, in some operations where no adequate and direct access to target anatomies is available, continuum robots can be more practical, as they provide curvilinear and flexible access. However, their inherent deformable design makes it difficult to accurately estimate their 3D shape during the operation in real-time. Despite extensive model-based research that relies on kinematics and mechanics, accurate shape sensing of continuum robots remains challenging. The state-of-the-art tracking technologies, including optical trackers, EM tracking systems, and intraoperative imaging modalities, are also unsuitable for this task, as they all have shortcomings. Optical fiber shape sensing solutions offer various advantages compared to other tracking modalities and can provide high-resolution shape measurements in real-time. However, commercially available fiber shape sensors are expensive and have limited accuracy.
In this thesis, we propose two cost-effective fiber shape sensing solutions based on multiple single-mode fibers with FBG (fiber Bragg grating) arrays and eccentric FBGs. First, we present the fabrication and calibration process of two shape sensing prototypes based on multiple single-mode fibers with semi-rigid and super-elastic substrates. Then, we investigate the sensing mechanism of edge-FBGs, which are eccentric Bragg gratings inscribed off-axis in the fiber's core. Finally, we present a deep learning algorithm to model edge-FBG sensors that can directly predict the sensor's shape from its signal and does not require any calibration or shape reconstruction steps.
In general, depending on the target application, each of the presented fiber shape sensing solutions can be used as a suitable tracking device. The developed fiber sensor with the semi-rigid substrate has a working channel in the middle and can accurately measure small deflections with an average tip error of 2.7 mm. The super-elastic sensor is suitable for measuring medium to large deflections, where a centimeter range tip error is still acceptable. The tip error in such super-elastic sensors is higher compared to semi-rigid sensors (9.9-16.2 mm in medium and large deflections, respectively), as there is a trade-off between accuracy and flexibility in substrate-based fiber sensors. Edge-FBG sensor, as the best performing sensing mechanism among the investigated fiber shape sensors, can achieve a tip accuracy of around 2 mm in complex shapes, where the fiber is heavily deflected. The developed edge-FBG shape sensing solution can compete with the state-of-the-art distributed fiber shape sensors that cost 30 times more.