FoCa: a modular treatment planning system for proton radiotherapy with research and educational purposes

A deep-learning-based surrogate model for Monte-Carlo simulations of the linear energy transfer in primary brain tumor patients treated with proton-beam radiotherapy

Objective.This study explores the use of neural networks (NNs) as surrogate models for Monte-Carlo (MC) simulations in predicting the dose-averaged linear energy transfer (LETd) of protons in proton-beam therapy based on the planned dose distribution and patient anatomy in the form of computed tomography (CT) images. As LETdis associated with variability in the relative biological effectiveness (RBE) of protons, we also evaluate the implications of using NN predictions for normal tissue complication probability (NTCP) models within a variable-RBE context.Approach.The predictive performance of three-dimensional NN architectures was evaluated using five-fold cross-validation on a cohort of brain tumor patients (n= 151). The best-performing model was identified and externally validated on patients from a different center (n= 107). LETdpredictions were compared to MC-simulated results in clinically relevant regions of interest. We assessed the impact on NTCP models by leveraging LETdpredictions to derive RBE-weighted doses, using the Wedenberg RBE model.Main results.We found NNs based solely on the planned dose distribution, i.e. without additional usage of CT images, can approximate MC-based LETddistributions. Root mean squared errors (RMSE) for the median LETdwithin the brain, brainstem, CTV, chiasm, lacrimal glands (ipsilateral/contralateral) and optic nerves (ipsilateral/contralateral) were 0.36, 0.87, 0.31, 0.73, 0.68, 1.04, 0.69 and 1.24 keV µm-1, respectively. Although model predictions showed statistically significant differences from MC outputs, these did not result in substantial changes in NTCP predictions, with RMSEs of at most 3.2 percentage points.Significance.The ability of NNs to predict LETdbased solely on planned dose distributions suggests a viable alternative to compute-intensive MC simulations in a variable-RBE setting. This is particularly useful in scenarios where MC simulation data are unavailable, facilitating resource-constrained proton therapy treatment planning, retrospective patient data analysis and further investigations on the variability of proton RBE.

Deep learning based linear energy transfer calculation for proton therapy

Objective.This study aims to address the limitations of traditional methods for calculating linear energy transfer (LET), a critical component in assessing relative biological effectiveness (RBE). Currently, Monte Carlo (MC) simulation, the gold-standard for accuracy, is resource-intensive and slow for dose optimization, while the speedier analytical approximation has compromised accuracy. Our objective was to prototype a deep-learning-based model for calculating dose-averaged LET (LETd) using patient anatomy and dose-to-water (DW) data, facilitating real-time biological dose evaluation and LET optimization within proton treatment planning systems.Approach. 275 4-field prostate proton Stereotactic Body Radiotherapy plans were analyzed, rendering a total of 1100 fields. Those were randomly split into 880, 110, and 110 fields for training, validation, and testing. A 3D Cascaded UNet model, along with data processing and inference pipelines, was developed to generate patient-specific LETddistributions from CT images and DW. The accuracy of the LETdof the test dataset was evaluated against MC-generated ground truth through voxel-based mean absolute error (MAE) and gamma analysis.Main results.The proposed model accurately inferred LETddistributions for each proton field in the test dataset. A single-field LETdcalculation took around 100 ms with trained models running on a NVidia A100 GPU. The selected model yielded an average MAE of 0.94 ± 0.14 MeV cm-1and a gamma passing rate of 97.4% ± 1.3% when applied to the test dataset, with the largest discrepancy at the edge of fields where the dose gradient was the largest and counting statistics was the lowest.Significance.This study demonstrates that deep-learning-based models can efficiently calculate LETdwith high accuracy as a fast-forward approach. The model shows great potential to be utilized for optimizing the RBE of proton treatment plans. Future efforts will focus on enhancing the model's performance and evaluating its adaptability to different clinical scenarios.

In vivo production of fluorine-18 in a chicken egg tumor model of breast cancer for proton therapy range verification

Range verification of clinical protontherapy systems via positron-emission tomography (PET) is not a mature technology, suffering from two major issues: insufficient signal from low-energy protons in the Bragg peak area and biological washout of PET emitters. The use of contrast agents including ¹⁸O, ⁶⁸Zn or ⁶³Cu, isotopes with a high cross section for low-energy protons in nuclear reactions producing PET emitters, has been proposed to enhance the PET signal in the last millimeters of the proton path. However, it remains a challenge to achieve sufficient concentrations of these isotopes in the target volume. Here we investigate the possibilities of ¹⁸O-enriched water (18-W), a potential contrast agent that could be incorporated in large proportions in live tissues by replacing regular water. We hypothesize that 18-W could also mitigate the problem of biological washout, as PET (¹⁸F) isotopes created inside live cells would remain trapped in the form of fluoride anions (F-), allowing its signal to be detected even hours after irradiation. To test our hypothesis, we designed an experiment with two main goals: first, prove that 18-W can incorporate enough ¹⁸O into a living organism to produce a detectable signal from ¹⁸F after proton irradiation, and second, determine the amount of activity that remains trapped inside the cells. The experiment was performed on a chicken embryo chorioallantoic membrane tumor model of head and neck cancer. Seven eggs with visible tumors were infused with 18-W and irradiated with 8-MeV protons (range in water: 0.74 mm), equivalent to clinical protons at the end of particle range. The activity produced after irradiation was detected and quantified in a small-animal PET-CT scanner, and further studied by placing ex-vivo tumours in a gamma radiation detector. In the acquired images, specific activity of ¹⁸F (originating from 18-W) could be detected in the tumour area of the alive chicken embryo up to 9 h after irradiation, which confirms that low-energy protons can indeed produce a detectable PET signal if a suitable contrast agent is employed. Moreover, dynamic PET studies in two of the eggs evidenced a minimal effect of biological washout, with 68% retained specific ¹⁸F activity at 8 h after irradiation. Furthermore, ex-vivo analysis of 4 irradiated tumours showed that up to 3% of oxygen atoms in the targets were replaced by ¹⁸O from infused 18-W, and evidenced an entrapment of 59% for specific activity of ¹⁸F after washing, supporting our hypothesis that F- ions remain trapped within the cells. An infusion of 18-W can incorporate ¹⁸O in animal tissues by replacing regular water inside cells, producing a PET signal when irradiated with low-energy protons that could be used for range verification in protontherapy. ¹⁸F produced inside cells remains entrapped and suffers from minimal biological washout, allowing for a sharper localization with longer PET acquisitions. Further studies must evaluate the feasibility of this technique in dosimetric conditions closer to clinical practice, in order to define potential protocols for its use in patients.

Engaging medical physics students in active and authentic learning through the use of monte-carlo simulation and inverse treatment planning

PURPOSE

Engagement and participation of students with the learning process has been recognised as a growing problem across the higher education sector. The aim of this study was to investigate the value and impact of introducing Problem-based Learning (PBL) activities into a radiotherapy physics unit of a postgraduate medical physics course.

METHODS

Computer-based problem solving activities on 1) monte-carlo modelling of a linear accelerator and 2) inverse radiotherapy treatment planning were designed and implemented into a one-semester unit on radiotherapy physics. The value and impact of the activities on the student learning were evaluated through student surveys, a focus group, and peer observation of the sessions by members of the learning design team. Student attendance and grade profile data are also reported.

RESULTS

Overall the results indicated that students had a positive experience with the new problem solving activities that were implemented. Survey responses from a number of students indicated a desire for increased theoretical and technical support prior to and during activities. Another underlying theme that emerged from survey and focus group response was the perceived lack of reward in terms of marks for their efforts working on the learning activities. This may have influenced students' choices around attendance and participation. No significant changes were noted in the overall grades achieved in the unit.

CONCLUSIONS

Students appreciated the more hands-on approach to learning in the form of more authentic activities that they could directly relate to clinical radiotherapy. Further work is required to update and integrate assessment into the new learning delivery model more directly.

Biological and Mechanical Synergies to Deal With Proton Therapy Pitfalls: Minibeams, FLASH, Arcs, and Gantryless Rooms

Proton therapy has advantages and pitfalls comparing with photon therapy in radiation therapy. Among the limitations of protons in clinical practice we can selectively mention: uncertainties in range, lateral penumbra, deposition of higher LET outside the target, entrance dose, dose in the beam path, dose constraints in critical organs close to the target volume, organ movements and cost. In this review, we combine proposals under study to mitigate those pitfalls by using individually or in combination: (a) biological approaches of beam management in time (very high dose rate “FLASH” irradiations in the order of 100 Gy/s) and (b) modulation in space (a combination of mini-beams of millimetric extent), together with mechanical approaches such as (c) rotational techniques (optimized in partial arcs) and, in an effort to reduce cost, (d) gantry-less delivery systems. In some cases, these proposals are synergic (e.g., FLASH and minibeams), in others they are hardly compatible (mini-beam and rotation). Fixed lines have been used in pioneer centers, or for specific indications (ophthalmic, radiosurgery,…), they logically evolved to isocentric gantries. The present proposals to produce fixed lines are somewhat controversial. Rotational techniques, minibeams and FLASH in proton therapy are making their way, with an increasing degree of complexity in these three approaches, but with a high interest in the basic science and clinical communities. All of them must be proven in clinical applications.

Fast robust dose calculation on GPU for high-precision 1H, 4He, 12C and 16O ion therapy: the FRoG platform

Radiotherapy with protons and heavier ions landmarks a novel era in the field of high-precision cancer therapy. To identify patients most benefiting from this technologically demanding therapy, fast assessment of comparative treatment plans utilizing different ion species is urgently needed. Moreover, to overcome uncertainties of actual in-vivo physical dose distribution and biological effects elicited by different radiation qualities, development of a reliable high-throughput algorithm is required. To this end, we engineered a unique graphics processing unit (GPU) based software architecture allowing rapid and robust dose calculation. FRoG, Fast Recalculation on GPU, currently operates with four particle beams available at Heidelberg Ion Beam Therapy center, i.e., raster-scanning proton (¹H), helium (⁴He), carbon (¹²C) and oxygen ions (¹⁶O). FRoG enables comparative analysis of different models for estimation of physical and biological effective dose in 3D within minutes and in excellent agreement with the gold standard Monte Carlo (MC) simulation. This is a crucial step towards development of next-generation patient specific radiotherapy.

Teaching treatment planning for protons with educational open-source software: experience with FoCa and matRad

Open‐source, MATLAB‐based treatment planning systems FoCa and matRAD were used in a pilot project for training prospective medical physicists and postgraduate physics students in treatment planning and beam modeling techniques for proton therapy. In the four exercises designed, students learnt how proton pencil beams are modeled and how dose is calculated in three‐dimensional voxelized geometries, how pencil beam scanning plans (PBS) are constructed, the rationale behind the choice of spot spacing in patient plans, and the dosimetric differences between photon IMRT and proton PBS plans. Sixty students of two courses participated in the pilot project, with over 90% of satisfactory rating from student surveys. The pilot experience will certainly be continued.

Performance of a hybrid Monte Carlo-Pencil Beam dose algorithm for proton therapy inverse planning

PURPOSE

Analytical algorithms have a limited accuracy when modeling very heterogeneous tumor sites. This work addresses the performance of a hybrid dose optimizer that combines both Monte Carlo (MC) and pencil beam (PB) dose engines to get the best trade-off between speed and accuracy for proton therapy plans.

METHODS

The hybrid algorithm calculates the optimal spot weights (w) by means of an iterative optimization process where the dose at each iteration is computed by using a precomputed dose influence matrix based on the conventional PB plus a correction term c obtained from a MC simulation. Updates of c can be triggered as often as necessary by calling the MC dose engine with the last corrected values of w as input. In order to analyze the performance of the hybrid algorithm against dose calculation errors, it was applied to a simplistic water phantom for which several test cases with different errors were simulated, including proton range uncertainties. Afterwards, the algorithm was used in three clinical cases (prostate, lung, and brain) and benchmarked against full MC-based optimization. The influence of different stopping criteria in the final results was also investigated.

RESULTS

The hybrid algorithm achieved excellent results provided that the estimated range in a homogeneous material is the same for the two dose engines involved, i.e., PB and MC. For the three patient cases, the hybrid plans were clinically equivalent to those obtained with full MC-based optimization. Only a single update of c was needed in the hybrid algorithm to fulfill the clinical dose constraints, which represents an extra computation time to obtain c that ranged from 1 (brain) to 4 min (lung) with respect to the conventional PB-based optimization, and an estimated average gain factor of 14 with respect to full MC-based optimization.

CONCLUSION

The hybrid algorithm provides an improved trade-off between accuracy and speed. This algorithm can be immediately considered as an option for improving dose calculation accuracy of commercial analytical treatment planning systems, without a significant increase in the computation time (≪5 min) with respect to current PB-based optimization.

Range optimization for mono- and bi-energetic proton modulated arc therapy with pencil beam scanning

The development of rotational proton therapy plans based on a pencil-beam-scanning (PBS) system has been limited, among several other factors, by the energy-switching time between layers, a system-dependent parameter that ranges between a fraction of a second and several seconds. We are investigating mono- and bi-energetic rotational proton modulated arc therapy (PMAT) solutions that would not be affected by long energy switching times. In this context, a systematic selection of the optimal proton energy for each arc is vital. We present a treatment planning comparison of four different range selection methods, analyzing the dosimetric outcomes of the resulting treatment plans created with the ranges obtained. Given the patient geometry and arc definition (gantry and couch trajectories, snout elevation) our in-house treatment planning system (TPS) FoCa was used to find the maximum, medial and minimum water-equivalent thicknesses (WETs) of the target viewed from all possible field orientations. Optimal ranges were subsequently determined using four methods: (1) by dividing the max/min WET interval into equal steps, (2) by taking the average target midpoints from each field, (3) by taking the average WET of all voxels from all field orientations, and (4) by minimizing the fraction of the target which cannot be reached from any of the available angles. After the range (for mono-energetic plans) or ranges (for bi-energetic plans) were selected, the commercial clinical TPS in use in our institution (Varian Eclipse^™) was used to produce the PMAT plans using multifield optimization. Linear energy transfer (LET) distributions of all plans were also calculated using FoCa and compared among the different methods. Mono- and bi-energetic PMAT plans, composed of a single 180° arc, were created for two patient geometries: a C-shaped target located in the mediastinal area of a thoracic tissue-equivalent phantom and a small brain tumor located directly above the brainstem. All plans were optimized using the same procedure to (1) achieve target coverage, (2) reduce dose to OAR and (3) limit dose hot spots in the target. Final outcomes were compared in terms of the resulting dose and LET distributions. Data shows little significant differences among the four studied methods, with superior results obtained with mono-energetic plans. A streamlined systematic method has been implemented in an in-house TPS to find the optimal range to maximize target coverage with rotational mono- or bi-energetic PBS rotational plans by minimizing the fraction of the target that cannot be reached by any direction.