Key points are not available for this paper at this time.
Real-time endoscopic navigation requires accurate, radiation-free three-dimensional (3D) reconstruction of the shape of the endoscope insertion segment to reduce procedural risks and improve operational consistency. Wavelength-division-multiplexed fiber-optic shape sensing achieves high signal-to-noise ratio and robust real-time performance, yet practical accuracy is often limited by external random torsion, packaging-induced geometric parameter deviations, and the absence of reliable insertion segment length estimation. In addition, spectral overlap between adjacent fiber Bragg gratings (FBGs) can lead to demodulation failure under large curvatures. Here, we develop a systematic 3D endoscopic navigation framework using a multicore dense fiber Bragg grating shape sensor. A torsion-elimination sensor package combined with a twist-compensation and parameter-calibration method markedly improves curvature and bending-angle estimation, reducing the corresponding errors from 0.809 to 0.327 m −1 and from 22.67 to 0.123 rad, respectively. As a result, the maximum shape reconstruction relative error of two-dimensional and 3D shape reconstruction decreases from 4.15% and 8.53% to 0.84% and 2.67%. To support robust navigation during insertion, we further propose an insertion-segment-matching algorithm to estimate insertion length and localize the dynamic tip zero point, which is validated in colorectal phantom experiments to achieve stable full-length navigation. Moreover, a spectral-shape-matching algorithm is introduced to resolve spectral overlap among adjacent FBGs, extending the measurable curvature range by approximately 3.4× while maintaining a curvature measurement relative error of 5.20%. These results demonstrate a practical route to improving the accuracy and robustness of fiber-optic endoscopic navigation and highlight the potential to enhance the safety and consistency of minimally invasive procedures.
Zhao et al. (Fri,) studied this question.