Figure 2
From: PET image segmentation using a Gaussian mixture model and Markov random fields
Volume error achieved by various methods for different NEMA sphere diameters, and SBRs. (a) 42% thresholding, (b) EMGMM, (c) GMRF, (d) statistical approach acting on MIP [40], (e) fuzzy locally adaptive Bayesian approach [38], and (f) maximum a posteriori GMRF approach with subsequent deconvolution [39]. (a), (b), and (c) use VOIs of 14×14×20 voxels; (d), (e), and (f) were copied visually from [38,40] and [39]. Note that the diameter ranges vary in (a-f).