Embedding machine learning based toxicity models within radiotherapy treatment plan optimization.

Objective.This study addresses radiation-induced toxicity (RIT) challenges in radiotherapy (RT) by developing a personalized treatment planning framework. It leverages patient-specific data and dosimetric information to create an optimization model that limits adverse side effects using constraints learned from historical data.Approach.The study uses the optimization with constraint learning (OCL) framework, incorporating patient-specific factors into the optimization process. It consists of three steps: optimizing the baseline treatment plan using population-wide dosimetric constraints; training a machine learning (ML) model to estimate the patient's RIT for the baseline plan; and adapting the treatment plan to minimize RIT using ML-learned patient-specific constraints. Various predictive models, including classification trees, ensembles of trees, and neural networks, are applied to predict the probability of grade 2+ radiation pneumonitis (RP2+) for non-small cell lung (NSCLC) cancer patients three months post-RT. The methodology is assessed with four high RP2+ risk NSCLC patients, with the goal of optimizing the dose distribution to constrain the RP2+ outcome below a pre-specified threshold. Conventional and OCL-enhanced plans are compared based on dosimetric parameters and predicted RP2+ risk. Sensitivity analysis on risk thresholds and data uncertainty is performed using a toy NSCLC case.Main results.Experiments show the methodology's capacity to directly incorporate all predictive models into RT treatment planning. In the four patients studied, mean lung dose and V20 were reduced by an average of 1.78 Gy and 3.66%, resulting in an average RP2+ risk reduction from 95% to 42%. Notably, this reduction maintains tumor coverage, although in two cases, sparing the lung slightly increased spinal cord max-dose (0.23 and 0.79 Gy).Significance.By integrating patient-specific information into learned constraints, the study significantly reduces adverse side effects like RP2+ without compromising target coverage. This unified framework bridges the gap between predicting toxicities and optimizing treatment plans in personalized RT decision-making.

The influence of anisotropy on the clinical target volume of brain tumor patients.

Objective.Current radiotherapy guidelines for glioma target volume definition recommend a uniform margin expansion from the gross tumor volume (GTV) to the clinical target volume (CTV), assuming uniform infiltration in the invaded brain tissue. However, glioma cells migrate preferentially along white matter tracts, suggesting that white matter directionality should be considered in an anisotropic CTV expansion. We investigate two models of anisotropic CTV expansion and evaluate their clinical feasibility.Approach.To incorporate white matter directionality into the CTV, a diffusion tensor imaging (DTI) atlas is used. The DTI atlas consists of water diffusion tensors that are first spatially transformed into local tumor resistance tensors, also known as metric tensors, and secondly fed to a CTV expansion algorithm to generate anisotropic CTVs. Two models of spatial transformation are considered in the first step. The first model assumes that tumor cells experience reduced resistance parallel to the white matter fibers. The second model assumes that the anisotropy of tumor cell resistance is proportional to the anisotropy observed in DTI, with an 'anisotropy weighting parameter' controlling the proportionality. The models are evaluated in a cohort of ten brain tumor patients.Main results.To evaluate the sensitivity of the model, a library of model-generated CTVs was computed by varying the resistance and anisotropy parameters. Our results indicate that the resistance coefficient had the most significant effect on the global shape of the CTV expansion by redistributing the target volume from potentially less involved gray matter to white matter tissue. In addition, the anisotropy weighting parameter proved useful in locally increasing CTV expansion in regions characterized by strong tissue directionality, such as near the corpus callosum.Significance.By incorporating anisotropy into the CTV expansion, this study is a step toward an interactive CTV definition that can assist physicians in incorporating neuroanatomy into a clinically optimized CTV.

A spot-specific range uncertainty framework for robust optimization of proton therapy treatments.

PURPOSE: An accurate estimation of range uncertainties is essential to exploit the potential of proton therapy. According to Paganetti's study, a value of 2.4% (1.5 standard deviation) is currently recommended for planning robust treatments with Monte Carlo dose engines. This number is based on a dominant contribution from the mean excitation energy of tissues. However, it was recently shown that expressing tissues as a mixture of water and "dry" material in the CT calibration process allowed for a significant reduction of this uncertainty. We thus propose an adapted framework for pencil beam scanning robust optimization. First, we move towards a spot-specific range uncertainty (SSRU) determination. Second, we use the water-based formalism to reduce range uncertainties and, potentially, to spare better the organs at risk. METHODS: The stoichiometric calibration was adapted to provide a molecular decomposition (including water) of each voxel of the CT. The SSRU calculation was implemented in MCsquare, a fast Monte Carlo dose engine dedicated to proton therapy. For each spot, a ray-tracing method was used to propagate molecular I-values uncertainties and obtain the corresponding effective range uncertainty. These were then combined with other sources of range uncertainties, according to Paganetti's study of 2012. The method was then assessed on three head-and-neck patients. Two plans were optimized for each patient: the first one with the classical 2.4% flat range uncertainty (FRU), the second one with the variable range uncertainty. Both plans were then compared in terms of target coverage and OAR mean dose reduction. Robustness evaluations were also performed, using the SSRU for both plans in order to simulate errors as realistically as possible. RESULTS: For patient 1, it was found that the median SSRU was 1.04% (1.5 standard deviation), yielding, therefore, a very large reduction from the 2.4% FRU. All three SSRU plans were found to have a very good robustness level at a 90% confidence interval while sparing OAR better than the classical plan. For instance, in nominal cases, average reductions in the mean dose of 15.7, 8.4, and 13.2% were observed in the left parotid, right parotid, and pharyngeal constrictor muscle, respectively. As expected, the classical plans showed a higher but unnecessary level of robustness. CONCLUSIONS: Promising results of the SSRU framework were observed on three head-and-neck cases, and more patients should now be considered. The method could also benefit to other tumor sites and, in the long run, the variable part of the range uncertainty could be generalized to other sources of uncertainty in order to move towards more and more patient-specific treatments.

Exploring trade-offs in treatment planning for brain tumor cases with a probabilistic definition of the clinical target volume.

PURPOSE: This study demonstrates how a novel probabilistic clinical target volume (CTV) concept-the clinical target distribution (CTD)-can be used to navigate the trade-off between target coverage and organ sparing with a semi-interactive treatment planning approach. METHODS: Two probabilistic treatment planning methods are presented that use tumor probabilities to balance tumor control with organ-at-risk (OAR) sparing. The first method explores OAR dose reduction by systematically discarding x % $x\%$ of CTD voxels with an unfavorable dose-to-probability ratio from the minimum dose coverage objective. The second method sequentially expands the target volume from the GTV edge, calculating the CTD coverage versus OAR sparing trade-off after dosing each expansion. Each planning method leads to estimated levels of tumor control under specific statistical models of tumor infiltration: an independent tumor islets model and contiguous circumferential tumor growth model. The methods are illustrated by creating proton therapy treatment plans for two glioblastoma patients with the clinical goal of sparing the hippocampus and brainstem. For probabilistic plan evaluation, the concept of a dose-expected-volume histogram is introduced, which plots the dose to the expected tumor volume ⟨ v ⟩ $\langle v \rangle$ considering tumor probabilities. RESULTS: Both probabilistic planning approaches generate a library of treatment plans to interactively navigate the planning trade-offs. In the first probabilistic approach, a significant reduction of hippocampus dose could be achieved by excluding merely 1% of CTD voxels without compromising expected tumor control probability (TCP) or CTD coverage: the hippocampus D 2 % $D_{2\%}$ dose reduces with 9.5 and 5.3 Gy for Patient 1 and 2, while the TCP loss remains below 1%. Moreover, discarding up to 10% of the CTD voxels does not significantly diminish the expected CTD dose, even though evaluation with a binary volume suggests poor CTD coverage. In the second probabilistic approach, the expected CTD D ⟨ 98 % ⟩ $D_{\langle 98\%\rangle }$ and TCP depend more strongly on the extent of the high-dose region: the target volume margin cannot be reduced by more than 2 mm if one aims at keeping the expected CTD D ⟨ 98 % ⟩ $D_{\langle 98\%\rangle }$ loss and TCP loss under 1 Gy and 2%, respectively. Therefore, there is less potential for improved OAR sparing without compromising TCP or expected CTD coverage. CONCLUSIONS: This study proposes and implements treatment planning strategies to explore trade-offs using tumor probabilities.

Integrating Structure Propagation Uncertainties in the Optimization of Online Adaptive Proton Therapy Plans.

Currently, adaptive strategies require time- and resource-intensive manual structure corrections. This study compares different strategies: optimization without manual structure correction, adaptation with physician-drawn structures, and no adaptation. Strategies were compared for 16 patients with pancreas, liver, and head and neck (HN) cancer with 1-5 repeated images during treatment: 'reference adaptation', with structures drawn by a physician; 'single-DIR adaptation', using a single set of deformably propagated structures; 'multi-DIR adaptation', using robust planning with multiple deformed structure sets; 'conservative adaptation', using the intersection and union of all deformed structures; 'probabilistic adaptation', using the probability of a voxel belonging to the structure in the optimization weight; and 'no adaptation'. Plans were evaluated using reference structures and compared using a scoring system. The reference adaptation with physician-drawn structures performed best, and no adaptation performed the worst. For pancreas and liver patients, adaptation with a single DIR improved the plan quality over no adaptation. For HN patients, integrating structure uncertainties brought an additional benefit. If resources for manual structure corrections would prevent online adaptation, manual correction could be replaced by a fast 'plausibility check', and plans could be adapted with correction-free adaptation strategies. Including structure uncertainties in the optimization has the potential to make online adaptation more automatable.

Modeling the propagation of tumor fronts with shortest path and diffusion models-implications for the definition of the clinical target volume.

Objective.The overarching objective is to make the definition of the clinical target volume (CTV) in radiation oncology less subjective and more scientifically based. The specific objective of this study is to investigate similarities and differences between two methods that model tumor spread beyond the visible gross tumor volume (GTV): (1) the shortest path model, which is the standard method of adding a geometric GTV-CTV margin, and (2) the reaction-diffusion model.Approach.These two models to capture the invisible tumor 'fire front' are defined and compared in mathematical terms. The models are applied to example cases that represent tumor spread in non-uniform and anisotropic media with anatomical barriers.Main results.The two seemingly disparate models bring forth traveling waves that can be associated with the front of tumor growth outward from the GTV. The shape of the fronts is similar for both models. Differences are seen in cases where the diffusive flow is reduced due to anatomical barriers, and in complex spatially non-uniform cases. The diffusion model generally leads to smoother fronts. The smoothness can be controlled with a parameter defined by the ratio of the diffusion coefficient and the proliferation rate.Significance.Defining the CTV has been described as the weakest link of the radiotherapy chain. There are many similarities in the mathematical description and the behavior of the common geometric GTV-CTV expansion method, and the definition of the CTV tumor front via the reaction-diffusion model. Its mechanistic basis and the controllable smoothness make the diffusion model an attractive alternative to the standard GTV-CTV margin model.

Improved healthy tissue sparing in proton therapy of lung tumors using statistically sound robust optimization and evaluation.

INTRODUCTION: Robust planning is essential in proton therapy for ensuring adequate treatment delivery in the presence of uncertainties. For both robust optimization and evaluation, commonly-used techniques can be overly conservative in selecting error scenarios and lack in providing quantified confidence levels. In this study, established techniques are compared to comprehensive alternatives to assess the differences in target coverage and organ at risk (OAR) dose. METHOD: Thirteen lung cancer patients were planned. Two robust optimization methods were used: scenario selection from marginal probabilities (SSMP) based on using maximum setup and range error values and scenario selection from joint probabilities (SSJP) that selects errors on a predefined 90% hypersurface. Two robust evaluation methods were used: conventional evaluation (CE) based on generating error scenarios from combinations of maximum errors of each uncertainty source and statistical evaluation (SE) via the Monte Carlo dose engine MCsquare which considers scenario probabilities. RESULTS: Plans optimized using SSJP had, on average, 0.5 Gy lower dose in CTV D98(worst-case) than SSMP-optimized plans. When evaluated using SE, 92.3% of patients passed our clinical threshold in both optimization methods. Average gains in OAR sparing were recorded when transitioning from SSMP to SSJP: esophagus (0.6 Gy D2(nominal), 0.9 Gy D2(worst-case)), spinal cord (3.9 Gy D2(nominal), 4.1 Gy D2(worst-case)) heart (1.1 Gy Dmean, 1.9% V30), lungs-GTV (1.0 Gy Dmean , 1.9% V30). CONCLUSION: Optimization using SSJP yielded significant OAR sparing in all recorded metrics with a target robustness within our clinical objectives, provided that a more statistically sound robustness evaluation method was used.

Introducing a probabilistic definition of the target in a robust treatment planning framework.

The 'clinical target distribution' (CTD) has recently been introduced as a promising alternative to the binary clinical target volume (CTV). However, a comprehensive study that considers the CTD, together with geometric treatment uncertainties, was lacking. Because the CTD is inherently a probabilistic concept, this study proposes a fully probabilistic approach that integrates the CTD directly in a robust treatment planning framework. First, the CTD is derived from a reported microscopic tumor infiltration model such that it explicitly features the probability of tumor cell presence in its target definition. Second, two probabilistic robust optimization methods are proposed that evaluate CTD coverage under uncertainty. The first method minimizes the expected-value (EV) over the uncertainty scenarios and the second method minimizes the sum of the expected value and standard deviation (EV-SD), thereby penalizing the spread of the objectives from the mean. Both EV and EV-SD methods introduce the CTD in the objective function by using weighting factors that represent the probability of tumor presence. The probabilistic methods are compared to a conventional worst-case approach that uses the CTV in a worst-case optimization algorithm. To evaluate the treatment plans, a scenario-based evaluation strategy is implemented that combines the effects of microscopic tumor infiltrations with the other geometric uncertainties. The methods are tested for five lung tumor patients, treated with intensity-modulated proton therapy. The results indicate that for the studied patient cases, the probabilistic methods favor the reduction of the esophagus dose but compensate by increasing the high-dose region in a low conflicting organ such as the lung. These results show that a fully probabilistic approach has the potential to obtain clinical benefits when tumor infiltration uncertainties are taken into account directly in the treatment planning process.

Accelerated robust optimization algorithm for proton therapy treatment planning.

PURPOSE: Robust optimization is a computational expensive process resulting in long plan computation times. This issue is especially critical for moving targets as these cases need a large number of uncertainty scenarios to robustly optimize their treatment plans. In this study, we propose a novel worst-case robust optimization algorithm, called dynamic minimax, that accelerates the conventional minimax optimization. Dynamic minimax optimization aims at speeding up the plan optimization process by decreasing the number of evaluated scenarios in the optimization. METHODS: For a given pool of scenarios (e.g., 63 = 7 setup  × 3 range  × 3 breathing phases), the proposed dynamic minimax algorithm only considers a reduced number of candidate-worst scenarios, selected from the full 63 scenario set. These scenarios are updated throughout the optimization by randomly sampling new scenarios according to a hidden variable P, called the "probability acceptance function," which associates with each scenario the probability of it being selected as the worst case. By doing so, the algorithm favors scenarios that are mostly "active," that is, frequently evaluated as the worst case. Additionally, unconsidered scenarios have the possibility to be re-considered, later on in the optimization, depending on the convergence towards a particular solution. The proposed algorithm was implemented in the open-source robust optimizer MIROpt and tested for six four-dimensional (4D) IMPT lung tumor patients with various tumor sizes and motions. Treatment plans were evaluated by performing comprehensive robustness tests (simulating range errors, systematic setup errors, and breathing motion) using the open-source Monte Carlo dose engine MCsquare. RESULTS: The dynamic minimax algorithm achieved an optimization time gain of 84%, on average. The dynamic minimax optimization results in a significantly noisier optimization process due to the fact that more scenarios are accessed in the optimization. However, the increased noise level does not harm the final quality of the plan. In fact, the plan quality is similar between dynamic and conventional minimax optimization with regard to target coverage and normal tissue sparing: on average, the difference in worst-case D95 is 0.2 Gy and the difference in mean lung dose and mean heart dose is 0.4 and 0.1 Gy, respectively (evaluated in the nominal scenario). CONCLUSIONS: The proposed worst-case 4D robust optimization algorithm achieves a significant optimization time gain of 84%, without compromising target coverage or normal tissue sparing.

Towards fast and robust 4D optimization for moving tumors with scanned proton therapy.

PURPOSE: Robust optimization is becoming the gold standard for generating robust plans against various kinds of treatment uncertainties. Today, most robust optimization strategies use a pragmatic set of treatment scenarios (the so-called uncertainty set) consisting of combinations of maximum errors, of each considered uncertainty source (such as tumor motion, setup and image-conversion errors). This approach presents two key issues. First, a subset of considered scenarios is unnecessarily improbable which could potentially compromise the plan quality. Second, the resulting large uncertainty set leads to long plan computation times, which limits the potential for robust optimization as a standard clinical tool. In order to address these issues, a method is introduced which is able to preselect a limited set of relevant treatment error scenarios. METHODS: Uncertainties due to systematic setup errors, image-conversion errors and respiratory tumor motion are considered. A four-dimensional (4D)-equiprobability hypersurface is defined, which takes into account the joint probabilities of the above-mentioned uncertainty sources. Only scenarios that lie on the predefined 4D hypersurface are considered, guaranteeing statistical consistency of the uncertainty set. In this regard, twelve scenarios are selected that cover maximum spatial displacements of the tumor during breathing. Subsequently, additional scenarios are considered (sampled from the aforementioned 4D hypersurface) in order to cover any estimated residual range errors. Two different scenario-selection procedures were tested: (a) the maximum displacements (MD) method that only considers twelve scaled maximum displacement scenarios and (b) maximum displacements and residual range (MDR) method which, in addition to the scaled maximum displacement scenarios, considers additional maximum range uncertainty scenarios. The methods were tested for five lung cancer patients by performing comprehensive Monte Carlo robustness evaluations. RESULTS: A plan computation time gain of 78% is achieved by applying the MD method, whilst obtaining a target robustness of D 95 larger than 95% of the prescribed dose, for the worst-case scenario. Additionally, the MD method has the potential to be fully automatic which makes it a promising candidate for fast automatic planning workflows. The MDR method produced plans with excellent target robustness (D 99 larger than 95% of the prescribed dose, even for the worst-case scenario), whilst still obtaining a significant plan computation time gain of 57%. CONCLUSIONS: Two scenario-selection procedures were developed which achieved significant reduction of plan computation time and memory consumption, without compromising plan quality or robustness.