Method for evaluation of different MRI segmentation approaches

Common comparison indices for evaluating MRI segmentation are usually based on a percentage of volumetric measures of different tissue types, the correlation matrix between a segmentation result and the manual segmentation result. Since manual segmentation is not a true reference, the authors propose a new evaluation method, which is based on simulated MR images, the extracted cross covariance, and the inhomogeneity map from real MRI data, together with manual segmentation which are used to generate simulated MR images. The result shows that while common evaluation methods using manual segmentation as reference are subject to intra- and inter-rater variations and require a time consuming manual segmentation step for each study to evaluate, the proposed method is expected to provide faster and more objective measures for comparing different MR segmentation methods.