Post-Treatment Dosimetry Following Yttrium-90 Radioembolization: A Clinical Comparison of Nuclear Medicine and CT-Based Approaches
Abstract
Purpose
Present interim dosimetry analysis from a prospective, multi-center study (NCT05953337) comparing the mean absorbed dose (Dmean) derived from nuclear medicine (NM) and CT imaging of liver cancer patients undergoing yttrium-90-radioembolization (90Y-RE) with radiopaque Eye90 microspheres.
Methods
Twenty-seven patients were treated with Eye90 microspheres. Post-treatment, patients were imaged with CT and NM (SPECT [n=9], PET [n=18]) to measure the absorbed dose to the perfused volume (PV) and tumor (T). In NM-based dosimetry, Dmean was calculated using voxel S-values for PET and the local deposition method for SPECT. For CT, Dmean was calculated through calibration phantom analysis and convolution with a dose-voxel kernel. Dmean was compared using linear regression and Bland-Altman analysis. Correlation coefficients, percent difference ± standard error (bias), and 95% limits of agreement (precision) between Dmean are reported. For PV, Dmean was also compared to a ground truth, single compartment MIRD-based Dmean.
Results
When comparing Dmean from CT- and NM-based dosimetry to MIRD-based Dmean, CT demonstrates significantly improved agreement. There is a very strong correlation between CT- and MIRD-based Dmean to PV (R2=0.96). In NM, the correlation was far weaker (R2=0.56). The MIRD-based Dmean was generally greater than both CT- or NM-based Dmean, particularly NM. For CT, bias and precision were 44±13% and 64%, respectively. For NM, bias and precision were 70±16% and 78%. When comparing Dmean between CT and NM, there was a moderate correlation between for PV (R2=0.57) and a weak correlation for T (R2=0.36). Dmean was generally greater in CT-based dosimetry relative to NM. For PV, bias and precision were 31±19% and 94%, respectively. For T, bias and precision were 10±21% and 104%.
Conclusion
For PV, CT-based dosimetry revealed that Dmean maintained a significantly stronger correlation with the MIRD-based Dmean, relative to NM. CT-based dosimetry had a reduced bias and improved precision relative to NM-based dosimetry.