Analytical probabilistic modeling of RBE-weighted dose for ion therapy

A novel stochastic optimization method for handling misalignments of proton and photon doses in combined treatments

Objective. Combined proton–photon treatments, where most fractions are delivered with photons and only a few are delivered with protons, may represent a practical approach to optimally use limited proton resources. It has been shown that, when organs at risk (OARs) are located within or near the tumor, the optimal multi-modality treatment uses protons to hypofractionate parts of the target volume and photons to achieve near-uniform fractionation in dose-limiting healthy tissues, thus exploiting the fractionation effect. These plans may be sensitive to range and setup errors, especially misalignments between proton and photon doses. Thus, we developed a novel stochastic optimization method to directly incorporate these uncertainties into the biologically effective dose (BED)-based simultaneous optimization of proton and photon plans. Approach. The method considers the expected value E b and standard deviation σ b of the cumulative BED b in every voxel of a structure. For the target, a piecewise quadratic penalty function of the form b min − E b − 2 σ b + 2 is minimized, aiming for plans in which the expected BED minus two times the standard deviation exceeds the prescribed BED b min . Analogously, E b + 2 σ b − b max + 2 is considered for OARs. Main results. Using a spinal metastasis case and a liver cancer patient, it is demonstrated that the novel stochastic optimization method yields robust combined treatment plans. Tumor coverage and a good sparing of the main OARs are maintained despite range and setup errors, and especially misalignments between proton and photon doses. This is achieved without explicitly considering all combinations of proton and photon error scenarios. Significance. Concerns about range and setup errors for safe clinical implementation of optimized proton–photon radiotherapy can be addressed through an appropriate stochastic planning method.

Long short-term memory networks for proton dose calculation in highly heterogeneous tissues

PURPOSE

To investigate the feasibility and accuracy of proton dose calculations with artificial neural networks (ANNs) in challenging three-dimensional (3D) anatomies.

METHODS

A novel proton dose calculation approach was designed based on the application of a long short-term memory (LSTM) network. It processes the 3D geometry as a sequence of two-dimensional (2D) computed tomography slices and outputs a corresponding sequence of 2D slices that forms the 3D dose distribution. The general accuracy of the approach is investigated in comparison to Monte Carlo reference simulations and pencil beam dose calculations. We consider both artificial phantom geometries and clinically realistic lung cases for three different pencil beam energies.

RESULTS

For artificial phantom cases, the trained LSTM model achieved a 98.57% γ-index pass rate ([1%, 3 mm]) in comparison to MC simulations for a pencil beam with initial energy 104.25 MeV. For a lung patient case, we observe pass rates of 98.56%, 97.74%, and 94.51% for an initial energy of 67.85, 104.25, and 134.68 MeV, respectively. Applying the LSTM dose calculation on patient cases that were fully excluded from the training process yields an average γ-index pass rate of 97.85%.

CONCLUSIONS

LSTM networks are well suited for proton dose calculation tasks. Further research, especially regarding model generalization and computational performance in comparison to established dose calculation methods, is warranted.

Variance-based sensitivity analysis for uncertainties in proton therapy: A framework to assess the effect of simultaneous uncertainties in range, positioning, and RBE model predictions on RBE-weighted dose distributions

PURPOSE

Treatment plans in proton therapy are more sensitive to uncertainties than in conventional photon therapy. In addition to setup uncertainties, proton therapy is affected by uncertainties in proton range and relative biological effectiveness (RBE). While to date a constant RBE of 1.1 is commonly assumed, the actual RBE is known to increase toward the distal end of the spread-out Bragg peak. Several models for variable RBE predictions exist. We present a framework to evaluate the combined impact and interactions of setup, range, and RBE uncertainties in a comprehensive, variance-based sensitivity analysis (SA).

MATERIAL AND METHODS

The variance-based SA requires a large number (10⁴ -10⁵ ) of RBE-weighted dose (RWD) calculations. Based on a particle therapy extension of the research treatment planning system CERR we implemented a fast, graphics processing unit (GPU) accelerated pencil beam modeling of patient and range shifts. For RBE predictions, two biological models were included: The mechanistic repair-misrepair-fixation (RMF) model and the phenomenological Wedenberg model. All input parameters (patient position, proton range, RBE model parameters) are sampled simultaneously within their assumed probability distributions. Statistical formalisms rank the input parameters according to their influence on the overall uncertainty of RBE-weighted dose-volume histogram (RW-DVH) quantiles and the RWD in every voxel, resulting in relative, normalized sensitivity indices (S = 0: noninfluential input, S = 1: only influential input). Results are visualized as RW-DVHs with error bars and sensitivity maps.

RESULTS AND CONCLUSIONS

The approach is demonstrated for two representative brain tumor cases and a prostate case. The full SA including ∼ 3 × 10 4 RWD calculations took 39, 11, and 55 min, respectively. Range uncertainty was an important contribution to overall uncertainty at the distal end of the target, while the relatively smaller uncertainty inside the target was governed by biological uncertainties. Consequently, the uncertainty of the RW-DVH quantile D₉₈ for the target was governed by range uncertainty while the uncertainty of the mean target dose was dominated by the biological parameters. The SA framework is a powerful and flexible tool to evaluate uncertainty in RWD distributions and DVH quantiles, taking into account physical and RBE uncertainties and their interactions. The additional information might help to prioritize research efforts to reduce physical and RBE uncertainties and could also have implications for future approaches to biologically robust planning and optimization.

Analytical probabilistic modeling of dose-volume histograms

PURPOSE

Radiotherapy, especially with charged particles, is sensitive to executional and preparational uncertainties that propagate to uncertainty in dose and plan quality indicators, for example, dose-volume histograms (DVHs). Current approaches to quantify and mitigate such uncertainties rely on explicitly computed error scenarios and are thus subject to statistical uncertainty and limitations regarding the underlying uncertainty model. Here we present an alternative, analytical method to approximate moments, in particular expectation value and (co)variance, of the probability distribution of DVH-points, and evaluate its accuracy on patient data.

METHODS

We use Analytical Probabilistic Modeling (APM) to derive moments of the probability distribution over individual DVH-points based on the probability distribution over dose. By using the computed moments to parameterize distinct probability distributions over DVH-points (here normal or beta distributions), not only the moments but also percentiles, that is, α - DVHs, are computed. The model is subsequently evaluated on three patient cases (intracranial, paraspinal, prostate) in 30- and single-fraction scenarios by assuming the dose to follow a multivariate normal distribution, whose moments are computed in closed-form with APM. The results are compared to a benchmark based on discrete random sampling.

RESULTS

The evaluation of the new probabilistic model on the three patient cases against a sampling benchmark proves its correctness under perfect assumptions as well as good agreement in realistic conditions. More precisely, ca. 90% of all computed expected DVH-points and their standard deviations agree within 1% volume with their empirical counterpart from sampling computations, for both fractionated and single fraction treatments. When computing α - DVH, the assumption of a beta distribution achieved better agreement with empirical percentiles than the assumption of a normal distribution: While in both cases probabilities locally showed large deviations (up to ±0.2), the respective - DVHs for α={0.05,0.5,0.95} only showed small deviations in respective volume (up to ±5% volume for a normal distribution, and up to 2% for a beta distribution). A previously published model from literature, which was included for comparison, exhibited substantially larger deviations.

CONCLUSIONS

With APM we could derive a mathematically exact description of moments of probability distributions over DVH-points given a probability distribution over dose. The model generalizes previous attempts and performs well for both choices of probability distributions, that is, normal or beta distributions, over DVH-points.

Modeling RBE-weighted dose variations in irregularly moving abdominal targets treated with carbon ion beams

PURPOSE

To model four-dimensional (4D) relative biological effectiveness (RBE)-weighted dose variations in abdominal lesions treated with scanned carbon ion beam in case of irregular breathing motion.

METHODS

The proposed method, referred to as bioWED method, combines the simulation of tumor motion in a patient- and beam-specific water equivalent depth (WED)-space with RBE modeling, aiming at the estimation of RBE-weighted dose changes due to respiratory motion. The method was validated on a phantom, simulating gated and free breathing dose delivery, and on a patient case, for which free breathing irradiation was assumed and both amplitude and baseline breathing irregularities were simulated through a respiratory motion model. We quantified (a) the effect of motion on the equivalent uniform dose (EUD) and the RBE-weighted dose-volume histograms (DVH), by comparing the planned dose distribution with "ground truth" 4D RBE-weighted doses computed using 4D computed tomography data, and (ii) the estimation error, by comparing the doses estimated with the bioWED method to "ground truth" 4D RBE-weighted doses.

RESULTS

In the phantom validation, the estimation error on the EUD was limited with respect to the motion effect and the median estimation error on relevant RBE-weighted DVH metrics remained within 5%. In the patient study, the estimation error as computed on the EUD was smaller than the corresponding motion effect, exhibiting the largest values in the baseline irregularity simulation. However, the median estimation error over all simulations was below 3.2% considering relevant DVH metrics.

CONCLUSIONS

In the evaluated cases, the bioWED method showed proper accuracy when compared to deformable image registration-based 4D dose calculation. Therefore, it can be seen as a tool to test treatment plan robustness against irregular breathing motion, although its accuracy decreases as a function of increasing soft tissue deformation and should be evaluated on a larger patient dataset.

An evaluation method of clinical impact with setup, range, and radiosensitivity uncertainties in fractionated carbon-ion therapy

In light ion therapy, the dose concentration is highly sensitive to setup and range errors. Here we propose a method for evaluating the effects of these errors by using the correlation between fractions on tumour control probability (TCP) in carbon-ion therapy. This method incorporates the concept of equivalent stochastic dose (Cranmer-Sargison and Zavgorodni 2005 Phys. Med. Biol. 50 4097-109), which was defined as a dose that gives the mean expected survival fraction (SF) for the stochastically variable dose. The mean expected SFs were calculated while considering the correlation between fractions for setup and range errors. By using this SF, equivalent stochastic clinical doses (ESCD), which are weighted by relative biological effectiveness, of lung and prostate cases with varying errors were derived. To account for spatial dose heterogeneity, equivalent uniform stochastic clinical doses (EUSCD) were obtained by using the mean expected SF in the volume of interest. TCP curves were calculated for each assumed error considering inter-patient sensitivity variation with a fractionation effect. ESCD distributions, EUSCD, and TCP curves were affected by the inter-fraction correlation and the contribution of setup and range errors. Irradiated areas that could be affected by these errors can be visualized quantitatively by using the ESCD distribution. TCP curves for the errors of various conditions converged around the TCP curve in nominal conditions by using the EUSCD. EUSCD correlated well with TCP in setup and range errors when the errors were not large and was comparatively stably insensitive to uncertain biological parameters. The proposed evaluation method with EUSCD and TCP calculations will be useful to indicate tumour doses to improve realistic dose distributions in carbon-ion therapy.