The Impact of Microsphere Deposition Algorithm Complexity on Microdosimetry following Yttrium-90 Radioembolization

The microsphere spatial distribution following Yttrium-90 radioembolization (90Y-RE) is inherently nonuniform, resulting in substantial microscopic dose heterogeneity not captured by conventional macroscopic dosimetry models. The motivation for this study was to gain insight into the microscopic characteristics of the 90Y dose distribution and build a robust framework to further understand the relationship between microdosimetry and clinical outcomes. In this study, a stochastic microsphere deposition algorithm sampled histologically derived cumulative distribution functions (CDFs) governing microsphere cluster diameter (C_dia), distance between clusters (C_dist), and cluster population (C_pop). Six unique models were generated to examine the impact of algorithm complexity on the corresponding absorbed dose distribution, ranging from a completely uniform to fully stochastic reference model. A two-sample statistical Kolmogorov-Smirnov test compared C_dia, C_dist, and C_pop derived separately from discrete and continuous CDFs. Microdosimetry calculations were performed by convolving a high-resolution dose-voxel kernel with each model. The mean absorbed dose D_mean and various dose-volume metrics (D_x,x=1,5,10,50,90,95,99) were calculated and compared to the reference model to assess the impact of algorithm complexity on dose metric error (E_x). Published median values of C_dia, C_dist, and C_pop agreed well with simulated counterparts. There were no statistically significant differences in sampling between discrete and continuous CDFs for C_dia (p=0.083), C_dist (p=0.104), and C_pop (p=0.094). Convolution with the 90Y DVK resulted a -0.3% deviation compared to a single compartment dose estimate. Model comparisons suggest that sampling C_dist is critical for accurately modeling low-dose regions (E_99=16%), while sampling C_pop is critical for resolving absorbed dose hot spots (E_1=-12%). In contrast, sampling from C_dia had minimal impact on model accuracy. The results of this study provide the necessary framework to develop an improved understanding of the relationship between microdosimetry and clinical outcomes following 90Y-RE.