Estimation of time activity curves from positron emission tomography with NMF and Bootstrap 
Bjarni Bödvarsson, Lars Kai Hansen, Gitte Moos Knudsen, Claus Svarer

Abstract  Measurement of the cerebral input function in terms of the arterial plasma input time activity curve (TAC) is required for assessment and validation of the quantification of receptor binding parameters on the basis of dynamic positron emission tomography (PET) images and kinetic models [4]. A vascular input TAC can be estimated using nonnegative matrix factorization (NMF) [1] [2]. The scaling of the estimated TAC is found by making assumptions on the mixing matrix in the NMF model.
To estimate how precise the estimated curve are, a resampling scheme called Bootstrap[3] is applied to the NMF estimated TACs. The Bootstrap gives error bars on the curves and this can be used to validate the precision of the solution. Correlation with the sampled TAC is used to estimate the number of components in the NMF. 
Keywords  NMF, Bootstrap, PET 
Type  Conference paper [Abstract] 
Conference  Human Brain Mapping 2007 Chicago 
Editors  NeuroImage 
Year  2007 Vol. 36 No. S1 pp. S101 
Electronic version(s)  [pdf] 
BibTeX data  [bibtex] 
IMM Group(s)  Intelligent Signal Processing 