A generalized Renkin-Crone relation for Rb-82 GFADS analysis

Objectives Myocardial blood flow (MBF) estimation by Rb-82 dynamic PET typically requires the use of a generalized Renkin-Crone (gRC) empirical relation. Compared to conventional ROI sampling of the arterial input function (AIF), generalized factor analysis of dynamic sequences (GFADS) analysis generally yields AIFs with higher peaks and better noise properties but the use of previously published ROI-based gRC relations (Lortie et al. 2007; Prior et al. 2012) may result in biased MBF estimates. Here we develop a GFADS-specific gRC relation and demonstrate that it provides improved MBF estimates while preserving the noise properties of GFADS analysis.

Methods Normal volunteers (N=22) were recruited at two academic medical centers for N13 ammonia and Rb82 dynamic rest/stress PET exams. MBF was estimated by fitting N13 ammonia data to a two-compartment model. Rb82 data were processed using the GFADS algorithm (El Fahkri et al. 2005) and K1 was estimated by fitting a one-compartment model. A GFADS-gRC relation was then fit to the K1 values as a function of MBF. For comparison a conventional ROI-gRC relation was derived based on ROI-sampled AIFs. The GFADS-gRC relation was validated using a separate set of rest/stress Rb82 PET datasets (N=29). The concordance correlation coefficient (CCC) was used to assess agreement of MBF between GFADS and ROI analyses under two conditions: using ROI-gRC or GFADS-gRC to determine GFADS MBF.

Results The CCC between GFADS and ROI MBF was significantly improved when using the GFADS-gRC compared to the ROI-gRC (Table). Moreover the GFADS MBF mean standard error (0.023) was reduced by 67% compared to that of the ROI MBF (0.068).

Conclusions We have developed an empirical GFADS-specific gRC relation that provides improved agreement of MBF estimates with those of a conventional ROI-gRC while preserving the favorable noise properties of GFADS analysis.