Bayesian Image Decomposition Applied to Relaxographic Imaging

摘要

T_1 relaxographic imaging is a precise and accurate way to characterize tissue. A number of fast MRI acquisition techniques allow both spatial and magnetization recoveries to be well sampled in reasonable imaging times. However, two limitations common to the analysis of relaxographic imaging data are (1) the assumption of single exponential behavior for each image voxel and (2) the treatment of each pixel as an independent entity. The first assumption disregards tissue heterogeneity known to be present and reduces the information content that can be extracted. The latter assumption reduces both the modeling stability and the accuracy of extracted parameters. A new method that overcomes these limitations is presented here. The method, Bayesian Image Decomposition, recovers individual tissue type magnetization recovery curves and their corresponding tissue-specific relaxographic images (i.e. segmented images) from a series of inversion recovery images. The general form of the decomposition is given together with its specific implementation to longitudinal relaxographic imaging. The method is validated by comparison of the results with those of the standard method and by comparison across data sets. A specific advantage of the new method is the ability to determine fractional contributions of tissue subtypes to each image voxel.