Selection of massive bone allografts using shape-matching 3-dimensional registration

Abstract

Background and purpose Massive bone allografts are used when surgery causes large segmental defects. Shape-matching is the primary criterion for selection of an allograft. The current selection method, based on 2-dimensional template comparison, is inefficient for 3-dimensional complex bones. We have analyzed a 3-dimensional (3-D) registration method to match the anatomy of the allograft with that of the recipient.

Methods 3-D CT-based registration was performed to match the shapes of both bones. We used the registration to align the allograft volume onto the recipient's bone. Hemipelvic allograft selection was tested in 10 virtual recipients with a panel of 10 potential allografts, including one from the recipient himself (trap graft). 4 observers were asked to visually inspect the superposition of allograft over the recipient, to classify the allografts into 4 categories according to the matching of anatomic zones, and to select the 3 best matching allografts. The results obtained using the registration method were compared with those from a previous study on the template method.

Results Using the registration method, the observers systematically detected the trap graft. Selections of the 3 best matching allografts performed using registration and template methods were different. Selection of the 3 best matching allografts was improved by the registration method. Finally, reproducibility of the selection was improved when using the registration method.

Interpretation 3-D CT registration provides more useful information than the template method but the final decision lies with the surgeon, who should select the optimal allograft according to his or her own preferences and the needs of the recipient.

Acknowledgements

LP: study design, data analysis, and writing of manuscript; P-LD: statistical analysis and study design; OCa : data analysis and review of manuscript; OCo: study design and observer; CD: observer and review of manuscript; XB: study design, review of manuscript, and supervision.

Authors have received funding from grant 7.4570.06 from Fonds National de la Recherche Scientifique (FNRS-Televie, Belgium) and Fondation Belge contre le cancer SCIE 2006/20. We thank W. André and A. De Grauwe for their cooperation in this study.