Abstract
Dosimetry is integral to informed implementation of radiopharmaceutical therapies, enabling personalized treatment planning and ensuring patient safety by calculating absorbed doses to organs and tumors. As the therapeutic radiopharmaceutical field continues to expand, dosimetry software has emerged as a crucial tool for optimization of treatment efficacy. This review discusses key features and capabilities that current dosimetry software solutions have or should have in the future. We highlight the need for standardization across platforms to support consistent and accurate dose calculations. Furthermore, we explore opportunities for advancing software, such as incorporating biologically effective dose modeling and improving uncertainty quantification. Looking ahead, we advocate for expanding infrastructure for open data sets and fostering ongoing collaboration between vendors and end users to guide the field toward greater integration and efficacy.
Radiopharmaceutical therapy (RPT) is a rapidly growing field. In current clinical practice, most dosing regimens are based on fixed or empirically determined activities, neglecting the opportunity to personally tailor patient treatments with dosimetry. With dosimetry guidance, the treatment can be optimized for disease control and the absorbed doses to critical organs such as bone marrow and kidneys verified to ensure that absorbed dose thresholds are not exceeded (1).
With the recent arrival of well-developed dosimetry software applications, we are seeing a growing body of literature illuminating the potential impact of dosimetry. Concerning efficacy, some recent studies have provided evidence on the relationship between tumor-absorbed dose and the treatment outcome (2–5). Other studies have also reported on the dose–effect relationship between absorbed dose and hematotoxicity (6–8). For patients treated with 177Lu-labeled compounds, no threshold for kidney-absorbed dose for a given level of toxicity has been observed, indicating a potential for increasing administered activities and delivering higher doses to lesions (9).
Dosimetry for 177Lu-labeled applications is still not routinely performed in many nuclear medicine departments, despite the availability of regulatory-cleared (United States) and Conformité Européenne–marked (European Union) software solutions. The dosimetry software solutions available can differ considerably in style, scope, and function (10–12). The aim of this review is to provide a general overview on the features that are expected to be part of a dosimetry software solution and to suggest areas for improvement or enhancement for future development of such codes.
BEST DOSIMETRY PRACTICES
With growing interest in the clinical implementation of RPT dosimetry, it is essential to establish structured methodologies that can accommodate varying levels of rigor across different health care settings. The practice of calculating and reporting RPT has multiple steps. Resources, expertise, and procedures vary, and this variability propagates divergence in the field. In response, efforts have been made to bring structure across the field, such as MIRD pamphlet no. 16 (13) and the European Association of Nuclear Medicine dosimetry committee guidance for good practice of clinical dosimetry reporting (14). Further, International Commission on Radiation Units and Measurements report 96 (15) identified 3 categoric reporting levels for dosimetry in RPTs. Level 1 is typically implemented on the basis of the activity administered. The mean absorbed dose to the dosimetric treatment region and the region at risk should be reported, as appropriate to the treatment intent. In addition, the time–activity curve and the fitting function to determine the activity biodistribution, if applicable, should be reported. Level 2 recommendations apply to the prescription and reporting of state-of-the-art techniques so that the calculation of target volume exposure is patient-specific and meets prespecified uncertainty criteria. Depending on the agent, clinical circumstances, and ability to accurately collect the necessary data, this level should include adjustments to the prescribed activity based on the absorbed dose to the dosimetric treatment region and the region at risk. Level 3 focuses on creating and refining techniques for which reporting criteria are not yet standardized, with the goal of applying and improving them in clinical trials. At all levels, the reports generated should include all inputs to the calculations used to arrive at the prescribed absorbed dose or administered activity.
Therapeutic radiopharmaceuticals are regulated to ensure safety in handling and administration, though this oversight often neglects optimal efficacy. In the United States, the Food and Drug Administration approves these therapies, whereas the Nuclear Regulatory Commission and state programs oversee facility licensing and safety protocols. International bodies such as the European Medicines Agency and the International Atomic Energy Agency set global standards for radiation protection and quality assurance. However, these frameworks prioritize safety over clinical outcomes, overlooking the potential of dosimetry-driven, patient-tailored treatments. By advancing and adopting better dosimetry software, we have the opportunity to bridge this gap, improving both safety and efficacy in RPT.
Significant motivation for advancing dosimetry solutions has come from European legislation. In Europe, two legal frameworks govern the use of therapeutic radiopharmaceuticals: the pharmaceutical legislation (directive 2001/83/EC) and the radiation protection legislation (directive 2013/59/Euratom). Efforts on a European level are presently under way, to propose and implement solutions that reconcile the requirements of the different legal frameworks (www.simplerad.eu).
For implementing dosimetry in clinical practice, either to be compliant with International Commission on Radiation Units and Measurements report 96 or to be in accordance with the European legislation, the availability of robust and reliable software solutions for dosimetry is a prerequisite.
DOSIMETRY WORKFLOW
All personalized internal dosimetry is derived by projecting biodistribution measurements into absorbed dose calculations (Fig. 1). Much of the conceptual basis for internal dosimetry calculations was established in the MIRD formalism, a framework based on assigning source and target regions and summing the dose contributions across these regions (16). Proper application requires a thorough accounting of the spatiotemporal distribution of a radiotracer within the body, including its excretion patterns, to define the radiation introduced into the system. Dosimetry can then be calculated in target regions of interest using mathematic models of dose deposition. This conceptual framework is flexible enough to accommodate calculations at various scales, from micrometers (cellular level) to millimeters (voxel level) and centimeters (organ level).
Typically required steps for absorbed dose calculations within (blue) or optional (gray) within dosimetry software.
Organ-Level Dosimetry
The standard approach to dosimetry calculations for RPTs over the past decade has been to implement dosimetry at an organ level. Following the MIRD approach, the mean absorbed dose to an organ or tissue is calculated by multiplying the time-integrated activities (integral of the time–activity curve) in the source regions with the absorbed dose rate per unit activity factor (also called S values) for the source-target region pair (17). These factors are usually derived from Monte Carlo simulations modeling the energy deposition in stylized phantoms. For many radionuclides used for RPTs today, this approach is still valid provided proper organ and tissue mass corrections are made. However, differences may occur for organs or tissues with unusual geometries or other nonspheric small structures.
NOTEWORTHY
Dosimetry software has emerged as a key tool for optimization of treatment efficacy.
Key features and capabilities are described that current dosimetry software solutions have or should have in the future.
There is a need for standardization across dosimetry platforms to support consistent and accurate dose calculations.
Voxel-Level Dosimetry
Voxel-level dosimetry is a method used to calculate radiation dose distributions with high spatial precision by estimating absorbed doses for individual 3-dimensional units, or voxels, within the body (18) by taking account of individual-patient time-integrated activity distributions. The methodology integrates somewhat elegantly with radiology and radiotherapy applications, where tracer distribution and anatomic data are already inherently represented at the voxel level, aligning with how clinicians typically interpret anatomy. Voxel-level dosimetry also supports the visualization of dose maps overlaid on anatomic images, offering a detailed, personalized view of radiation distribution within tumors and surrounding tissues. This allows for more precise assessments of therapeutic efficacy and potential toxicity. Currently, commercial dosimetry software vendors favor and support voxel-level calculation because of its detailed approach. However, voxel-level dosimetry has limitations, such as challenges related to patient motion, limited statistical accuracy in dose calculations, and partial-volume effects (PVEs) that can affect precision in small structures.
Bone Marrow Dosimetry
For nonsolid organs such as the bone marrow, which are nonuniformly distributed throughout the body, traditional organ- and voxel-level dosimetry approaches may be inadequate. It is therefore common to use a surrogate measurement or hybrid dosimetry approach. Whole-body measurements acquired using external probes have been proposed as a simple approach that can correlate with hemotoxicity in specific RPTs (19,20). Alternatively, surrogate measurements of blood samples provide a more direct measurement of the primary source of activity irradiating the marrow (21–23). Quantitative imaging on a representative boney region may also support marrow dosimetry, focusing on regions where a large proportion of red marrow is likely situated, such as the lumbar spine or sacrum. In some cases, it may be necessary to combine data from the different sources and use data from both the external counting approaches and quantitative imaging.
90Y/166Ho Microsphere Treatment Planning
Although not classified as traditional RPT, radioembolization of liver malignancies using microspheres labeled with 90Y or 166Ho is often considered alongside RPTs. This is because the millions of microspheres act similarly to an unsealed liquid source and are managed within nuclear medicine departments. Their dosimetry is calculated with the aid of surrogate imaging and imaging-based platforms. However, unlike RPTs, this therapy does not involve radiotracer pharmacokinetics; rather, the radiation-emitting microspheres get lodged in the tumor vasculature and only the initial distribution of spheres and the isotope’s physical decay need to be considered with respect to dosimetry. Notably, these treatment modalities consistently incorporate dosimetry-based treatment planning for every patient.
Treatment activities are calculated using model-based dosimetry calculators, which range in complexity and include empiric MIRD single-compartment and MIRD multicompartment dose calculation models. The latter uses surrogate 99mTc-macroaggregated albumin imaging to aid in projecting anticipated multicompartment distribution. A comprehensive summary of these microsphere treatment planning models is provided in the European Association of Nuclear Medicine standard operational procedure on 99mTc-macroaggregated albumin pretherapy dosimetry and 90Y peritherapy dosimetry in liver radioembolization with 90Y microspheres (24) and in MIRD pamphlet no. 29 (25).
SOFTWARE FEATURES
Highly Recommended
The software features listed in this section are related to the most pertinent aspects of the dosimetry workflow. Their integration and adoption in dosimetry software are consequently highly endorsed.
Digital Imaging and Communications in Medicine (DICOM) Data Input
Dosimetry requires detailed information about patient morphology and isotope tracer distribution. These data, along with crucial patient and scan metadata, are provided in standard DICOM format for PET, SPECT, CT, and MR images. Dosimetry software should facilitate users in navigating DICOM file databases and importing the relevant files into the dosimetry system.
Image Calibration Factor
SPECT images do not inherently contain calibration information. Traditionally, SPECT is acquired in units of relative counts, although it is becoming more common for vendors to offer protocols that integrate quantification into image reconstruction. Various methods can be used for image calibration, and dosimetry software users must be able to input and correct calibration factors as needed for specific input data.
Administered Activity and Time of Administration
Recording the administration details is a vital aspect of the dosimetry calculation. Unfortunately, this information is not always recorded in DICOM metadata. Therefore, users need the ability to enter or modify administered activity and administration time. Ideally, software would have functionality to apply decay correction, correct the administered activity from a given measurement time, and allow residual activities to be set.
Image Segmentation
An important aspect of the dosimetry workflow is the segmentation of images to delineate dosimetric regions of interest. Users must have controls to navigate images and assist in segmenting volumes of interest (VOIs), either manually or with semiautomated methods or artificial intelligence–assisted segmentation tools. Functionality to navigate coregistered hybrid modalities is particularly important to merge radiopharmaceutical distributions with superior anatomic information. Users must therefore be able to segment VOIs on any single image and propagate these volumes to registered images. Additionally, the ability to propagate VOIs across images from different time points is desirable, alongside the means to edit and fine-tune the VOIs at the different time points.
Time-Integrated Activity Modeling
Radiopharmaceutical dosimetry requires modeling of both the spatial and the temporal distribution of radioactivity. Temporal modeling is achieved through curve fitting. Typically, data are organized by plotting either the measured activity or the measured dose rate versus time for a given region—organ or voxel. The time-integrated activity or absorbed dose is then generated by integrating these data. Analytic integration is achieved by fitting continuous data models to the available data. Radiopharmaceutical biokinetics are often adequately described by simple sums of exponential functions; for example, mono- or biexponential curves are frequently used to represent single- or dual-compartment clearance. Simpler approaches, such as trapezoidal integration, may also be used. The ideal fit function can vary depending on various factors, so dosimetry software should ideally offer flexibility in the fitting algorithm.
Absorbed Dose Calculation
There are several computational models for absorbed dose calculation, including organ dose factor (S value)–based calculation with and without mass scaling of the S values (26); Voxel S values or dose-point kernel convolution (27), with and without CT-based density scaling (28); and individual-patient Monte Carlo simulation (29). The different methods balance speed and accuracy. It is important for software users to understand the dosimetry algorithm being used within software and its limitations (30). However, when implemented and applied correctly, the choice of dosimetry algorithm should have minimal impact on mean organ- or lesion-absorbed doses.
Dose–Volume Histograms
A useful graphical representation for summarizing the heterogeneity of absorbed dose calculated at the voxel level is the dose–volume histogram (DVH). A DVH is created by plotting the percentage of the VOI against the absorbed dose it receives, a concept borrowed from its widespread use in external-beam radiation therapy (EBRT). The utility of a DVH is its ability to succinctly summarize a large amount of data in a clear 2-dimensional chart. Dosimetry software calculating at the voxel level should support the generation of DVHs. Summary metrics such as D90, D50, or other Dx values (31), where Dx represents the dose received by x% of the target volume, should also be provided to allow the user to determine the level of radiation dose received to a given percentage of the VOI.
DICOM Saving
The DICOM file format is universally used in health care settings and is supported by modern PACS systems for storing, retrieving, and managing digital images and related data. Modern dosimetry software should provide output results as both numeric data and images in DICOM format to support proper archiving.
Saving Output Data
Analysis of dosimetry data is often required outside the dosimetry application. It is therefore useful if the dosimetry software can provide output data (VOIs, dosimetry results, DVH metrics, etc.) in a common data readable format for example, such as a comma-separated variable file.
Saving Output Reports
Dosimetry information should be organized into a standard report format that can be exported and shared. Since local clinical and regulatory requirements can necessitate specific information to be included in reports, it is ideal for manufacturers to support local customization.
Recommended
Advisable features that are not currently available within all existing applications include the following:
Flexible Input
Image-based dosimetry may use a variety of imaging protocols. These include variation in the number and periodicity of acquired time points and the imaging technology used (PET, SPECT, or planar imaging). Hybrid SPECT and planar imaging protocols are also common (32), which require a different workflow than purely SPECT- or PET-based approaches. A software application that allows for flexibility in image input is therefore desirable and broadens its scope of application.
Image Registration and Segmentation
The registration of images to a common reference dataset is advantageous for a common set of VOIs to be used across time points and modalities and is a prerequisite for voxel-level dose calculations. Ideally, several registration options would be available: automatic, semiautomatic, (based on landmarks), or manual. Rigid registration preserves relative distances on the images using only image shifts and rotations, whereas deformable registration transforms the original image (33). Whichever method is used, it is important that users have the ability to validate and further refine the registration on the basis of their requirements.
Workflows and Saving of Intermediate Results
A common function available in most computer applications, but not currently available in all dosimetry software, is the ability to save individual steps within a workflow. This is useful not only for returning to a specific task when interrupted but also for verifying certain steps by other professionals. Separate saving of the VOIs is important for quality control and for potential later use in subsequent therapy cycles. Saving intermediate results such as VOI-wise activities or time-integrated activities may aid in reliability verification.
PVE Correction
It is generally acknowledged that a postreconstruction correction for PVEs is necessary for accurate dosimetry (34). Conventionally, this is achieved using recovery coefficients derived from a set of phantom measurements. However, this is not currently available in all software. Such a correction is desirable as it avoids the need to later correct the absorbed dose or manipulate values outside the software.
Absorbed Dose Calculation
Multiple absorbed dose calculation approaches exist, and it is desirable that software support use of multiple models, to allow for intercomparison. Furthermore, it is important that the calculation models take into account all radionuclide decay products (βs, αs, photons). It is also often desirable to be able to separate the absorbed dose contributions from different decay products within a radionuclide decay chain. Extension of the absorbed dose to other dosimetric quantities is also of increasing importance. This is particularly true for metrics such as the equieffective dose or the biologically effective dose, which incorporate bioeffect modeling (15).
Error Propagation and Uncertainty Analysis
Lastly, error propagation and uncertainty analysis are useful to provide users with information about the uncertainty of the derived absorbed dose value. It would be beneficial for the field of nuclear medicine if routine error propagation were applied and uncertainty values reported.
Optional
Additional features may aid in improving the usability and impact of absorbed dose software, such as the following:
Quantitative Reconstruction
It can be advantageous to perform camera-vendor–neutral quantitative image reconstruction within the dosimetry software and provide activity images directly in units of becquerels or becquerels per milliliter for the dosimetry workflow.
Implementation of Blood-Based Dosimetry
To date, we have not seen blood-based dosimetry implemented in commercial dosimetry software, despite bone marrow being the common primary organ at risk. While bone marrow is not easily visualized in 3-dimensional voxel images, the calculation remains important to patient management.
Standardized Output
The output format of different software programs should ideally be standardized or harmonized to support integration into common databases. Such databases could enable design and generation of quick local reports according to local needs and preferences. They can also support large-scale review and analysis needed for impactful research and dose optimization.
Surrogate Nuclide Dosimetry
There are increasing efforts to explore the possibility of therapy planning using theranostic pairs. It is therefore likely that the need to adapt time–activity curves across different radionuclides will increase. Examples include the use of 99mTc-macroaggregated albumin for therapy simulation of liver radioembolization. Similarly, 111In or 68Ga could be used for 177Lu RPT dosimetry planning (35).
Patient Clinical History and Absorbed Dose Tracking
The inclusion of a patient’s history, especially the dosimetric history, can aid in adjusting the subsequent therapy cycles as well as determining potential dose–effect relationships for the individual patient.
Customer Support
A vendor helpline aids when users face issues with the dosimetry software ranging from implementation and application to potential bugs.
OPPORTUNITIES FOR IMPROVEMENT
Ongoing collaboration between vendors and end users is crucial to advancing dosimetry software and unlocking its full potential in clinical practice. Together, these groups can tackle key areas for improvement:
Standardizing Image Calibration
Dosimetry software requires quantified images for accurate calculations. In SPECT imaging, calibration factors are typically determined by comparing the activity derived from the image with the known activity in a controlled phantom scan. Various protocols for obtaining these calibration factors are used, with certain methods being preferred in different geographic regions (36–38) and by different vendors (39).
As traceable image calibration is one of the most crucial parts in the dosimetry calculation process, there remain significant scope and opportunity for the scientific community and industry to adopt a standardized method for image quantification. Optimally, one harmonized standard protocol should be adopted in the future. Nevertheless, for all dosimetry studies, traceability of the dosimetry to primary standards is important to facilitate proper comparison between studies.
Harmonizing Clinical Protocols
In recent years, international efforts have been undertaken to propose harmonized clinical protocols regarding image acquisition, optimal image acquisition time-points, time–activity curve fitting, and calculation of absorbed doses (15,40,41). However, these recommendations are still not rigorously applied in clinical trials or in daily practice. Of particular importance is how dosimetry is implemented in first-in-human and early-phase clinical trials. The application of the methods described in the European Association of Nuclear Medicine guidance document (40) might pave the way to including patient-specific dosimetry in the future use of licensed radiopharmaceuticals. As recommendations for harmonized clinical standard operating procedures are established, harmonized calculation methods implemented within dosimetry software would boost its impact.
Improving Image Reconstruction Consistency
Activity quantification is typically based on SPECT imaging and possibly using PET imaging. These 3-dimensional emission images are not perfect depictions but rather are tomographic representations of activity distributions derived from acquisition data. Unfortunately, not all reconstruction algorithms produce identical results. Various algorithms are used to optimize image quality for different imaging scenarios, but they vary in metrics of noise, resolution, and quantification characteristics. Additionally, many reconstruction algorithms are proprietary, making direct comparisons more challenging.
Research has shown that differences in image reconstruction can have a significant impact on dosimetric analysis (42). Although efforts have been made to standardize reconstruction methods for RPT dosimetry applications (43), variability in quantification due to differences in image reconstruction remains a known challenge in the field. Addressing this variability is crucial for improving standardization across the field in future developments.
Enhancing VOI Segmentation
Target delineation is often the most time-consuming aspect of the absorbed dose calculation and contributes a large source of uncertainty (44). Historically, region- and volume-outlining tools in nuclear medicine software have been fairly rudimentary, being based on simple threshold or freehand techniques. For dosimetry in RPT applications, more sophisticated anatomic delineation is often required for complex organ substructures and heterogeneous lesion distributions. Machine learning has had a dramatic impact on radiotherapy for automatic segmentation. Transferring these algorithms to RPT outlining is gaining favor in commercial applications. However, validation and quality assurance of these algorithms are still in their infancy, requiring a collaborative effort from both the scientific community and industry to establish best practices for the technology.
Implementing PVE Corrections
The PVE, which is due to the limited resolution of PET and SPECT scanners, consistently leads to the underestimation of focal activity distribution in images, particularly affecting small, dosimetrically relevant structures such as tumors. Despite advances in reconstruction algorithms and improved resolution recovery techniques, the sensitivity of absorbed dose calculations to PVEs compels additional corrections. International guidance recommends the use of recovery coefficients, which although not entirely accurate offer a reasonable means of modeling corrections for partial-volume losses (45). Alternative approaches to recovery coefficient–based PVE correction include using smaller-than-region VOIs to better characterize mean uptake in a region (46) or using larger-than-region VOIs to robustly estimate total region uptake (47).
At present, the application of PVC in dosimetry software is not harmonized. Methodologies for generating recovery coefficients via system recovery curves vary, and standard phantoms with suitably sized inserts are not widely available. Scientific societies should guide the software manufacturers toward a reliable implementation of standardized corrections. Current efforts are under way within the Society of Nuclear Medicine and Molecular Imaging MIRD committee to establish procedures and tools for harmonized application of PVC for RPT applications (48).
Advancing Curve-Fitting Tools
Most commercial systems do not allow the users to investigate the validity of the fitting process. Several methods on how to check the validity of the fitting function have been proposed (49). MIRDfit, a recently published biodistribution fitting software solution by the MIRD committee of the Society of Nuclear Medicine and Molecular Imaging, provides essential features and advantages for choosing nuclear medicine–appropriate fitting functions (50).
Because the time-integrated activity curve-fitting approach has a significant impact on the dose calculation, it would be helpful if vendors were to provide more curve-fitting options and flexibility in future commercial software.
Balancing Voxel- and Organ-Level Dosimetry
The scientific community has developed reliable methods for calculating dosimetry at both organ and voxel levels, each with distinct advantages and limitations. Voxel-level dosimetry is ideal for detailed dose assessments in nonstandard geometries or suborgan regions, whereas organ-level dosimetry is better suited for normal-organ evaluations and model-based calculations, such as bone marrow dosimetry. However, limitations for voxel-level calculations include resolution-related artifacts (29), and organ-level methods can be limited by inadequate assumptions in computational models. Moving forward, the field should focus on standardizing dosimetric calculation methods to match specific applications, balancing utility, accuracy, and practicality.
Integrating Error Propagation and Uncertainty Analysis
Another item that is not fully developed in commercial dosimetry software is the propagation of uncertainties in calculating the absorbed doses. Although in 2018 the European Association of Nuclear Medicine dosimetry committee provided practical guidance on uncertainty analysis for molecular radiotherapy absorbed dose calculations (45), none of the principles described in that publication have been implemented in the commercial systems that are presently available. It remains important that a realistic assessment of absorbed dose calculation uncertainties be provided to clinicians interpreting the absorbed doses. We can speculate that the field has been slow to embrace this call because of the additional efforts required for integration, but RPT software presents a potential structure for implementing such calculations in ways that require minimal additional effort from the user. For example, uncertainties in curve fitting can be calculated within curve fitting software as exemplified in MIRDfit (50) and NUKfit (51), and the uncertainties can automatically be propagated into dose calculation as exemplified in MIRDcalc (52).
Incorporating Bioeffect Modeling
The radiobiology of RPTs cannot be reliably extrapolated directly from EBRT frameworks because it differs in terms of absorbed dose, dose rate, exposure protraction, dose distribution, the potential biologic activity of the targeting molecule, and the use of radionuclides that emit various particles (β, α, Auger, γ/X) and of low- and high-linear-energy-transfer radiation mixtures (53). Therefore, before the radiobiology-related quantities are implemented in software solutions, they need to be verified specifically for RPTs.
Provided these bioeffects are well established, their modeling and the corresponding radiobiology-related quantities could be implemented into software solutions. How this could be done is described in detail in International Commission on Radiation Units and Measurements report 96 (15).
Benchmarking Dosimetry Software
Differences in absorbed doses can still occur despite extensive internal testing by developers and larger numbers of publications comparing available dosimetry software codes. We identify 2 major categories of potential failure: user-related and software-related.
Manual input by users within any of the steps associated with the dosimetry workflow is prone to errors. Typographic mistakes are common and can easily lead to errors in the dosimetry output. Software with automated input is consequently more robust against these mistakes. Users with less expertise and experience may make further mistakes, such as switching the left and right kidneys, using the wrong units, omitting mass scaling for S values, including lesions within organs for absorbed dose calculation of healthy organ tissues, or using different mean energies for radioisotope emissions for the local dose deposition method (54). To prevent such avoidable mistakes that lead to differences in absorbed dose, we suggest that dosimetry experts undergo mandatory training, including globally available educational resources and dosimetry schools.
Despite rigorous and intensive testing, both commercial and noncommercial dosimetry software may have internal bugs. With no access to the internal algorithms, it is often difficult to track and reveal erroneous results. Benchmarking from developers may not include testing of all individual features, or the used patient datasets may not be appropriate to—or may even be unable to—uncover the mistake.
Some efforts have been taken to compare the accuracy of quantitative 177Lu SPECT/CT imaging and results from dosimetry calculations. The combination of pharmacokinetic modeling with anthropomorphic 3-dimensional printed phantoms (55) seems to be promising to provide input data for validation of dosimetry software.
No real benchmarking process for comparing the performance of different software solutions is currently available, thus demonstrating the need to develop reliable benchmarking tools that can evaluate different dosimetry software and their versions systematically. Such tools could be based on, for example, freely available patient datasets to enable comparison with existing absorbed dose results (56).
Fostering Open Data Sets
The scientific community is increasingly embracing open data to drive innovation and collaboration. To support standardization and community cooperation, the field should actively promote the development, deployment, and accessibility of open RPT patient datasets (56), establishing a foundation for modern and continuous community-driven advancement. This approach aligns with broader scientific trends emphasizing reproducibility, transparency, and collective progress. For dosimetry software, open data can enhance benchmarking accuracy, ensuring reliable and standardized practices across institutions.
CONCLUSION
The development and use of dosimetry software in RPT are essential for enhancing patient outcomes and ensuring treatment safety. With the rapidly expanding use of therapeutic radiopharmaceuticals, there is an increasing need for precise and individualized treatment planning that considers patient-specific factors. Several challenges remain in the implementation of personalized dosimetry in clinical settings. The variability in available software solutions, in terms of both their calculation algorithms and their adherence to standardized protocols, underscores the need for further harmonization. Commercial dosimetry workflows conventionally focus on calculations at either the organ or voxel level. However, with nonsolid organs, such as the bone marrow, often being the dose-limiting organ, incorporation of nonimage data is also highly desirable. There are several key features that all dosimetry software programs should strive to incorporate. An important aspect is the standardization of output data, in terms of both report functionality and DICOM data. Moreover, developments not only should focus on enhancing software features and incorporating advanced dosimetric models but also should take place alongside efforts by the scientific community to harmonize clinical protocols and standardize dosimetry methodologies.
DISCLOSURE
Michael Lassmann has received institutional grants from Novartis and Pentixapharm. Adam Kesner has served as a paid consultant for Boston Scientific. No other potential conflict of interest relevant to this article was reported.
Footnotes
Published online Jan. 8, 2025.
- © 2025 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication October 16, 2024.
- Accepted for publication December 6, 2024.
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.