MIRD Pamphlet No. 34, Part 2: Benchmarking of MIRDct Software for CT Organ Dose Estimation.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
Accurate estimation of organ and effective doses in CT imaging is essential for risk assessment, protocol optimization, and personalized care in diagnostic radiology and nuclear medicine. We systematically benchmarked MIRDct, a freely available mesh phantom-based CT dose calculation software, against established reference software (National Cancer Institute Dosimetry System for Computed Tomography [NCICT] and VirtualDose [Virtual Phantoms]) by evaluating agreement between organ-absorbed doses and effective doses across representative scanners, phantoms, and protocols. Methods: Organ absorbed and effective doses were calculated for adult and pediatric phantoms for whole-body (WB) and regional (head, chest, abdomen-pelvis [AP]) CT examinations. MIRDct uses mesh-based International Commission on Radiologic Protection (ICRP) reference phantoms with anatomically realistic organ surfaces, whereas NCICT and VirtualDose use voxel-based ICRP 110 and hybrid Rensselaer Polytechnic Institute/University of Florida phantom models, respectively. For each software, volumetric CT dose index (CTDIvol) values were obtained from the software interface using matched acquisition parameters; in MIRDct, these values were derived from scanner console-reported outputs for the corresponding protocol settings. Organ absorbed doses, dose coefficients, and effective doses were computed across 44 matched scanner-phantom-protocol configurations. Inter-software differences were summarized using medians and interquartile ranges. For regional protocols, organ-absorbed doses were stratified by irradiation category (in-field, partial-in-field, out-of-field), to assess field-dependent variability. Results: CTDIvol values reported by the 3 software tools showed close agreement across matched protocol configurations, with median inter-software differences not exceeding 7%. For in-field organs, dose coefficients from NCICT and VirtualDose generally agreed with MIRDct values within ±25% across adult and pediatric head, chest, AP, and WB protocols, indicating good agreement in the primary beam region. Larger relative deviations occurred for partial-in-field and out-of-field organs, where doses were scatter-dominated; however, absolute organ doses were less than 2 mGy, limiting clinical relevance. Effective dose estimates showed similar concordance: differences were below 25% for all VirtualDose comparisons except head scans and for WB protocols, whereas adult chest and AP protocols differed by up to 40% relative to NCICT. These differences were associated with variations in phantom anatomy and fixed, pre-tabulated CTDIvol reference values in NCICT and VirtualDose, compared with protocol-specific, console-reported CTDIvol inputs in MIRDct. Conclusion: MIRDct provides organ- and effective-dose estimates that are broadly consistent with established CT dosimetry tools, with agreement typically within ±25% for in-field organs and within a few milligray for absolute doses across adult and pediatric protocols. The use of mesh-based ICRP reference phantoms with anatomically realistic organ surfaces, protocol-specific CTDIvol inputs from the scanner console, and uncertainty propagation supports its application as a research tool for CT dose benchmarking, protocol optimization, and quality assurance in diagnostic CT and nuclear medicine.