Toward improved pediatric dosing guidelines for Tc-99m-MAG3 renal function imaging

1428 Objectives: Current clinical practice uses weight-based dosing for pediatric renal Tc-99m-MAG3 scans. In this study, we quantified the effects of the administered activity on the value of quantitative parameters extracted from dynamic MAG3 renal scans as a function of pediatric patient weight and body habitus.\n Methods: We used the University of Florida (UF) pediatric phantom database to model pediatric patients at 5 ages (newborn, age 1, 5, 10, and 15). For each age, male and female phantoms with variations in body weight (5th, 25th, 50th, 75th, 95th percentile) were generated. Time activity curves (TAC) extracted from pediatric patient studies acquired at Boston Children’s Hospital (BCH) were fitted with a (scaled) Gamma density function to model kidney biokinetics; the fitted parameters were used to generate an ensemble of TACs for phantom data simulation. A dynamic renal Tc-99m-MAG3 study consisting of 80 consecutive posterior images (15 s duration) were simulated modeling all imaging physics. The activity values of the kidney and the background regions of the UF phantom were set based on the dynamic uptake distribution from the BCH patient scans. For each patient variation (age, gender, body weight), we applied a clinically used, weight-based dosing scheme, (100% AA), and 6 lower AA values (at 1%, 2%, 5%, 10%, 20%, 50% of clinical AA), appropriately modeling noise statistics for each AA. From the simulated dynamic data, we estimated time to peak (t2p) of the TAC and residual (TAC value at the last time point divided by the TAC value at t2p), for the left and right kidney cortex and the whole kidney. The region masks for the cortex and the kidney were obtained using an OTSU segmentation method similar to clinical practice. From these data, we estimated the bias and coefficient of variation (COV) for t2p and residual for all AA values and patient variations. The sources of error for the parameter estimates came from quantum noise, realizations, attenuation, scatter, and limited resolution in the image data, and interactions of these physical effects with variations in anatomy and biokinetics. All these factors affected the biases and precisions of the segmentation and the estimated parameter values.\n Results: The biases of the two renal function parameters were relatively constant for moderate AA values (5%-100% of clinical for the residual bias in the left kidney cortex), with a rapid increase (up to 20%) for the lowest AA. The COV of the two renal function parameters varied linearly with AA on a log-log plot, suggesting a power-law relationship between the COV and AA. Different sized patients had values lying at different points along this line, thus requiring a different AA to provide a constant level of error in the estimated parameters.\n Conclusions: The data on the tradeoff between errors in parameter estimates and AA as a function of patient size suggest that results may be useful in understanding the tradeoff between diagnostic accuracy and AA. Combining this with data generated in earlier work on the tradeoff between AA and patient radiation dose can provide objective input for improved dosing decisions for pediatric patients.