Magnetic Resonance Image Analysis by Information
Yue Wang, Jianhua Xuan, Zsolt Roland Szabó · 2001
Quantitative analysis of magnetic resonance (MR) images is a powerful tool for image-guided diagnosis, monitoring, and intervention. The major tasks involve tissue quantification and image segmentation where both the pixel and context images are considered. To extract clinically useful information from images that might be lacking in prior knowledge, we introduce an unsu- pervised tissue characterization algorithm that is both statistically principled and patient specific. The method uses adaptive standard finite normal mixture and inhomogeneous Markov random field models, whose parameters are estimated using expectation-max- imization and relaxation labeling algorithms under information theoretic criteria. We demonstrate the successful applications of the approach with synthetic data sets and then with real MR brain images. method is adopted, both pixel and context images should be con- sidered. In this paper, we introduce an unsupervised image anal- ysis procedure, which is both statistically principled and patient specific, a feature especially important in cases with limited or no prior knowledge (4), (10). We assume that the MR images are single valued and the anatomy of the site may contain ab- normalities. The method involves adaptive use of standard finite normal mixture (SFNM) and inhomogeneous Markov random field (MRF) models, whose parameters are estimated using the fast expectation-maximization (EM) and modified iterated con- ditional modes (MICM) algorithms under a selected informa- tion theoretic criterion (15)-(17). The major difference of our study from previous research in the area (1)-(3), (5), (10) is as follows.