Porous polyethylene has become a popular alloplastic material for maxillofacial reconstruction. Despite its benefits, this material is radiolucent on conventional radiography. Computed tomography of traditional graphic segmentation based on gray levels was reported by a few authors, but the accuracy has not been described. The aim of this study was to image this material using segmentation based on Graph-cut algorithm. Another object was to compare accuracy of 2 algorithms. Between 2005 and 2010, 15 patients (12 females) participated in the study. Computed tomography was used for image acquisition. Graphic segmentation based on gray level involved use of the software OsiriX. Stroke-based specification for user interaction and Graph-cut segmentation involved a self-developed module. The accuracy of both segmentation algorithms was tested by comparing the volume of reconstructed Medpor with the actual volume measured by the draining method during surgery. Twenty-seven comparisons of the computed and actual volumes were gained. Linear regression was done between reconstructed and real volume. For the volume of Graph-cut algorithm, the regression equation was T = -0.19425 + 1.02692S; the difference between the population intercept and zero was not significant (t = -0.83, P = 0.4172). For the volume of gray-level algorithm, the regression equation was T = -1.15282 + 1.14101S; the difference between the population intercept and zero was significant (t = -3.04, P = 0.0055).These results demonstrated porous polyethylene can be well defined by computed tomography segmentation with Graph-cut and gray-level algorithms. Graph-cut algorithm has a better accuracy compared with that of gray-level algorithm.