Absorbed and effective dose estimates in HR-pQCT of the distal radius and tibia: virtual dosimetry with a GPU-accelerated Monte Carlo code

<i>Objective</i>. To obtain maps of absorbed dose and an estimate of the effective dose to an adult patient for a high resolution peripheral quantitative computed tomography (HR-pQCT) (XtremeCT II) examination of the distal tibia and radius, using a graphical processing unit (GPU)-based Monte Carlo (MC) code.<i>Approach</i>. We adapted the validated code gCTD (GPU-based CT Dose calculator), to replicate the HR-pQCT configuration. MC simulations were performed on digital phantoms of the tibia and radius obtained from bilateral scans of the ankle and wrist of a 25 year-old female volunteer. Scans were segmented using an ad hoc algorithm in Fiji. Simulations run on an NVIDIA GeForce RTX 3090 GPU board. MC dose estimates were validated via computed tomography dose index measurements.<i>Main Results.</i>We obtained the absorbed dose distribution in the skin, bone, bone marrow, fat, and muscle tissues. The effective dose for the HR-pQCT examination were 2.05<i>μ</i>Sv and 2.13<i>μ</i>Sv for the right and left tibia, and 1.48<i>μ</i>Sv and 1.49<i>μ</i>Sv for the right and left radius, respectively, with a Type A statistical uncertainty of 0.06% (<i>k</i>= 3) with 4.66 × 10¹¹photon histories. Corresponding effective dose conversion coefficients (<i>k-factors</i>) were 0.185<i>μ</i>Sv ∙ mGy^-1· cm^-1(tibia), and 0.133<i>μ</i>Sv mGy^-1· cm^-1(radius).<i>Significance.</i>We reported the first independent estimate of the effective dose for standard HR-pQCT clinical scans of the distal tibia and radius with the XtremeCT II scanner. Effective dose estimates (considering a total relative uncertainty of less than 40%) were lower than those indicated by the manufacturer and commonly reported for these scans. With 4.66 × 10⁹photon histories, the gCTD MC code can produce 3D dose maps from segmented HR-pQCT images in less than 12 s (GPU time), with 0.9% (<i>k</i>= 3) statistical uncertainty, making real-time personalized dose estimate feasible.