Dobbe J G G, Curnier F, Rondeau X, Streekstra G J
Department of Biomedical Engineering & Physics, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands.
Digisens R&D, Le Bourget-du-Lac, France.
Med Eng Phys. 2015 Jun;37(6):524-30. doi: 10.1016/j.medengphy.2015.03.008. Epub 2015 Apr 20.
Navigation for corrective osteotomy surgery requires patient-to-image registration. When registration is based on intraoperative 3-D cone-beam CT (CBCT) imaging, metal landmarks may be used that deteriorate image quality. This study investigates whether metal artifacts influence the precision of image-to-patient registration, either with or without intermediate user intervention during the registration procedure, in an application for corrective osteotomy of the distal radius. A series of 3-D CBCT scans is made of a cadaver arm with and without metal landmarks. Metal artifact reduction (MAR) based on inpainting techniques is used to improve 3-D CBCT images hampered by metal artifacts. This provides three sets of images (with metal, with MAR, and without metal), which enable investigating the differences in precision of intraoperative registration. Gray-level based point-to-image registration showed a better correlation coefficient if intraoperative images with MAR are used, indicating a better image similarity. The precision of registration without intermediate user intervention during the registration procedure, expressed as the residual angulation and displacement error after repetitive registration was very low and showed no improvement when MAR was used. By adding intermediate user intervention to the registration procedure however, precision was very high but was not affected by the presence of metal artifacts in the specific application.
矫正截骨手术的导航需要患者与图像配准。当基于术中三维锥形束CT（CBCT）成像进行配准时，可以使用金属标记物，但这会降低图像质量。本研究调查了在桡骨远端矫正截骨术的应用中，金属伪影是否会影响图像与患者配准的精度，无论在配准过程中是否有中间用户干预。对一具带有和不带有金属标记物的尸体手臂进行了一系列三维CBCT扫描。基于图像修复技术的金属伪影减少（MAR）用于改善受金属伪影影响的三维CBCT图像。这提供了三组图像（有金属、有MAR和无金属），从而能够研究术中配准精度的差异。基于灰度的点到图像配准在使用带有MAR的术中图像时显示出更好的相关系数，表明图像相似性更好。在配准过程中没有中间用户干预时的配准精度，以重复配准后的残余角度和位移误差表示，非常低，并且使用MAR时没有改善。然而，通过在配准过程中增加中间用户干预，精度非常高，但在特定应用中不受金属伪影存在的影响。
