PET was used to measure tumor blood flow, which is potentially valuable for
diagnosis and assessing the effects of therapy. To help visualize regional
differences in blood flow and to improve the accuracy of region-of-interes
t placement, a parametric imaging approach was developed and compared with
the standard region-of-interest method. Methods: Five patients with renal c
ell metastases in the thorax were studied using [O-15]water and dynamic PET
. To assess the reproducibility of the blood flow measurements, multiple wa
ter studies were performed on each patient. Model fitting was done on a pix
el-by-pixel basis using several different formulations of the standard sing
le-compartment model. Results: The tumors studied spanned a wide range of b
lood flows, varying from 0.4 to 4.2 mL/min/g. These values were generally h
igh compared with those of most other tissues, which meant that the tumors
could be readily identified in parametric images of flow. The different mod
el formulations produced images with different characteristics, and no mode
l was entirely valid throughout the field of view. Although tumor blood flo
w measured from the parametric images was largely unbiased with respect to
a standard regional method, large errors were observed with certain models
in regions of low flow. The most robust model throughout the field of view
had only 1 free parameter and, compared with a regional method, gave rise t
o a flow bias of 0.3% +/- 3.1% for tumor and 16% +/- 11% for low-flow soft
tissue (muscle plus fat). With this model, tumor blood flow was measured wi
th an SD of 7.6% +/- 4.0%. Conclusion: Parametric imaging provides a conven
ient way of visualizing regional changes in blood flow, which may be valuab
le in studies of tumor blood flow.