Unsupervised partial volume estimation in single-channel image data

Partial volume effects are present in nearly all medical imaging data. These artifacts blur the boundaries between different regions, making accurate delineation of anatomical structures difficult. Here, the authors propose a method for unsupervised estimation of partial volume effects in single-channel image data. Based on a statistical image model, an algorithm is derived for estimating both partial volumes and the means of the different tissue classes in the image. To compensate for the ill-posed nature of the estimation problem, the authors employ a Bayesian approach that places a prior probability model on the parameters. They demonstrate on simulated and real images that the new algorithm is superior in several respects to the fuzzy and Gaussian clustering algorithms that have previously been used for modeling partial volume effects