Robustness Recipes for Minimax Robust Optimization in Intensity Modulated Proton Therapy for Oropharyngeal Cancer Patients

Impact of robust treatment planning on single- and multi-field optimized plans for proton beam therapy of unilateral head and neck target volumes

BACKGROUND

Proton beam therapy is promising for the treatment of head and neck cancer (HNC), but it is sensitive to uncertainties in patient positioning and particle range. Studies have shown that the planning target volume (PTV) concept may not be sufficient to ensure robustness of the target coverage. A few planning studies have considered irradiation of unilateral HNC targets with protons, but they have only taken into account the dose on the nominal plan, without considering anatomy changes occurring during the treatment course.

METHODS

Four pencil beam scanning (PBS) proton therapy plans were calculated for 8 HNC patients with unilateral target volumes: single-field (SFO) and multi-field optimized (MFO) plans, either using the PTV concept or clinical target volume (CTV)-based robust optimization. The dose was recalculated on computed tomography (CT) scans acquired during the treatment course. Doses to target volumes and organs at risk (OARs) were compared for the nominal plans, cumulative doses considering anatomical changes, and additional setup and range errors in each fraction. If required, the treatment plan was adapted, and the dose was compared with the non-adapted plan.

RESULTS

All nominal plans fulfilled the clinical specifications for target coverage, but significantly higher doses on the ipsilateral parotid gland were found for both SFO approaches. MFO PTV-based plans had the lowest robustness against range and setup errors. During the treatment course, the influence of the anatomical variation on the dose has shown to be patient specific, mostly independent of the chosen planning approach. Nine plans in four patients required adaptation, which led to a significant improvement of the target coverage and a slight reduction in the OAR dose in comparison to the cumulative dose without adaptation.

CONCLUSIONS

The use of robust MFO optimization is recommended for ensuring plan robustness and reduced doses in the ipsilateral parotid gland. Anatomical changes occurring during the treatment course might degrade the target coverage and increase the dose in the OARs, independent of the chosen planning approach. For some patients, a plan adaptation may be required.

Impact of model and dose uncertainty on model-based selection of oropharyngeal cancer patients for proton therapy

BACKGROUND

Proton therapy is becoming increasingly available, so it is important to apply objective and individualized patient selection to identify those who are expected to benefit most from proton therapy compared to conventional intensity modulated radiation therapy (IMRT). Comparative treatment planning using normal tissue complication probability (NTCP) evaluation has recently been proposed. This work investigates the impact of NTCP model and dose uncertainties on model-based patient selection.

MATERIAL AND METHODS

We used IMRT and intensity modulated proton therapy (IMPT) treatment plans of 78 oropharyngeal cancer patients, which were generated based on automated treatment planning and evaluated based on three published NTCP models. A reduction in NTCP of more than a certain threshold (e.g. 10% lower NTCP) leads to patient selection for IMPT, referred to as 'nominal' selection. To simulate the effect of uncertainties in NTCP-model coefficients (based on reported confidence intervals) and planned doses on the accuracy of model-based patient selection, the Monte Carlo method was used to sample NTCP-model coefficients and doses from a probability distribution centered at their nominal values. Patient selection accuracy within a certain sample was defined as the fraction of patients which had similar selection in both the 'nominal' and 'sampled' scenario.

RESULTS

For all three NTCP models, the median patient selection accuracy was found to be above 70% when only NTCP-model uncertainty was considered. Selection accuracy decreased with increasing uncertainty resulting from differences between planned and delivered dose. In case of excessive dose uncertainty, selection accuracy decreased to 60%.

CONCLUSION

Model and dose uncertainty highly influence the accuracy of model-based patient selection for proton therapy. A reduction of NTCP-model uncertainty is necessary to reach more accurate model-based patient selection.

The impact of treatment accuracy on proton therapy patient selection for oropharyngeal cancer patients

BACKGROUND AND PURPOSE

The impact of treatment accuracy on NTCP-based patient selection for proton therapy is currently unknown. This study investigates this impact for oropharyngeal cancer patients.

MATERIALS AND METHODS

Data of 78 patients was used to automatically generate treatment plans for a simultaneously integrated boost prescribing 70 Gy_RBE/54.25 Gy_RBE in 35 fractions. IMRT treatment plans were generated with three different margins; intensity modulated proton therapy (IMPT) plans for five different setup and range robustness settings. Four NTCP models were evaluated. Patients were selected for proton therapy if NTCP reduction was ≥10% or ≥5% for grade II or III complications, respectively.

RESULTS

The degree of robustness had little impact on patient selection for tube feeding dependence, while the margin had. For other complications the impact of the robustness setting was noticeably higher. For high-precision IMRT (3 mm margin) and high-precision IMPT (3 mm setup/3% range error), most patients were selected for proton therapy based on problems swallowing solid food (51.3%) followed by tube feeding dependence (37.2%), decreased parotid flow (29.5%), and patient-rated xerostomia (7.7%).

CONCLUSIONS

Treatment accuracy has a significant impact on the number of patients selected for proton therapy. Therefore, it cannot be ignored in estimating the number of patients for proton therapy.

Dual-energy CT based proton range prediction in head and pelvic tumor patients

BACKGROUND AND PURPOSE

To reduce range uncertainty in particle therapy, an accurate computation of stopping-power ratios (SPRs) based on computed tomography (CT) is crucial. Here, we assess range differences between the state-of-the-art CT-number-to-SPR conversion using a generic Hounsfield look-up table (HLUT) and a direct patient-specific SPR prediction (RhoSigma) based on dual-energy CT (DECT) in 100 proton treatment fields.

MATERIAL AND METHODS

For 25 head-tumor and 25 prostate-cancer patients, the clinically applied treatment plan, optimized using a HLUT, was recalculated with RhoSigma as CT-number-to-SPR conversion. Depth-dose curves in beam direction were extracted for both dose distributions in a regular grid and range deviations were determined and correlated to SPR differences within the irradiated volume.

RESULTS

Absolute (relative) mean water-equivalent range shifts of 1.1mm (1.2%) and 4.1mm (1.7%) were observed in the head-tumor and prostate-cancer cohort, respectively. Due to the case dependency of a generic HLUT, range deviations within treatment fields strongly depend on the tissues traversed leading to a larger variation within one patient than between patients.

CONCLUSIONS

The magnitude of patient-specific range deviations between HLUT and the more accurate DECT-based SPR prediction is clinically relevant. A clinical application of the latter seems feasible as demonstrated in this study using medically approved systems from CT acquisition to treatment planning.

Potential proton and photon dose degradation in advanced head and neck cancer patients by intratherapy changes

Purpose

Evaluation of dose degradation by anatomic changes for head‐and‐neck cancer (HNC) intensity‐modulated proton therapy (IMPT) relative to intensity‐modulated photon therapy (IMRT) and identification of potential indicators for IMPT treatment plan adaptation.

Methods

For 31 advanced HNC datasets, IMPT and IMRT plans were recalculated on a computed tomography scan (CT) taken after about 4 weeks of therapy. Dose parameter changes were determined for the organs at risk (OARs) spinal cord, brain stem, parotid glands, brachial plexus, and mandible, for the clinical target volume (CTV) and the healthy tissue outside planning target volume (PTV). Correlation of dose degradation with target volume changes and quality of rigid CT matching was investigated.

Results

Recalculated IMPT dose distributions showed stronger degradation than the IMRT doses. OAR analysis revealed significant changes in parotid median dose (IMPT) and near maximum dose (D_1ml) of spinal cord (IMPT, IMRT) and mandible (IMPT). OAR dose parameters remained lower in IMPT cases. CTV coverage (V_95%) and overdose (V_107%) deteriorated for IMPT plans to (93.4 ± 5.4)% and (10.6 ± 12.5)%, while those for IMRT plans remained acceptable. Recalculated plans showed similarly decreased PTV conformity, but considerable hotspots, also outside the PTV, emerged in IMPT cases. Lower CT matching quality was significantly correlated with loss of PTV conformity (IMPT, IMRT), CTV homogeneity and coverage (IMPT). Target shrinkage correlated with increased dose in brachial plexus (IMRT, IMPT), hotspot generation outside the PTV (IMPT) and lower PTV conformity (IMRT).

Conclusions

The study underlines the necessity of precise positioning and monitoring of anatomy changes, especially in IMPT which might require adaptation more often. Since OAR doses remained typically below constraints, IMPT plan adaptation will be indicated by target dose degradations.

Coverage-based constraints for IMRT optimization

Radiation therapy treatment planning requires an incorporation of uncertainties in order to guarantee an adequate irradiation of the tumor volumes. In current clinical practice, uncertainties are accounted for implicitly with an expansion of the target volume according to generic margin recipes. Alternatively, it is possible to account for uncertainties by explicit minimization of objectives that describe worst-case treatment scenarios, the expectation value of the treatment or the coverage probability of the target volumes during treatment planning. In this note we show that approaches relying on objectives to induce a specific coverage of the clinical target volumes are inevitably sensitive to variation of the relative weighting of the objectives. To address this issue, we introduce coverage-based constraints for intensity-modulated radiation therapy (IMRT) treatment planning. Our implementation follows the concept of coverage-optimized planning that considers explicit error scenarios to calculate and optimize patient-specific probabilities [Formula: see text] of covering a specific target volume fraction [Formula: see text] with a certain dose [Formula: see text]. Using a constraint-based reformulation of coverage-based objectives we eliminate the trade-off between coverage and competing objectives during treatment planning. In-depth convergence tests including 324 treatment plan optimizations demonstrate the reliability of coverage-based constraints for varying levels of probability, dose and volume. General clinical applicability of coverage-based constraints is demonstrated for two cases. A sensitivity analysis regarding penalty variations within this planing study based on IMRT treatment planning using (1) coverage-based constraints, (2) coverage-based objectives, (3) probabilistic optimization, (4) robust optimization and (5) conventional margins illustrates the potential benefit of coverage-based constraints that do not require tedious adjustment of target volume objectives.

Superiority in Robustness of Multifield Optimization Over Single-Field Optimization for Pencil-Beam Proton Therapy for Oropharynx Carcinoma: An Enhanced Robustness Analysis

PURPOSE

To compare the difference in robustness of single-field optimized (SFO) and robust multifield optimized (rMFO) proton plans for oropharynx carcinoma patients by an improved robustness analysis.

METHODS AND MATERIALS

We generated rMFO proton plans for 11 patients with oropharynx carcinoma treated with SFO intensity modulated proton therapy with simultaneous integrated boost prescription. Doses from both planning approaches were compared for the initial plans and the worst cases from 20 optimization scenarios of setup errors and range uncertainties. Expected average dose distributions per range uncertainty were obtained by weighting the contributions from the respective scenarios with their expected setup error probability, and the spread of dose parameters for different range uncertainties were quantified. Using boundary dose distributions created from 56 combined setup error and range uncertainty scenarios and considering the vanishing influence of setup errors after 30 fractions, we approximated realistic worst-case values for the total treatment course. Error bar metrics derived from these boundary doses are reported for the clinical target volumes (CTVs) and organs at risk (OARs).

RESULTS

The rMFO plans showed improved CTV coverage and homogeneity while simultaneously reducing the average mean dose to the constrictor muscles, larynx, and ipsilateral middle ear by 5.6 Gy, 2.0 Gy, and 3.9 Gy, respectively. We observed slightly larger differences during robustness evaluation, as well as a significantly higher average brainstem maximum and ipsilateral parotid mean dose for SFO plans. For rMFO plans, the range uncertainty-related spread in OAR dose parameters and many error bar metrics were found to be superior. The SFO plans showed a lower global maximum dose for single-scenario worst cases and a slightly lower mean oral cavity dose throughout.

CONCLUSIONS

An enhanced robustness analysis has been proposed and implemented into clinical systems. The benefit of better CTV coverage and OAR dose sparing in oropharynx carcinoma patients by rMFO compared with SFO proton plans is preserved in a robustness analysis with consideration of setup error and range uncertainty.

Beam angle evaluation to improve inter-fraction motion robustness in pelvic lymph node irradiation with proton therapy

BACKGROUND

Proton therapy dose distributions are sensitive to range variations, e.g. arising from inter-fraction organ motion. The aim of this study was to evaluate the inter-fraction motion robustness of proton beams from different beam angles in irradiation of pelvic lymph nodes (LNs).

MATERIAL AND METHODS

Planning CT (pCT) and multiple repeat CT (rCT) scans of 18 prostate cancer patients were used. Considering left and right LNs separately, the average water equivalent path length (WEPL) over all ray paths in the beams eye view of the LNs were calculated for all gantry/couch angle combinations across all rCTs versus the corresponding pCT. Single beam proton plans were optimized on the pCT for all gantry angles (0° couch) and were re-calculated on all rCTs for each respective patient. WEPL and dose parameters were extracted and a statistical clustering analysis performed to identify patient sub-populations in terms of patterns in which angles were robust.

RESULTS

The WEPL analysis showed a general pattern of least variation for 0° couch beam angles where three minima were found across gantry angles for the left LNs and two for the right LNs. The clustering analysis identified three patient sub-groups for the left LNs and two groups for the right LNs. The dose calculations showed similar results as the WEPL variation, e.g. for the left LNs angles around 25°-35°, 100°-110°, and 160°-170° were consistently preferable for both target and organs at risk.

CONCLUSIONS

Sub-populations of patients with similar patterns of WEPL variations across beam angles were identified. The most robust angles found for WEPL variations were also confirmed by the dose/volume analysis.

Worst case optimization for interfractional motion mitigation in carbon ion therapy of pancreatic cancer

INTRODUCTION

The efficacy of radiation therapy treatments for pancreatic cancer is compromised by abdominal motion which limits the spatial accuracy for dose delivery - especially for particles. In this work we investigate the potential of worst case optimization for interfractional offline motion mitigation in carbon ion treatments of pancreatic cancer.

METHODS

We implement a worst case optimization algorithm that explicitly models the relative biological effectiveness of carbon ions during inverse planning. We perform a comparative treatment planning study for seven pancreatic cancer patients. Treatment plans that have been generated using worst case optimization are compared against (1) conventional intensity-modulated carbon ion therapy, (2) single field uniform dose carbon ion therapy, and (3) an ideal yet impractical scenario relying on daily re-planning. The dosimetric quality and robustness of the resulting treatment plans is evaluated using reconstructions of the daily delivered dose distributions on fractional control CTs.

RESULTS

Idealized daily re-planning consistently gives the best dosimetric results with regard to both target coverage and organ at risk sparing. The absolute reduction of D 95 within the gross tumor volume during fractional dose reconstruction is most pronounced for conventional intensity-modulated carbon ion therapy. Single field uniform dose optimization exhibits no substantial reduction for six of seven patients and values for D 95 for worst case optimization fall in between. The treated volume (D>95 % prescription dose) outside of the gross tumor volume is reduced by a factor of two by worst case optimization compared to conventional optimization and single field uniform dose optimization. Single field uniform dose optimization comes at an increased radiation exposure of normal tissues, e.g. ≈2 Gy (RBE) in the mean dose in the kidneys compared to conventional and worst case optimization and ≈4 Gy (RBE) in D 1 in the spinal cord compared to worst case optimization.

CONCLUSION

Interfractional motion substantially deteriorates dose distributions for carbon ion treatments of pancreatic cancer patients. Single field uniform dose optimization mitigates the negative influence of motion on target coverage at an increased radiation exposure of normal tissue. Worst case optimization enables an exploration of the trade-off between robust target coverage and organ at risk sparing during inverse treatment planning beyond margin concepts.

[Robust treatment planning in proton therapy]

The concentration of the dose delivered by protons at the end of their path, the Bragg peak, has the potential to improve external radiotherapy treatments. Unfortunately, the main strength of the protons, their finite range, is also their greatest weakness. Any uncertainty on the range may lead to inadequate target coverage or excessive toxicity. The uncertainties have multiple origins and include, among others, ballistic errors, morphological modifications or inaccurate estimations of the physical quantities necessary to predict the proton range. Uncertainties have been part of daily practice in conventional radiotherapy with X-rays for a long time. However, dose distributions delivered with X-rays are much less sensitive to uncertainties than the ones delivered with protons. This relative insensitivity enabled the management of uncertainties through safety margins using a simple formalism. The conditions of validity of this formalism are much more restrictive for proton therapy, leading to the need of developing new tools and adapted strategies to manage accurately these uncertainties. The objective of this paper is to present a vision for the management of uncertainties in proton therapy in the continuity of formalisms established for X-rays. The latter are first summarized before discussing the necessary developments in order to consistently apply them to protons.

Fast and accurate sensitivity analysis of IMPT treatment plans using Polynomial Chaos Expansion

The highly conformal planned dose distribution achievable in intensity modulated proton therapy (IMPT) can severely be compromised by uncertainties in patient setup and proton range. While several robust optimization approaches have been presented to address this issue, appropriate methods to accurately estimate the robustness of treatment plans are still lacking. To fill this gap we present Polynomial Chaos Expansion (PCE) techniques which are easily applicable and create a meta-model of the dose engine by approximating the dose in every voxel with multidimensional polynomials. This Polynomial Chaos (PC) model can be built in an automated fashion relatively cheaply and subsequently it can be used to perform comprehensive robustness analysis. We adapted PC to provide among others the expected dose, the dose variance, accurate probability distribution of dose-volume histogram (DVH) metrics (e.g. minimum tumor or maximum organ dose), exact bandwidths of DVHs, and to separate the effects of random and systematic errors. We present the outcome of our verification experiments based on 6 head-and-neck (HN) patients, and exemplify the usefulness of PCE by comparing a robust and a non-robust treatment plan for a selected HN case. The results suggest that PCE is highly valuable for both research and clinical applications.

The price of robustness; impact of worst-case optimization on organ-at-risk dose and complication probability in intensity-modulated proton therapy for oropharyngeal cancer patients

PURPOSE

To quantify the impact of the degree of robustness against setup errors and range errors on organ-at-risk (OAR) dose and normal tissue complication probabilities (NTCPs) in intensity-modulated proton therapy for oropharyngeal cancer patients.

MATERIAL AND METHODS

For 20 oropharyngeal cases (10 unilateral and 10 bilateral), robust treatment plans were generated using 'minimax' worst-case optimization. We varied the robustness against setup errors ('setup robustness') from 1 to 7mm and the robustness against range errors ('range robustness') from 1% to 7% (+1mm). We evaluated OAR doses and NTCP-values for xerostomia, dysphagia and larynx edema.

RESULTS

Varying the degree of setup robustness was found to have a considerably larger impact than varying the range robustness. Increasing setup robustness from 1mm to 3, 5, and 7mm resulted in average NTCP-values to increase by 1.9, 4.4 and 7.5 percentage point, whereas they increased by only 0.4, 0.8 and 1.2 percentage point when increasing range robustness from 1% to 3%, 5% and 7%. The degree of setup robustness was observed to have a clinically significant impact in bilateral cases in particular.

CONCLUSIONS

For oropharyngeal cancer patients, minimizing setup errors should be given a higher priority than minimizing range errors.