In the paper that is at the origin of ClinicaDL (
doi:10.1016/j.media.2020.101694), we compared two types of preprocessing in the context of Alzheimer's disease diagnostic and prognostic, a minimal one (corresponding to Clinica's
t1-linear pipeline) and an extensive one. As we didn't see a clear interest in having an extensive preprocessing in terms of classification accuracy, we decided not to maintain this pipeline in ClinicaDL.
The documentation mentions the preprocessing steps that may be performed, not necessarily that are performed. We will clarify that.