We present a prototype of a flexible nitinol needle (φ 1.0 mm and length 172 mm) integrated with an array of 12 Fiber Bragg Grating (FBG) sensors. These sensors measure the axial strain, which enables the computation of the needle curvature. We reconstruct the three-dimensional (3-D) needle shape from the curvature. Experiments are performed where the needle is deflected in free space. The maximum errors between the experiments and beam theory-based model are 0.20 mm (in-plane deflection with single bend), 0.51 mm (in-plane deflection with double bend), and 1.66 mm (out-of-plane). We also describe kinematics-based and mechanics-based models for predicting the 3-D needle shape during insertion into soft tissue. We perform experiments where the needle is inserted into a soft-tissue simulant, and the 3-D needle shape is reconstructed using the FBG sensors. We compare the reconstructed needle shape to deflection obtained from camera images and our models. The maximum error between the experiments and the camera images is 0.74 mm. The maximum errors between the kinematics-based and mechanics-based models and the camera images are 3.77 mm and 2.20 mm, respectively. This study demonstrates that deflection models and needles integrated with FBG sensors have the potential to be used in combination with clinical imaging modalities in order to enable accurate needle steering.

Needle insertion procedures are commonly used for surgical interventions. In this paper, we develop a threedimensional (3D) closed-loop control algorithm to robotically steer flexible needles with an asymmetric tip towards a target in a soft-tissue phantom. Twelve Fiber Bragg Grating (FBG) sensors are embedded on the needle shaft. FBG sensors measure the strain applied on the needle during insertion. A method is developed to reconstruct the needle shape using the strain data obtained from the FBG sensors. Four experimental cases are conducted to validate the reconstruction method (singlebend, double-bend, 3D double-bend and drilling insertions). In the experiments, the needle is inserted 120 mm into a softtissue phantom. Camera images are used as a reference for the reconstruction experiments. The results show that the mean needle tip accuracy of the reconstruction method is 1.8 mm. The reconstructed needle shape is used as feedback for the steering algorithm. The steering algorithm estimates the region that the needle can reach during insertion, and controls the needle to keep the target in this region. Steering experiments are performed for 110 mm insertion, and the mean targeting accuracy is 1.3 mm. The results demonstrate the capability of using FBG sensors to robotically steer needles.