Title: Generalizable Patterns in Neuroimaging: How Many Principal Components?

Authors: Lars Kai Hansen, Jan Larsen Finn Årup Nielsen
Department of Mathematical Modeling - Building 321
Technical University of Denmark, DK-2800 Lyngby, Denmark
Phone: +45 4525 3889,3923
Fax: +45 4587 2599
E-mail: lkhansen,jl@imm.dtu.dk

Stephen C. Strother
PET Imaging Service, VA Medical Center, and Radiology+Neurology Depts.
University of Minnesota,
Minneapolis, Minnesota, USA

Egill Rostrup
Danish Center for Magnetic Resonance,
Hvidovre Hospital, Denmark

Robert Savoy
Massachusetts General Hospital,
Boston, USA

Claus Svarer, Olaf B. Paulson
Neurobiology Research Unit,
Rigshospitalet, Copenhagen, Denmark


Generalization can be defined quantitatively and can be used to assess the performance of Principal Component Analysis (PCA). The generalizability of PCA depends on the number of principal components retained in the analysis. We provide analytic and test set estimates of generalization. We show how the generalization error can be used to select the number of principal components in two analyses of functional Magnetic Resonance Imaging activation sets.

Techn. Report., submitted for publication, 1998.