Factor analysis and kinetic modeling for PET: are they relevant for clinical applications?

Anatomic and functional aspects of biologic processes are studied with multi-modality imaging. PET can perform this in-vivo and almost real-time. The dynamics and kinetics of new drugs, their optimal dose and individual response need to be determined. In human cancer, non-responders to therapy need to be separated from responders, and patients surveyed to detect recurrence. We explored the potential of factor analysis (FA) by extracting vessels and tissues from the dynamic images. With thresholding techniques a volume-of-interest (VOI) is delineated and regional blood clearance (input function) and tissue uptake (output function) measured. kinetic modeling (KM) estimates the rate constants between compartments representing the underlying biologic pathways. Tracers used were: ¹¹C-acetate, ¹³N-ammonia, ¹⁸F-fluoride, ¹⁸F- FDOPA, ¹⁸F-FLT and ¹⁸F-FDG. Dynamic studies were acquired with high temporal sampling during the first 2 minutes, decreasing over the next phase to 5 min frames. Total image acquisition varied from 20-75 min dependent on tracer and organ investigated. Breast and prostate cancer, brain tumor, myocardial perfusion, and osteoporosis were evaluated. FA was successful in all cases and produced clear parametric images. Accurate input functions could be derived for all situations. Tissue or tumor uptake can be low and images noisy, yielding sub-optimal output functions. Comparison of the FA input function to plasma blood samples in breast cancer showed excellent concordance. KM provided rate constants that did not reveal significant differences between input curves from blood samples vs FA derived. KM may help elucidating the metabolic pathways, e.g. ¹⁸F-FDOPA revealed a different mechanism for tumor than striatum. ¹¹C-acetate kinetics in prostate cancer differ completely from those in myocardium. ¹⁸F-FLT uptake reflects cell proliferation and shows the greates- t decrease within 1 week after treatment in responders. Our procedure is automated and virtually user independent; this processing tool facilitates PET as a reproducible quantification method in routine clinical applications.