Impact of interpatient variability on organ dose estimates according to MIRD schema: Uncertainty and variance-based sensitivity analysis.

PURPOSE: Variance-based sensitivity analysis (SA) is described and applied to the radiation dosimetry model proposed by the Committee on Medical Internal Radiation Dose (MIRD) for the organ-level absorbed dose calculations in nuclear medicine. The uncertainties in the dose coefficients thus calculated are also evaluated. METHODS: A Monte Carlo approach was used to compute first-order and total-effect SA indices, which rank the input factors according to their influence on the uncertainty in the output organ doses. These methods were applied to the radiopharmaceutical (S)-4-(3-18 F-fluoropropyl)-L-glutamic acid (18 F-FSPG) as an example. Since 18 F-FSPG has 11 notable source regions, a 22-dimensional model was considered here, where 11 input factors are the time-integrated activity coefficients (TIACs) in the source regions and 11 input factors correspond to the sets of the specific absorbed fractions (SAFs) employed in the dose calculation. The SA was restricted to the foregoing 22 input factors. The distributions of the input factors were built based on TIACs of five individuals to whom the radiopharmaceutical 18 F-FSPG was administered and six anatomical models, representing two reference, two overweight, and two slim individuals. The self-absorption SAFs were mass-scaled to correspond to the reference organ masses. RESULTS: The estimated relative uncertainties were in the range 10%-30%, with a minimum and a maximum for absorbed dose coefficients for urinary bladder wall and heart wall, respectively. The applied global variance-based SA enabled us to identify the input factors that have the highest influence on the uncertainty in the organ doses. With the applied mass-scaling of the self-absorption SAFs, these factors included the TIACs for absorbed dose coefficients in the source regions and the SAFs from blood as source region for absorbed dose coefficients in highly vascularized target regions. For some combinations of proximal target and source regions, the corresponding cross-fire SAFs were found to have an impact. CONCLUSION: Global variance-based SA has been for the first time applied to the MIRD schema for internal dose calculation. Our findings suggest that uncertainties in computed organ doses can be substantially reduced by performing an accurate determination of TIACs in the source regions, accompanied by the estimation of individual source region masses along with the usage of an appropriate blood distribution in a patient's body and, in a few cases, the cross-fire SAFs from proximal source regions.

Feasibility of reducing differences in estimated doses in nuclear medicine between a patient-specific and a reference phantom.

The feasibility of reducing the differences between patient-specific internal doses and doses estimated using reference phantoms was evaluated. Relatively simple adjustments to a polygon-surface ICRP adult male reference phantom were applied to fit selected individual dimensions using the software Rhinoceros®4.0. We tested this approach on two patient-specific phantoms: the biggest and the smallest phantoms from the Helmholtz Zentrum München library. These phantoms have unrelated anatomy and large differences in body-mass-index. Three models approximating each patient's anatomy were considered: the voxel and the polygon-surface ICRP adult male reference phantoms and the adjusted polygon-surface reference phantom. The Specific Absorbed Fractions (SAFs) for internal photon and electron sources were calculated with the Monte Carlo code EGSnrc. Employing the time-integrated activity coefficients of a radiopharmaceutical (S)-4-(3-18F-fluoropropyl)-l-glutamic acid and the calculated SAFs, organ absorbed-dose coefficients were computed following the formalism promulgated by the Committee on Medical Internal Radiation Dose. We compared the absorbed-dose coefficients between each patient-specific phantom and other models considered with emphasis on the cross-fire component. The corresponding differences for most organs were notably lower for the adjusted reference models compared to the case when reference models were employed. Overall, the proposed approach provided reliable dose estimates for both tested patient-specific models despite the pronounced differences in their anatomy. To capture the full range of inter-individual anatomic variability more patient-specific phantoms are required. The results of this test study suggest a feasibility of estimating patient-specific doses within a relative uncertainty of 25% or less using adjusted reference models, when only simple phantom scaling is applied.

Effect of blood activity on dosimetric calculations for radiopharmaceuticals.

The objective of this work was to investigate the influence of the definition of blood as a distinct source on organ doses, associated with the administration of a novel radiopharmaceutical for positron emission tomography-computed tomography (PET/CT) imaging-(S)-4-(3-18F-fluoropropyl)-L-glutamic acid (18F-FSPG). Personalised pharmacokinetic models were constructed based on clinical PET/CT images from five healthy volunteers and blood samples from four of them. Following an identifiability analysis of the developed compartmental models, person-specific model parameters were estimated using the commercial program SAAM II. Organ doses were calculated in accordance to the formalism promulgated by the Committee on Medical Internal Radiation Dose (MIRD) and the International Commission on Radiological Protection (ICRP) using specific absorbed fractions for photons and electrons previously derived for the ICRP reference adult computational voxel phantoms. Organ doses for two concepts were compared: source organ activities in organs parenchyma with blood as a separate source (concept-1); aggregate activities in perfused source organs without blood as a distinct source (concept-2). Aggregate activities comprise the activities of organs parenchyma and the activity in the regional blood volumes (RBV). Concept-1 resulted in notably higher absorbed doses for most organs, especially non-source organs with substantial blood contents, e.g. lungs (92% maximum difference). Consequently, effective doses increased in concept-1 compared to concept-2 by 3-10%. Not considering the blood as a distinct source region leads to an underestimation of the organ absorbed doses and effective doses. The pronounced influence of the blood even for a radiopharmaceutical with a rapid clearance from the blood, such as 18F-FSPG, suggests that blood should be introduced as a separate compartment in most compartmental pharmacokinetic models and blood should be considered as a distinct source in dosimetric calculations. Hence, blood samples should be included in all pharmacokinetic and dosimetric studies for new tracers if possible.