A BAYESIAN-APPROACH TO SUBVOXEL TISSUE CLASSIFICATION IN NMR MICROSCOPIC IMAGES OF TRABECULAR BONE

NMR microscopy is currently being used as an investigational tool for the evaluation of micromorphometric parameters of trabecular bone as a possible means to assess its strength. Since, typically, the image vo xel size is not significantly smaller than individual trabecular eleme nts, partial volume blurring can be a major complication for accurate tissue classification. In this paper, a Bayesian segmentation techniqu e is reported that achieves improved subvoxel tissue classification. E ach voxel is subdivided either into eight subvoxels twice the original resolution, or up to four subvoxels along the transaxial direction an d the subvoxels optimally classified as either bone or marrow. Based o n a statistical model for partial volume blurring, the likelihood for the number of marrow subvoxels in each voxel can be computed on the ba sis of its measured signal. To resolve the ambiguity of the location o f the marrow subvoxels, a Gibbs distribution is introduced to model th e interaction between the subvoxels. Neighboring subvoxel pairs with t he same tissue label are encouraged, and pairs with distinct labels ar e penalized. The segmentation is achieved by maximizing the a posterio ri probability of the label image using the block ICM (iterative condi tional mode) algorithm. The potential of the proposed technique is dem onstrated in real and synthetic NMR microscopic images.