Effective and organ doses assessment in neuroradiologic and hemodynamic procedures: comparison among three different simulation software

Introduction

The goal of this work is to compare three different software for effective (ED) and organ (OD) dose estimation to analyze the reliability and investigate the main differences.

Methods

Both thirty neuroradiologic and hemodynamic procedures were analyzed retrospectively choosing a representative sample in terms of kerma-area product and complexity. Each procedure was simulated using NCIRF, VirtualDose-IR Batch Utility and PCXMC. The three software differ for computational method and phantom used and they also need different parameters from RDSR or experimental data to meet the requirements. Bland-Altman method was used to assess the agreement among software.

Results

Large variations in some ODs were observed for all the software, ranging from −70 % to 109 % for neuroradiology and −73 % to 92 % for hemodynamics. This variation is greatly reduced when considering ED, from −23 % to 19 % and from −12 % to 18 % respectively. Bland-Altman analysis showed a poor agreement in OD estimations, with mean differences from 0.8 to 92 mGy and variations up to 200 mGy. The variability in ED is less evident, with mean differences from 1.3 to 3.7 mSv for neuroradiology and from 0.6 mSv to 3.4 mSv for hemodynamics, although in individual cases these differences reached 7.6 mSv and 11 mSv respectively.

Conclusions

Differences principally arise from input parameters, models and phantoms which strongly affect ODs. Nevertheless, the overall EDs is confirmed to be a robust quantity with discrepancies lower than 4 mSv (20 %), showing differences comparable to uncertainties in radiological dosimetry and confirming its utility in population dose assessment.