PARAMETER-ESTIMATION AND TISSUE SEGMENTATION FROM MULTISPECTRAL MR-IMAGES

A statistical method is developed to classify tissue types and to segm ent the corresponding tissue regions from relaxation time T1, T2, and proton density P(D) weighted magnetic resonance images. The method ass umes that the distribution of image intensities associated with each t issue type can be expressed as a multivariate likelihood function of t hree weighted signal intensity values (T1, T2, P(D)) at each location within that tissue regions. The method further assumes that the underl ying tissue regions are piecewise contiguous and can be characterized by a Markov random field prior. In classifying the tissue types, the m ethod models the likelihood of realizing the images as a finite multiv ariate-mixture function. The class parameters associated with the tiss ue types (i.e., the weighted intensity means, variances and correlatio n coefficients of the multivariate function, as well as the number of voxels within regions of the tissue types) are estimated by maximum li kelihood. The estimation fits the class parameters to the image data v ia the expectation-maximization algorithm. The number of classes assoc iated with the tissue types is determined by the information criterion of minimum description length. The method segments the tissue regions , given the estimated class parameters, by maximum a posteriori probab ility. The prior is constructed by the tissue-region membership of the first- and second-order neighborhood. The method is tested by a few s ets of T1, T2, and P(D) weighted images of the brain acquired with a 1 .5 Tesla whole body scanner. The number of classes and the associated class parameters are automatically estimated. The regions of different brain tissues are satisfactorily segmented.