Towards better quantification than Standard Uptake Value Ratios for radiotracers following reference region models

1413 Objectives: Many positron emission tomography (PET) studies make use of the standardized uptake value ratio (SUVR) outcome, which is computed based on a predefined short time interval post injection. A drawback of this approach is that the radiotracer does not always achieve equilibrium during this period. Individual variability in radiotracer delivery or clearance can contribute to higher bias and variance in the SUVR, in comparison to the true underlying distribution volume ratio value (DVRT). This work extends a recently proposed correction for the total distribution volume (VT) estimates in the absence of a true equilibrium[1] to correct SUVR estimates. Methods: Our proposed formula for correcting the apparent SUVR estimates takes the following form: DVRC = SUVR k2,ref/[k2,ref - βref + SUVR βt/R1], where DVRC is the corrected SUVR, k2,ref is the population-based k2 parameter of the reference region, βref is the clearance rate of the tracer from the reference tissue, βt is the clearance rate of the tracer from the target tissue, and R1 is defined as the ratio K1,t/K1,ref, which is typically near 1. We first validate this correction formula by performing simulations for 11C-LSN3172176, a tracer that specifically binds to the muscarinic M1 acetylcholine receptor. SUVR values were computed as the average ratio of activity in various brain regions as compared to the reference region (cerebellum) 60-80 minutes post injection. Both SUVR and DVRC values were then compared with the simulated DVRT values. This correction was then tested on a cohort of 22 human subjects who underwent dynamic PET scans after a bolus 11C-LSN3172176 injection on the HRRT scanner. The correction formula was found to be particularly sensitive to the errors in the estimation of βt from 60-80 minutes of PET data. A novel regression-based estimator was therefore developed to compute βt for this time duration. DVRT values were estimated across nine different brain regions in these subjects from a 1TCM fit of the full dynamic scan data, and compared with SUVR and DVRC estimated from data 60-80 min post injection. Results: In simulations, SUVR, on average, overestimated the true DVR across different brain regions (mean: 25.7%, range: [4.8, 37.2]%). The standard deviation of the error in DVRT estimation by SUVR was, on average, 12.2% of DVRT (range: [11.4, 14.3]%). DVRC reduced the mean bias and the mean standard deviation of the error in DVRT prediction to -0.3% (range: [-2.7, 2.6]%) and 4.0% (range: [2.8, 5.3]%) respectively. The human data revealed similar trends, as SUVR overestimated the true DVR across all brain regions (mean bias: 21.0%, range: [4.8, 30.3]%). With the exception of the ventral striatum (the region with highest binding), DVRC reduced the bias across all regions (mean bias: -5.0%, range: [-17.1, 2.2]%). Reduction in the bias magnitude was statistically significant for the caudate (p < 0.05), brain regions in the cerebral cortex (p < 0.0001) and the hippocampus (p < 0.0001) using two-sampled t-test (α = 0.05, one-tailed). The reduction in bias magnitude in the putamen approached significance (p=0.065). DVRC also reduced the standard deviation of the error in DVRT prediction from 13.2% (range: [9.7, 14.2]%) to 9.6% (range: [7.1, 11.0]%). Conclusions: These preliminary results suggest that this correction can reduce both the bias and variance associated with using raw SUVR values. Future studies will further refine this correction approach and characterize its performance on other tracers. References1.Hillmer, A.T. and Carson, R.E., 2020. Quantification of PET infusion studies without true equilibrium: A tissue clearance correction. Journal of Cerebral Blood Flow & Metabolism, 40(4), pp.860-874.