A comparison of methods for adapting $^{177}{\rm Lu}$ dose-voxel-kernels to tissue inhomogeneities

A single-institution experience with 177Lu RPT workflow improvements and qualifying the SPECT/CT imaging for dosimetry

Background and purpose

Implementing any radiopharmaceutical therapy (RPT) program requires a comprehensive review of system readiness, appropriate workflows, and training to ensure safe and efficient treatment delivery. A quantitative assessment of the dose delivered to targets and organs at risk (OAR) using RPT is possible by correlating the absorbed doses with the delivered radioactivity. Integrating dosimetry into an established RPT program demands a thorough analysis of the necessary components and system fine-tuning. This study aims to report an optimized workflow for molecular radiation therapy using 177Lu with a primary focus on integrating patient-specific dosimetry into an established radiopharmaceutical program in a radiation oncology setting.

Materials and methods

We comprehensively reviewed using the Plan-Do-Check-Act (PDCA) cycle, including efficacy and accuracy of delivery and all aspects of radiation safety of the RPT program. The GE Discovery SPECT/CT 670DR™ system was calibrated per MIM protocol for dose calculation on MIM SurePlan™ MRT software. Jaszcak Phantom with 15-20 mCi of 177Lu DOTATATE with 2.5 µM EDTA solution was used, with the main energy window defined as 208 keV ±10% (187.6 to 229.2 keV); the upper scatter energy window was set to 240 keV ±5% (228 to 252 keV), while the lower scatter energy window was 177.8 keV ±5% (168.9 to 186.7 keV). Volumetric quality control tests and adjustments were performed to ensure the correct alignment of the table, NM, and CT gantry on SPECT/CT. A comprehensive end-to-end (E2E) test was performed to ensure workflow, functionality, and quantitative dose accuracy.

Results

Workflow improvements and checklists are presented after systematically analyzing over 400 administrations of 177Lu-based RPT. Injected activity to each sphere in the NEMA Phantom scan was quantified, and the MIM Sureplan MRT reconstruction images calculated activities within ±12% of the injected activity. Image alignment tests on the SPECT/CT showed a discrepancy of more than the maximum tolerance of 2.2 mm on any individual axis. As a result of servicing the machine and updating the VQC and COR corrections, the hybrid imaging system was adjusted to achieve an accuracy of <1 mm in all directions.

Conclusion

Workflows and checklists, after analysis of system readiness and adequate training for staff and patients, are presented. Hardware and software components for patient-specific dosimetry are presented with a focus on hybrid image registration and correcting any errors that affect dosimetric quantification calculation. Moreover, this manuscript briefly overviews the necessary quality assurance requirements for converting diagnostic images into dosimetry measurement tools and integrating dosimetry for RPT based on 177Lu.

Averaging of absorbed doses: How matter matters

BACKGROUND

Dosimetry in radionuclide therapy often requires the calculation of average absorbed doses within and between spatial regions, for example, for voxel-based dosimetry methods, for paired organs, or across multiple tumors. Formation of such averages can be made in different ways, starting from different definitions.

PURPOSE

The aim of this study is to formally specify different averaging strategies for absorbed doses, and to compare their results when applied to absorbed dose distributions that are non-uniform within and between regions.

METHODS

For averaging within regions, two definitions of the average absorbed dose are considered: the simple average over the region (the region average) and the average when weighting by the mass density (density-weighted region average). The latter is shown to follow from the definition of mean absorbed dose according to the ICRU, and to be consistent with the MIRD formalism. For averaging between different spatial regions, three definitions follow: the volume-weighted, the mass-weighted, and the unweighted average. With respect to characterizing non-uniformity, the different average definitions lead to the use of dose-volume histograms (DVHs) (region average), dose-mass histograms (DMHs) (density-weighted region average), and unweighted histograms (unweighted average). Average absorbed doses are calculated for three worked examples, starting from the different definitions. The first, schematic, example concerns the calculation of the average absorbed dose between two regions with different volumes or mass densities. The second, stylized, example concerns voxel-based dosimetry, for which the average absorbed-dose rate within a region is calculated. The geometries studied include three 177 Lu-filled voxelized spheres, where the sphere masses are held constant while the material compositions, densities, and volumes are varied. For comparison, the mean absorbed-dose rates obtained using unit-density sphere S-values are also included. The third example concerns SPECT/CT-based tumor dosimetry for five patients undergoing therapy with 177 Lu-PSMA and six patients undergoing therapy with 177 Lu-DOTA-TATE, for which the average absorbed-dose rates across multiple tumors are calculated. For the second and third examples, analyses also include representations by histograms.

RESULTS

Example 1 shows that the average absorbed doses, calculated using different definitions, can differ considerably if the masses and absorbed doses for two regions are markedly different. From example 2 it is seen that the density-weighted region average is stable under different activity and density distributions and is also in line with results using S-values. In contrast, the region average varies as function of the activity distribution. In example 3, the absorbed dose rates for individual tumors differ by (1.1 ± 4.3)% and (-0.1 ± 0.4)% with maximum deviations of +34.4% and -1.4% for 177 Lu-PSMA and 177 Lu-DOTA-TATE, respectively, when calculated as region averages or density-weighted region averages, with largest deviations obtained when the density is non-uniform. The average absorbed doses calculated across all tumors are similar when comparing mass-weighted and volume-weighted averages but these differ substantially from unweighted averages.

CONCLUSION

Different strategies for averaging of absorbed doses within and between regions can lead to substantially different absorbed-dose estimates. At reporting of radionuclide therapy dosimetry, it is important to specify the averaging strategy applied.

Patient-specific biokinetics and hybrid 2D/3D approach integration in OEDIPE software: Application to radioiodine therapy

BACKGROUND

The progression of targeted radionuclide therapy requires the development of dosimetry software accounting for patient-specific biokinetics. New functionalities were thus developed in the OEDIPE software, to deal with multiple 3D images or multiple planar images and a SPECT image.

MATERIEL & METHOD

Methods were implemented to recover patient biokinetics in volumes of interest. If several 3D SPECT images are available, they are registered to a reference CT scan. When several planar images and a single SPECT are available, the planar images are registered to the SPECT and counts of the planar images converted to activity. To validate these developments, six SPECT/CT and planar images of a Jaszczak phantom containing I-131 were acquired at different dates. Cumulated activity was estimated in each sphere using the SPECT/CT images only or the planar series associated to one SPECT/CT. Biokinetics and doses in lesions and in the lungs of a patient treated with I-131 for differentiated thyroid cancer were then estimated using four planar images and a SPECT/CT scan. Whole-body retention data were used to compare the biokinetics obtained from the planar and SPECT data.

RESULTS

Activities and cumulated activities estimated using OEDIPE in the phantom spheres agreed well with the reference values for both approaches. Results obtained for the patient compared well with those derived from whole-body retention data.

CONCLUSION

The implemented features allow automatic evaluation of patient-specific biokinetics from different series of patient images, enabling patient-specific dosimetry without the need for external software to estimate the cumulated activities in different VOIs.

Updating 90Y Voxel S-Values including internal Bremsstrahlung: Monte Carlo study and development of an analytical model

PURPOSE

Internal Bremsstrahlung (IB) is a process accompanying β-decay but neglected in Voxel S-Values (VSVs) calculation. Aims of this work were to calculate, through Monte Carlo (MC) simulation, updated 90Y-VSVs including IB, and to develop an analytical model to evaluate 90Y-VSVs for any voxel size of practical interest.

METHODS

GATE (Geant4 Application for Tomographic Emission) was employed for simulating voxelized geometries of soft tissue, with voxels sides l ranging from 2 to 6 mm, in steps of 0.5 mm. The central voxel was set as a homogeneous source of 90Y when IB photons are not modelled. For each l, the VSVs were computed for 90Y decays alone and for 90Y + IB. The analytical model was then built through fitting procedures of the VSVs including IB contribution.

RESULTS

Comparing GATE-VSVs with and without IB, differences between + 25% and + 30% were found for distances from the central voxel larger than the maximum β-range. The analytical model showed an agreement with MC simulations within ± 5% in the central voxel and in the Bremsstrahlung tails, for any l value examined, and relative differences lower than ± 40%, for other distances from the source.

CONCLUSIONS

The presented 90Y-VSVs include for the first time the contribution due to IB, thus providing a more accurate set of dosimetric factors for three-dimensional internal dosimetry of 90Y-labelled radiopharmaceuticals and medical devices. Furthermore, the analytical model constitutes an easy and fast alternative approach for 90Y-VSVs estimation for non-standard voxel dimensions.

Voxel-S-Value based 3D treatment planning methods for Y-90 microspheres radioembolization based on Tc-99m-macroaggregated albumin SPECT/CT

Partition model (PM) for Y-90 microsphere radioembolization is limited in providing 3D dosimetrics. Voxel-S-Values (VSV) method has good agreement with Monte Carlo (MC) simulations for 3D absorbed dose conversion. We propose a new VSV method and compare its performance along with PM, MC and other VSV methods for Y-90 RE treatment planning based on Tc-99m MAA SPECT/CT. Twenty Tc-99m-MAA SPECT/CT patient data are retrospectively analyzed. Seven VSV methods are implemented: (1) local energy deposition; (2) liver kernel; (3) liver kernel and lung kernel; (4) liver kernel with density correction (LiKD); (5) liver kernel with center voxel scaling (LiCK); (6) liver kernel and lung kernel with density correction (LiLuKD); (7) proposed liver kernel with center voxel scaling and lung kernel with density correction (LiCKLuKD). Mean absorbed dose and maximum injected activity (MIA) obtained by PM and VSV are evaluated against MC results, and 3D dosimetrics generated by VSV are compared with MC. LiKD, LiCK, LiLuKD and LiCKLuKD have the smallest deviation in normal liver and tumors. LiLuKD and LiCKLuKD have the best performance in lungs. MIAs are similar by all methods. LiCKLuKD could provide MIA consistent with PM, and precise 3D dosimetrics for Y-90 RE treatment planning.

Voxel-S-value methods adapted to heterogeneous media for quantitative Y-90 microsphere radioembolization dosimetry

PURPOSE

The absorbed dose estimation from Voxel-S-Value (VSV) method in heterogeneous media is suboptimal as VSVs are calculated in homogeneous media. The aim of this study is to develop and evaluate new VSV methods in order to enhance the accuracy of Y-90 microspheres absorbed dose estimation in liver, lungs, tumors and lung-liver interface regions.

METHODS

Ten patients with Y-90 microspheres SPECT/CT and PET/CT data, six of whom had additional Tc-99m-macroaggregated albumin SPECT/CT data, were analyzed from the Deep Blue Data Repository. Seven existing VSV methods along with three newly proposed VSV methods were evaluated: liver and lung kernel with center voxel scaling (LiLuCK), liver kernel with density correction and lung kernel with center voxel scaling (LiKDLuCK), liver kernel with center voxel scaling and lung kernel with density correction (LiCKLuKD). Monte Carlo (MC) results were regarded as the gold standard. Absolute absorbed dose errors (%AADE) of these methods for the liver, lungs, tumors, upper liver, and lower lungs were assessed.

RESULTS

Liver and tumor's median %AADE of all methods were <3% for three types of imaging data. In the lungs, however, three recently proposed VSV methods provided median %AADEs of less than 7%, whereas the differences exceeded 20% for existing methods that did not use a lung kernel. LiCKLuKD could achieve median %AADE <2% in the liver, upper liver and tumors, and median %AADE <7% in the lungs and lower lungs in three types of data.

CONCLUSION

All methods are consistent with MC in the liver and tumors. Methods with tissue-specific kernel and effective correction achieve smaller errors in lungs. LiCKLuKD has comparable results with MC in absorbed dose estimation of Y-90 radioembolization for all target regions.

DblurDoseNet: A deep residual learning network for voxel radionuclide dosimetry compensating for single-photon emission computerized tomography imaging resolution

PURPOSE

Current methods for patient-specific voxel-level dosimetry in radionuclide therapy suffer from a trade-off between accuracy and computational efficiency. Monte Carlo (MC) radiation transport algorithms are considered the gold standard for voxel-level dosimetry but can be computationally expensive, whereas faster dose voxel kernel (DVK) convolution can be suboptimal in the presence of tissue heterogeneities. Furthermore, the accuracies of both these methods are limited by the spatial resolution of the reconstructed emission image. To overcome these limitations, this paper considers a single deep convolutional neural network (CNN) with residual learning (named DblurDoseNet) that learns to produce dose-rate maps while compensating for the limited resolution of SPECT images.

METHODS

We trained our CNN using MC-generated dose-rate maps that directly corresponded to the true activity maps in virtual patient phantoms. Residual learning was applied such that our CNN learned only the difference between the true dose-rate map and DVK dose-rate map with density scaling. Our CNN consists of a 3D depth feature extractor followed by a 2D U-Net, where the input was 11 slices (3.3 cm) of a given Lu-177 SPECT/CT image and density map, and the output was the dose-rate map corresponding to the center slice. The CNN was trained with nine virtual patient phantoms and tested on five different phantoms plus 42 SPECT/CT scans of patients who underwent Lu-177 DOTATATE therapy.

RESULTS

When testing on virtual patient phantoms, the lesion/organ mean dose-rate error and the normalized root mean square error (NRMSE) relative to the ground truth of the CNN method was consistently lower than DVK and MC, when applied to SPECT images. Compared to DVK/MC, the average improvement for the CNN in mean dose-rate error was 55%/53% and 66%/56%; and in NRMSE was 18%/17% and 10%/11% for lesion and kidney regions, respectively. Line profiles and dose-volume histograms demonstrated compensation for SPECT resolution effects in the CNN-generated dose-rate maps. The ensemble noise standard deviation, determined from multiple Poisson realizations, was improved by 21%/27% compared to DVK/MC. In patients, potential improvements from CNN dose-rate maps compared to DVK/MC were illustrated qualitatively, due to the absence of ground truth. The trained residual CNN took about 30 s on a single GPU (Tesla V100) to generate a 512 × 512 × 130 dose-rate map for a patient.

CONCLUSION

The proposed residual CNN, trained using phantoms generated from patient images, has potential for real-time patient-specific dosimetry in clinical treatment planning due to its demonstrated improvement in accuracy, resolution, noise, and speed over the DVK/MC approaches.

Clinical Evaluation of a Three-Dimensional Internal Dosimetry Technique for Liver Radioembolization with 90Y Microspheres Using Dose Voxel Kernels

Background: The purpose of this study was to develop a rapid, reliable, and efficient tool for three-dimensional (3D) dosimetry treatment planning and post-treatment evaluation of liver radioembolization with ⁹⁰Y microspheres, using tissue-specific dose voxel kernels (DVKs) that can be used in everyday clinical practice. Materials and Methods: Two tissue-specific DVKs for ⁹⁰Y were calculated through Monte Carlo (MC) simulations. DVKs for the liver and lungs were generated, and the dose distribution was compared with direct MC simulations. A method was developed to produce a 3D dose map by convolving the calculated DVKs with the activity biodistribution derived from clinical single-photon emission computed tomography (SPECT) or positron emission tomography (PET) images. Image registration for the SPECT or PET images with the corresponding computed tomography scans was performed before dosimetry calculation. The authors first compared the DVK convolution dosimetry with a direct full MC simulation on an XCAT anthropomorphic phantom. They then tested it in 25 individual clinical cases of patients who underwent ⁹⁰Y therapy. All MC simulations were carried out using the GATE MC toolkit. Results: Comparison of the measured absorbed dose using tissue-specific DVKs and direct MC simulation on 25 patients revealed a mean difference of 1.07% ± 1.43% for the liver and 1.03% ± 1.21% for the tumor tissue, respectively. The largest difference between DVK convolution and full MC dosimetry was observed for the lung tissue (10.16% ± 1.20%). The DVK statistical uncertainty was <0.75% for both media. Conclusions: This semiautomatic algorithm is capable of performing rapid, accurate, and efficient 3D dosimetry. The proposed method considers tissue and activity heterogeneity using tissue-specific DVKs. Furthermore, this method provides results in <1 min, making it suitable for everyday clinical practice.

Dosimetry for Optimized, Personalized Radiopharmaceutical Therapy

Radiopharmaceutical therapy (RPT) has grown rapidly over the last decade for treatment of numerous cancer types. Dosimetric guidance, as with other radiotherapy modalities, has benefitted patients by reducing the incidence of side effects and improving overall survival in populations treated under this paradigm. Development of tools and techniques for dosimetry-guided therapy is ongoing, with numerous the Food and Drug Administration-cleared products reaching the U.S. market in 2019. Safe use of commercial dosimetry platforms requires a deep understanding of the underlying physical principles and thoroughly vetted input data. Likewise, interpretation of dosimetry results relies on an understanding of radiobiological principles, and the principles of uncertainty propagation. In this article, we review strategies commonly employed for dosimetry-guided RPT - including quantitative imaging, dose calculation methods, and modeling of dose across time-points. Additionally, we review recent literature evidence (2013-2020) demonstrating the efficacy of personalized RPT.

Dose voxel kernel prediction with neural networks for radiation dose estimation

BACKGROUND

Currently there is an ever increasing interest in Lu-177 targeted radionuclide therapies, which target neuro-endocrine and prostate tumours. For a patient-specific treatment, an individual dosimetry based on SPECT/CT imaging is necessary. The aim of this study is to introduce a dosimetry method, where dose voxel kernels (DVK) are predicted by a neural network.

METHODS

Kidneys are considered one of the most critical organs in any radionuclide therapy. Hence we chose kidneys of 26 patients, who underwent Lu-177-DOTATOC or PSMA therapy, as target organs for our dosimetric method. First of all, density kernels with a size of 9×9×9 voxels were considered, and the corresponding DVKs were calculated by Monte Carlo simulations. These kernels were used to train a neural network (NN), which received a density kernel as input and predicted a DVK at the output. To predict the dose distribution of an entire kidney, the organ had to be partitioned into a large number of density kernels. Afterwards the DVKs were predicted by a trained NN, and employed to reconstruct the whole-organ dose distribution by convolution with the activity distribution. This method was compared to the standard method where the activity distribution is convolved with a DVK based on a homogeneous soft tissue kernel.

RESULTS

The number of training kernels amounted to 52,274 density kernels with corresponding MC-derived DVKs. The results serve as proof of principle of the newly proposed method and showed that the NN approach yielded superior results compared to the standard method with no additional computational effort.

CONCLUSION

The NN approach is an accurate and highly competitive dosimetric method to precisely estimate absorbed radiation dose in critical organs like kidneys in clinical routine. To further improve the results, a larger number of DVKs needs to be computed by Monte Carlo simulations. An extension of the method to other organs is easily conceivable.