Surface Reconstruction for Computer Vision-Based Craniofacial Surgery

Abstract

High-energy traumatic impact of the craniofacial skeleton is an inevitable consequence of today's fast paced society. The work presented in the paper addresses the problem of craniofacial reconstruction using two popular surface matching algorithms namely the iterative closest point (ICP) algorithm and the data aligned rigidity constrained exhaustive search (DARCES) algorithm. The two algorithms are first applied individually and then in combination to achieve the desired reconstruction. The synergistic combination of the DARCES and ICP algorithms is found to yield higher reconstruction accuracy in much shorter execution time compared to the ICP algorithm used in isolation. The local surface irregularities on the fracture surfaces are exploited using a fuzzy set theoretic approach and a curvature-based method. Incorporation of the knowledge of the surface irregularities by means of two reward-penalty schemes in the hybrid DARCES-ICP algorithm is shown to result in robust and accurate surface reconstruction. Experimental results on computer tomography (CT) image data obtained from fractured human mandibles are presented.