Evaluation of several multi-atlas methods for PSEUDO-CT generation in brain MRI-PET attenuation correction

Radiodensity maps for attenuation correction are not readily available in PET-MR. We investigated several such methods to build a patient-specific pseudo-CT, based on a multi-atlas registration and label propagation approach, including a novel maximum probability method. A database of 27 MR/CT image pairs (atlases) was used. We implemented a state-of-the-art algorithm (Ref) along with two variants and our new maximum probability (MaxProb) method. A pseudo-CT was generated for each subject of the database and assessed by comparison with the real CT in a leave-one-out scheme. Jaccard overlap indices were computed per tissue class (air, soft tissue, and bone). Voxel classification error was also assessed. The Jaccard indices and voxel classification error showed that the MaxProb method performance (7.81±1.00%) was only equivalent to Ref (7.69±1.00%). The influence of the number of atlases fused in pseudo-CT generation was studied, showing that an optimal number improves accuracy. By adjusting the number of fused atlases, it was possible to significantly improve on Ref (7.57±1.05%). Methods now need evaluation on additional databases and quantitative PET data to identify potential benefits in PET analysis.