3D Shape Reconstruction Algorithm for Interventional Catheter Based on HPO-BPNN

Aiming at the problems of various nonlinear characteristics caused by fiber twist in the manufacturing and packaging process of interventional catheter shape sensor, this paper proposes a fiber grating shape sensor configuration and its packaging method. The HPO-BPNN algorithm is employed to predict discrete curvature and the direction angles of curvature vectors, thus avoiding complex sensor parameter calibration steps. Additionally, the Frenet-Serret equation is integrated to achieve three-dimensional posture reconstruction of the catheter. Through the multi-curvature calibration experiment of the designed sensor, the average error at the end of the 2D curves is 21.15mm (3.91%), and for the 3D curves, it is 26.43mm (4.89%), thereby verifying the feasibility of the data-driven shape reconstruction algorithm.