An Incremental Registration Method for Endoscopic Sinus and Skull Base Surgery Navigation: from Phantom Study to Clinical Trials

Wenjie Li,Jingfan Fan,Shaowen Li,Zhao Zheng,Zhaorui Tian,Danni Ai,Hong Song,Xiaohong Chen,Jian Yang

DOI: https://doi.org/10.1002/mp.15941

IF: 4.506

2023-01-01

Medical Physics

Abstract:Purpose Surface-based image-to-patient registration in current surgical navigation is mainly achieved by a 3D scanner, which has several limitations in clinical practice such as uncontrollable scanning range, complicated operation, and even high failure rate. An accurate, robust, and easy-to-perform image-to-patient registration method is urgently required. Methods An incremental point cloud registration method was proposed for surface-based image-to-patient registration. The point cloud in image space was extracted from the computed tomography (CT) image, and a template matching method was applied to remove the redundant points. The corresponding point cloud in patient space was incrementally collected by an optically tracked pointer, while the nearest point distance (NPD) constraint was applied to ensure the uniformity of the collected points. A coarse-to-fine registration method under the constraints of coverage ratio (CR) and outliers ratio (OR) was then proposed to obtain the optimal rigid transformation from image to patient space. The proposed method was integrated in the recently developed endoscopic navigation system, and phantom study and clinical trials were conducted to evaluate the performance of the proposed method. Results The results of the phantom study revealed that the proposed constraints greatly improved the accuracy and robustness of registration. The comparative experimental results revealed that the proposed registration method significantly outperform the scanner-based method, and achieved comparable accuracy to the fiducial-based method. In the clinical trials, the average registration duration was 1.24 +/- 0.43 min, the target registration error (TRE) of 294 marker points (59 patients) was 1.25 +/- 0.40 mm, and the lower 97.5% confidence limit of the success rate of positioning marker points exceeds the expected value (97.56% vs. 95.00%), revealed that the accuracy of the proposed method significantly met the clinical requirements (TRE <= 2 mm, p < 0.05). Conclusions The proposed method has both the advantages of high accuracy and convenience, which were absent in the scanner-based method and the fiducial-based method. Our findings will help improve the quality of endoscopic sinus and skull base surgery.