Dosimetric impact of computational phantom selection in passive scattering proton therapy: Development and validation of an extended MIRD prostate phantom using Geant4 Monte Carlo simulation.

Abstract

Accurate anatomical modelling is important for reliable Monte Carlo (MC) proton dosimetry, yet the influence of computational phantom selection on clinically relevant dose prediction remains insufficiently quantified. In this study, an anatomically extended Medical Internal Radiation Dose (MIRD) computational phantom was developed for passive scattering proton therapy of prostate cancer and systematically compared with the anatomically realistic reference male phantom recommended by the International Commission on Radiological Protection using a validated Geant4 MC simulation framework. The extended phantom incorporated the prostate, bladder, rectum, urethra, and intraprostatic tumour volumes. At the same time, a clinical passive-scattering beamline at the CATANA proton therapy facility was modelled to generate spread-out Bragg peaks for treatment planning. The simulation framework demonstrated excellent agreement with experimental depth-dose measurements, yielding a practical proton range (R₉₀) deviation of 1.10 mm and a gamma passing rate of 98.63% under the 3%/2 mm criterion, confirming the accuracy of the dosimetric model. Following dose normalisation, both phantom models achieved identical target coverage (D₉₅=74 Gy) and conformity (CI = 0.95). In contrast, anatomical representation produced measurable differences in proton transport characteristics, including a 5.29% variation in proton range and a 72.4% difference in distal falloff. Quantitative dosimetric analysis further demonstrated differences in internal target dose distribution and organ-at-risk dose metrics despite comparable overall treatment quality, indicating that conventional target coverage indices alone do not fully reflect the dosimetric consequences of anatomical modelling. Although all evaluated organ-at-risk doses remained below commonly accepted clinical dose constraints, the observed variations in dose deposition highlight the sensitivity of proton transport to tissue geometry and heterogeneity. These findings suggest that computational phantom selection represents an important modelling factor influencing Monte Carlo proton dose prediction and that mathematically defined phantoms can provide a flexible platform for methodological and dosimetric investigations when their inherent anatomical simplifications are appropriately considered.