Background: Regional cerebral perfusion measured by single photon emis
sion computed tomography (SPECT) was examined as a preclinical predict
or of the development of Alzheimer's disease (AD). Methods: Singular v
alue decomposition was used to produce 20 SPECT factors (known as vect
ors) (n = 152). Vector scores were then computed for four groups (n =
136), differing in cognitive status: Group 1-normal controls at both b
aseline and follow-up; Group 2-subjects with ''questionable'' AD at bo
th baseline and follow-up; Group 3-subjects with questionable AD at ba
seline who converted to AD on follow-up (Converters); Group 4-subjects
with AD at baseline. All SPECT data in the analyses were gathered at
baseline. Results: The four groups could be distinguished on the basis
of their baseline SPECT data (p less than or equal to 0.00005; hit ra
te = 83%), Regional decreases in perfusion were most prominent among C
onverters in the hippocampalamygdaloid complex, the posterior cingulat
e, the anterior thalamus, and the anterior cingulate. Inclusion of apo
lipoprotein E status did not significantly improve the discrimination.
Conclusions: SPECT data gathered and analyzed in this manner may be u
seful as one aspect of the preclinical prediction of AD. Three of the
four brain regions important for discriminating Converters from normal
controls involve a distributed brain network pertaining to memory, su
ggesting that this network may be selectively affected in the earliest
stages of AD.