Author: Ben Piacitelli, Celeste Winters 👨🔬
Affiliation: OHSU, Department of Diagnostic Radiology, Oregon Health & Science University 🌍
Purpose: To compare recovery coefficients measured with Lu-177 and Tc-99m with the goal of assessing accuracy of the Tc-99m surrogate method and comparing RC reproducibility between scanners of the same make and model.
Methods: NEMA phantom scans were conducted with Tc-99m and Lu-177 on three Siemens Pro.Specta SPECT/CT scanners, using our standard Lu-177 dosimetry protocol (BroadQuant, OSCGMM, 8i16s, CTAC, ME collimators, 15 seconds/frame, 120 projections). All scanners were calibrated with a Tc-99m point source and volume source using ME collimators for quantitative reconstruction. Activity concentration sphere-to-background ratio was 12.25:1. CT scans were contoured by placing ROIs on the respective spheres. RC values were calculated and plotted against sphere volume, then RC curves were fitted. An additional in-scan calibration factor was applied to Tc-99m images for a second analysis.
Results: Inter-scanner Lu-177 measurements varied across the 3 scanners by an average of 7.4% (min:1.7%, max:12.9%). Tc-99m measurements showed an average difference of 10.4% (min:6.0%, max:19.6%). When measured on xSPECT reconstructions, RC mean values measured with Tc-99m vs. Lu-177 demonstrated an average absolute error of 12.2% (min:7.0%, max:17.9%) for Scanner 1, 14.3% for Scanner 2 (min:7.8%, max:24.8%), and 11.8% for Scanner 3 (min: 6.8%, max:18.1%). However, when a correction factor for the known activity in the phantom background was applied, errors dropped to 3.4%, 5.3%, 10.8%, respectively. In comparison, Graves et al. reported an absolute error for mean RC of 3% for spheres in the CTN3 phantom (sphere diameter 13 - 37 mm).
Conclusion: Analysis of the RC curves demonstrates that Tc-99m may produce acceptable results for deriving Lu-177 RC curves, however, additional validation of your scanners sensitivity calibration needs to be completed as a prerequisite. Comparisons of Lu-177 measurements across three identical scanners show a 7.4% absolute error, suggesting that RC measurements may not be suitable to use for different scanners.