Quantitation of regional cerebral blood flow corrected for partial volume effect using O-15 water and PET: I. Theory, error analysis, and stereologic comparison.

Limited spatial resolution of positron emission tomography (PET) can cause significant underestimation in the observed regional radioactivity concentration (so-called partial volume effect or PVE) resulting in systematic errors in estimating quantitative physiologic parameters. The authors have formulated four mathematical models that describe the dynamic behavior of a freely diffusible tracer (H215O) in a region of interest (ROI) incorporating estimates of regional tissue flow that are independent of PVE. The current study was intended to evaluate the feasibility of these models and to establish a methodology to accurately quantify regional cerebral blood flow (CBF) corrected for PVE in cortical gray matter regions. Five monkeys were studied with PET after IV H2(15)O two times (n = 3) or three times (n = 2) in a row. Two ROIs were drawn on structural magnetic resonance imaging (MRI) scans and projected onto the PET images in which regional CBF values and the water perfusable tissue fraction for the cortical gray matter tissue (hence the volume of gray matter) were estimated. After the PET study, the animals were killed and stereologic analysis was performed to assess the gray matter mass in the corresponding ROIs. Reproducibility of the estimated parameters and sensitivity to various error sources were also evaluated. All models tested in the current study yielded PVE-corrected regional CBF values (approximately 0.8 mL x min(-1) x g(-1) for models with a term for gray matter tissue and 0.5 mL x min(-1) x g(-1) for models with a term for a mixture of gray matter and white matter tissues). These values were greater than those obtained from ROIs tracing the gray matter cortex using conventional H2(15)O autoradiography (approximately 0.40 mL x min(-1) x g(-1)). Among the four models, configurations that included two parallel tissue compartments demonstrated better results with regards to the agreement of tissue time-activity curve and the Akaike's Information Criteria. Error sensitivity analysis suggested the model that fits three parameters of the gray matter CBF, the gray matter fraction, and the white matter fraction with fixed white matter CBF as the most reliable and suitable for estimating the gray matter CBF. Reproducibility with this model was 11% for estimating the gray matter CBF. The volume of gray matter tissue can also be estimated using this model and was significantly correlated with the results from the stereologic analysis. However, values were significantly smaller compared with those measured by stereologic analysis by 40%, which can not be explained by the methodologic errors. In conclusion, the partial volume correction was essential in quantitation of regional CBF. The method presented in this article provided the PVE-corrected regional CBF in the cortical gray matter tissue. This study also suggests that further studies are required before using MRI derived anatomic information for PVE correction in PET.