Jean-Noël Badel

Simulation of a 6 MV Elekta Precise Linac photon beam using GATE/GEANT4.

The GEANT4-based GATE Monte Carlo (MC) platform was initially focused on PET and SPECT simulations. The new release v6.0 (February 2010) proposes new tools dedicated for radiation therapy simulations. In this work, we investigated some part of this extension and proposed a general methodology for Linac simulations. Details of the modeling of a 6 MV photon beam delivered by an Elekta Precise Linac, with radiation fields ranging from 5 × 5 to 30 × 30 cm(2) at the isocenter are presented. Comparisons were performed with measurements in water. The simulations were performed in two stages: first, the patient-independent part was simulated and a phase space (PhS) was built above the secondary collimator. Then, a multiple source model (MSM) derived from the PhS was proposed to simulate the photon fluence interacting with the patient-dependent part. The selective bremsstrahlung splitting (SBS) variance reduction technique proposed in GATE was used in order to speed up the accelerator head simulation. Further investigations showed that the SBS can be safely used without biasing the simulations. Additional comparisons with full simulations performed on the EGEE grid, in a single stage from the electron source to the water phantom, allowed the evaluation of the MSM. The proposed MSM allowed for calculating depth dose and transverse profiles in 48 hours on a single 2.8 GHz CPU, with a statistical uncertainty of 0.8% for a 10 × 10 cm(2) radiation field, using voxels of 5 × 5 × 5 mm(3). Good agreement between simulations and measurements in water was observed, with dose differences of about 1% and 2% for depth doses and dose profiles, respectively. Additional gamma index comparisons were performed; more than 90% of the points for all simulations passed the 3%/3 mm gamma criterion. To our knowledge, this feasibility study is the first one illustrating the potential of GATE for external radiotherapy applications.

3D absorbed dose distribution estimated by Monte Carlo simulation in radionuclide therapy with a monoclonal antibody targeting synovial sarcoma

BackroundRadiolabeled OTSA101, a monoclonal antibody targeting synovial sarcoma (SS) developed by OncoTherapy Science, was used to treat relapsing SS metastases following a theranostic procedure: in case of significant 111In-OTSA101 tumor uptake and favorable biodistribution, patient was randomly treated with 370/1110 MBq 90Y-OTSA101. Monte Carlo-based 3D dosimetry integrating time-activity curves in VOI was performed on 111In-OTSA101 repeated SPECT/CT. Estimated absorbed doses (AD) in normal tissues were compared to biological side effects and to the admitted maximal tolerated absorbed dose (MTD) in normal organs. Results in the tumors were also compared to disease evolution.ResultsBiodistribution and tracer quantification were analyzed on repeated SPECT/CT acquisitions performed after injection of 111In-OTSA101 in 19/20 included patients. SPECT images were warped to a common coordinates system with deformable registration. Volumes of interest (VOI) for various lesions and normal tissues were drawn on the first CT acquisition and reported to all the SPECT images. Tracer quantification and residence time of 111In-OTSA101 in VOI were used to evaluate the estimated absorbed doses per MBq of 90Y-OTSA101 by means of Monte Carlo simulations (GATE). A visual scale analysis was applied to assess tumor uptake (grades 0 to 4) and results were compared to the automated quantification. Results were then compared to biological side effects reported in the selected patients treated with 90Y-OTSA101 but also to disease response to treatment.After screening, 8/20 patients were treated with 370 or 1110 MBq 90Y-OTSA101. All demonstrated medullary toxicity, only one presented with transient grade 3 liver toxicity due to disease progression, and two patients presented with transient grade 1 renal toxicity. Median absorbed doses were the highest in the liver (median, 0.64 cGy/MBq; [0.27 −1.07]) being far lower than the 20 Gy liver MTD, and the lowest in bone marrow (median, 0.09 cGy/MBq; [0.02 −0.18]) being closer to the 2 Gy bone marrow MTD. Most of the patients demonstrated progressive disease on RECIST criteria during patient follow-up. 111In-OTSA101 tumors tracer uptake visually appeared highly heterogeneous in inter- and intra-patient analyses, independently of tumor sizes, with variable kinetics. The majority of visual grades corresponded to the automated computed ones. Estimated absorbed doses in the 95 supra-centimetric selected lesions ranged from 0.01 to 0.71 cGy per injected MBq (median, 0.22 cGy/MBq). The maximal tumor AD obtained was 11.5 Gy.Conclusions3D dosimetry results can explain the observed toxicity and tumors response. Despite an intense visual 111In-OTSA101 liver uptake, liver toxicity was not the dose limiting factor conversely to bone marrow toxicity. Even though tumors 111In-OTSA101 avidity was visually obvious for treated patients, the low estimated tumors AD obtained by 3D dosimetry explain the lack of tumor response.

Image-based SPECT calibration based on the evaluation of the Fraction of Activity in the Field of View

BackgroundSPECT quantification is important for dosimetry in targeted radionuclide therapy (TRT) and the calibration of SPECT images is a crucial stage for image quantification. The current standardized calibration protocol (MIRD 23) uses phantom acquisitions to derive a global calibration factor in specific conditions. It thus requires specific acquisitions for every clinical protocols. We proposed an alternative and complementary image-based calibration method that allows to determine a calibration factor adapted to each patient, radionuclide, and acquisition protocol and that may also be used as an additional independent calibration.ResultsThe proposed method relies on a SPECT/CT acquisition of a given region of interest and an initial whole-body (WB) planar image. First, the conjugate view of WB planar images is computed after scatter and attenuation correction. 3D SPECT images are reconstructed with scatter, attenuation, and collimator-detector response (CDR) corrections and corrected from apparent dead-time. The field of view (FOV) of the SPECT image is then projected on the corrected WB planar image. The fraction of activity located in the area corresponding to the SPECT FOV is then calculated based on the counts on the corrected WB planar image. The Fraction of Activity in Field Of View (FAF) is then proposed to compute the calibration factor as the total number of counts in the SPECT image divided by this activity. Quantification accuracy was compared with the standard calibration method both with phantom experiments and on patient data.Both standard and image-based calibrations give good accuracy on large region of interest on phantom experiments (less than 7% of relative difference compared to ground truth). Apparent dead-time correction allows to reduce the uncertainty associated with standard calibration from 2.5 to 1%. The differences found between both methods were lower than the uncertainty range of the standard calibration (<3%). In patient data, although no ground truth was available, both methods give similar calibration factor (average difference 3.64%).ConclusionsA calibration factor may be computed directly from the acquired SPECT image providing that a WB planar image is also available and if both acquisitions are performed before biological elimination. This method does not require to perform phantom acquisition for every different acquisition conditions and may serve to double check the calibration with an independent factor.

Voxel‐based multimodel fitting method for modeling time activity curves in SPECT images

Purpose: Estimating the biodistribution and the pharmacokinetics from time‐sequence SPECT images on a per‐voxel basis is useful for studying activity nonuniformity or computing absorbed dose distributions by convolution of voxel kernels or Monte‐Carlo radiation transport. Current approaches are either region‐based, thus assuming uniform activity within the region, or voxel‐based but using the same fitting model for all voxels. Methods: We propose a voxel‐based multimodel fitting method (VoMM) that estimates a fitting function for each voxel by automatically selecting the most appropriate model among a predetermined set with Akaike criteria. This approach can be used to compute the time integrated activity (TIA) for all voxels in the image. To control fitting optimization that may fail due to excessive image noise, an approximated version based on trapezoid integration, named restricted method, is also studied. From this comparison, the number of failed fittings within images was estimated and analyzed. Numerical experiments were used to quantify uncertainties and feasibility was demonstrated with real patient data. Results: Regarding numerical experiments, root mean square errors of TIA obtained with VoMM were similar to those obtained with bi‐exponential fitting functions, and were lower (< 5% vs. > 10%) than with single model approaches that consider the same fitting function for all voxels. Failure rates were lower with VoMM and restricted approaches than with single‐model methods. On real clinical data, VoMM was able to fit 90% of the voxels and led to less failed fits than single‐model approaches. On regions of interest (ROI) analysis, the difference between ROI‐based and voxel‐based TIA estimations was low, less than 4%. However, the computation of the mean residence time exhibited larger differences, up to 25%. Conclusions: The proposed voxel‐based multimodel fitting method, VoMM, is feasible on patient data. VoMM leads organ‐based TIA estimations similar to conventional ROI‐based method. However, for pharmacokinetics analysis, studies of spatial heterogeneity or voxel‐based absorbed dose assessment, VoMM could be used preferentially as it prevents model overfitting.