Patient Registration via Topologically Encoded Depth Projection Images in Spine Surgery

Abstract

Accurate and efficient patient registration is essential for surgical image-guidance. Here, we present a registration pipeline to establish spatial correspondence between tracked intraoperative stereovision (iSV) and preoperative computed tomography (pCT) for spine surgery. First, depth projection images encoding the common vertebral dorsal surface “height” were generated from pCT and iSV. For pCT, vertebral pose was adjusted when necessary based on anatomic landmarks. For iSV, multiple reconstructed surfaces were combined to generate a unified projection image with accounting of overlapped regions to maximize the sampling of the surgical scene. Rigid registration between the resulting projection images produced an initial alignment for refined registration using an improved iterative closest point algorithm. The technique was applied to four explanted porcine spines in a total of eight poses. Registration accuracy was assessed using bone-implanted mini screws. The average fiducial registration error and target registration error (TRE) for ground-truth probe registration was $0.50\,{\pm }\,0.08$ and $0.63\,{\pm }\,0.08$, respectively. The accuracy for iSV registration was $1.77\,{\pm }\,0.31\,\text {mm}$ in TRE and was $2.01\,{\pm }\,0.44\,\text {mm}$ for surface reconstruction. The entire registration completed within 2 min. These results suggest potential for application of the method in human patients.