A database of simulated PET volumes with anatomical variability

We propose an original method to simulate a database of realistic simulated PET data accounting for inter-subject anatomical variability, which is based on the use of an automatic labelling method, a functional model and a Monte Carlo based PET simulator. This database has the merit to provide the community with realistic simulated PET studies for commonly used radiotracers, regarding the signal degradation as well as the anatomical and functional characteristics. Consequently, it may be of particular interest to evaluate image processing methods in a fully controlled environment. Whereas the purpose of this study is to present the overall methodology to generate such a bank of simulated data, the actual database is in constant evolution with the addition of anatomical and functional models, and so far, it contains simulated [/sup 18/F]FDG, [/sup 18/F]Dopa and [/sup 11/C]Raclopride, generated from 17 different numerical head models.