Abstract
1769
Objectives 15O-Water has previously been validated for the measurement of regional pulmonary blood flow (PBF) using the Kety model (Mintun 1986). However, a recognized challenge in implementing this model is the high covariance between PBF (K1) and transit delay (td) in the parameter estimation, potentially introducing a large bias in the PBF estimate (Richard 2002). The goal of this work is to develop a strategy to uncouple PBF and td using differing estimation schemes and time series framing.
Methods Bolus 15O-water studies were performed on 12 beagles scanned 3 times each using a GE Discovery VCT PET/CT scanner. PET framing (from listmode) was investigated for 1s/, 2s/ and 3s/frame until bolus passage through tissue. An image derived input function was obtained from the right ventricle. The model included terms for PBF, partition coefficient (λ) and td. Parameter estimates were obtained using lsqnonlin in Matlab with uniform weighting for TACs obtained from regions of high and low perfused lung tissue.
Results 1s framing showed the least covariance between PBF and td, and highest sensitivity to small differences in td, with all estimates being < 2s. Compared to 1s framing, there was significant overestimation of 12.6±2.8% and 8.9±3.1% in PBF for 2s frames, and 65.6±14.3% and 30.0±13.4% for 3s frames in high and low perfused regions, respectively. This corresponded to overestimation of 2.2±3.0% and 3.2±2.7% in td for 2s frames, and 72.4±7.5% and 39.9±9.6% for 3s frames in high and low perfused regions.
Conclusions Short frames (~1s) are required at the time of bolus arrival to accurately estimate td and, in turn, reduce the delay dependent bias in PBF. The strategy of fixing td in estimating PBF will produce systematic variability in PBF due to regional variations in bolus arrival. Denoising strategies (e.g. spatial smoothing) may be implemented to improve SNR and precision in PBF estimation at the voxel level.
Research Support NIH UL1TR000427 NIH KL2TR000428