University of Patras, Faculty of Medicine, Department of Medical Physics, Patras, Greece.
Med Eng Phys. 2011 May;33(4):479-89. doi: 10.1016/j.medengphy.2010.11.016. Epub 2010 Dec 22.
Closed intramedullary nailing is a common technique for treatment of femur and tibia fractures. The most challenging step in this procedure is the precise placement of the lateral screws that stabilize the fragmented bone. The present work concerns the development and the evaluation of a method to accurately identify in the 3D space the axes of the nail hole canals. A limited number of projection images are acquired around the leg with the help of a C-arm. On two of them, the locking hole entries are interactively selected and a rough localization of the hole axes is performed. Perpendicularly to one of them, cone-beam computed tomography (CBCT) reconstructions are produced. The accurate identification and localization of the hole axes are done by an identification of the centers of the nail holes on the tomograms and a further 3D linear regression through principal component analysis (PCA). Various feature-based approaches (RANSAC, least-square fitting, Hough transform) have been compared for best matching the contours and the centers of the holes on the tomograms. The robustness of the suggested method was investigated using simulations. Programming is done in Matlab and C++. Results obtained on synthetic data confirm very good localization accuracy - mean translational error of 0.14 mm (std=0.08 mm) and mean angular error of 0.84° (std=0.35°) at no radiation excess. Successful localization can be further used to guide a surgeon or a robot for correct drilling the bone along the nail openings.
髓内钉固定是治疗股骨和胫骨骨折的常用技术。该手术过程中最具挑战性的步骤是精确放置稳定碎骨的外侧螺钉。本工作涉及开发和评估一种在三维空间中准确识别钉孔通道轴线的方法。在 C 臂的帮助下，在腿部周围采集少量投影图像。在其中两张图像上，交互选择锁定孔入口，并进行孔轴的粗略定位。在其中一张图像的垂直方向上，生成锥束 CT（CBCT）重建。通过在断层图像上识别钉孔的中心，并通过主成分分析（PCA）进行进一步的 3D 线性回归，完成孔轴的准确识别和定位。已经比较了各种基于特征的方法（RAN-SAC、最小二乘拟合、Hough 变换），以最佳匹配断层图像上的轮廓和孔的中心。使用模拟研究了所建议方法的鲁棒性。编程使用 Matlab 和 C++完成。在合成数据上获得的结果证实了非常好的定位精度-无辐射过量时的平均平移误差为 0.14 毫米（std=0.08 毫米），平均角度误差为 0.84°（std=0.35°）。成功的定位可以进一步用于指导外科医生或机器人沿着钉孔正确钻孔。
