Because the use of factor analysis has been proposed for extracting pure ph
ysiologic temporal or spatial information from dynamic nuclear medicine ima
ges, factor analysis should be capable of robustly estimating regional myoc
ardial blood flow (rMBF) using (H2O)-O-15 PET without additional (CO)-O-15
PET, which is a cumbersome procedure for patients. Therefore, we measured r
MBF using time-activity curves (TACs) obtained from factor analysis of dyna
mic myocardial (H2O)-O-15 PET images without the aid of (CO)-O-15 PET. Meth
ods: (H2O)-O-15 PET of six healthy dogs at rest and during stress was perfo
rmed simultaneously with microsphere studies using Sr-85, Sc-46, and Sn-113
. We performed factor analysis in two steps after reorienting and masking t
he images to include only the cardiac region. The first step discriminated
each factor in the spatial distribution and acquired the input functions, a
nd the second step extracted regional-tissue TACs, Image-derived input func
tions obtained by factor analysis were compared with those obtained by the
sampling method. rMBF calculated using a compartmental model with tissue TA
Cs from the second step of the factor analysis was compared with rMBF measu
red by microsphere studies. Results: Factor analysis was successful for all
the dynamic (H2O)-O-15 PET images. The input functions obtained by factor
analysis were nearly equal to those obtained by arterial blood sampling, ex
cept for the expected delay. The correlation between rMBF obtained by facto
r analysis and rMBF obtained by microsphere studies was good (r = 0.95). Th
e correlation between rMBF obtained by the region-of-interest method and rM
BF obtained by microsphere studies was also good (r = 0.93). Conclusion: rM
BF can be measured robustly by factor analysis using dynamic myocardial (H2
O)-O-15 PET images without additional (CO)-O-15 blood-pool PET.