Direct comparison of single-scan autoradiographic with multiple-scan least-squares fitting approaches to PET CMRO2 estimation.

The time course of local cerebral radioactivity concentration after bolus inhalation of oxygen gas labeled with O-15 was measured in a rapid dynamic sequence of positron tomographic images. Four normal subjects were studied at rest. In each study, 15 multiple-slice image sets were acquired over a 3-min period in a Scanditronix model 384 tomograph. The radioactivity concentration in arterial blood was measured at 1-s intervals by means of a pump-fed flow-through detector. Pump effluent was directed to discrete samples that were separated into plasma and cell fractions to estimate the accumulation of labeled, recirculating water arising from systemic metabolism. Stereotactically matched scans of local cerebral blood flow and volume were acquired in the same imaging session, and the derived values were used as fixed parameters in the model fits of the time courses of pixel radioactivity in the oxygen study. Rapid nonlinear least-squares parameter optimization was used to estimate simultaneously the local CMRO2 and the brain/blood relative distribution volume for water in each image pixel. The same scan data were combined into effective single frames of various starting times and durations for analysis using the single-scan ("autoradiographic") approach to CMRO2 estimation, which requires a presumed value for relative distribution volume. Oxygen use values derived using this approach were observed to be strongly dependent on the relative distribution volume value chosen, particularly for long study durations. However, for each gray matter region of interest studied, a uniform value for the relative distribution volume existed such that the estimated CMRO2 values were independent of the starting time and duration of the single scan used, and were furthermore the same as that yielded by the multiple-scan least-squares fitting of the total time course in the same region. We conclude that the properties of the single-scan and multiple-scan approaches are very similar at the same total study duration, provided that the value selected for the water relative distribution volume brings the measured and computed tissue time courses into correspondence.