Sinogram analysis of aperture optimization by iterative least-squares in volumetric modulated arc therapy

Inverse planning of lung radiotherapy with photon and proton beams using a discrete ordinates Boltzmann solver

Objective. A discrete ordinates Boltzmann solver has recently been developed for use as a fast and accurate dose engine for calculation of photon and proton beams. The purpose of this study is to apply the algorithm to the inverse planning process for photons and protons and to evaluate the impact that this has on the quality of the final solution.Approach.The method was implemented into an iterative least-squares inverse planning optimiser, with the Boltzmann solver used every 20 iterations over the total of 100 iterations. Elemental dose distributions for the intensity modulation and the dose changes at the intermediate iterations were calculated by a convolution algorithm for photons and a simple analytical model for protons. The method was evaluated for 12 patients in the heterogeneous tissue environment encountered in radiotherapy of lung tumours. Photon arc and proton arc treatments were considered in this study. The results were compared with those for use of the Boltzmann solver solely at the end of inverse planning or not at all.Main results.Application of the Boltzmann solver at the end of inverse planning shows the dose heterogeneity in the planning target volume to be greater than calculated by convolution and empirical methods, with the median root-mean-square dose deviation increasing from 3.7 to 5.3 for photons and from 1.9 to 3.4 for proton arcs. Use of discrete ordinates throughout inverse planning enables homogeneity of target coverage to be maintained throughout, the median root-mean-square dose deviation being 3.6 for photons and 2.3 for protons. Dose to critical structures is similar with discrete ordinates and conventional methods. Time for inverse planning with discrete ordinates takes around 1-2 h using a contemporary computing environment.Significance.By incorporating the Boltzmann solver into an iterative least squares inverse planning optimiser, accurate dose calculation in a heterogeneous medium is obtained throughout inverse planning, with the result that the final dose distribution is of the highest quality.

A discrete ordinates Boltzmann solver for application to inverse planning of photons and protons

The aim of this work is to develop a discrete ordinates Boltzmann solver that can be used for calculation of absorbed dose from both photons and protons within an inverse planning optimiser, so as to perform accurate dose calculation throughout the whole of the inverse planning process. With photons, five transport sweeps were performed to obtain scattered photon fluence, and unscattered electron fluence was then obtained and used as a fixed source for solution of the electron transport equations. With protons, continuous slowing down was treated as a fixed source, and five transport sweeps were used to calculate scattered fluence. The total electron or proton fluence was multiplied by the stopping power ratio for the transport medium to obtain absorbed dose. The method was evaluated in homogeneous media and in a lung case where the planning target volume was surrounded by low-density lung material. Photon arc, proton passive scattering and proton arc treatments were considered. The results were compared to a clinically validated convolution dose calculation for photons, and with an analytical method for protons. In water-equivalent media, the discrete ordinates method agrees with the alternative algorithms to within 2%. Convergence is found to be sufficiently complete for water-, lung- and bone-equivalent materials after five iterations. The dose calculated by the relatively simple angular quadrature is seen to be very close to that calculated by a more comprehensive quadrature. For inhomogeneous lung plans, the method shows more heterogeneity of dose to the planning target volume than the comparative methods. The discrete ordinates Boltzmann solver provides a general framework for dose calculation with both photons and protons. The method is suitable for incorporation into an inverse planning optimiser, so that accurate dose calculation in a heterogeneous medium can be obtained throughout inverse planning, with the result that the final dose distribution is as predicted by the optimiser.

A recurrent neural network for rapid detection of delivery errors during real-time portal dosimetry

Background and purpose

Real-time portal dosimetry compares measured images with predicted images to detect delivery errors as the radiotherapy treatment proceeds. This work aimed to investigate the performance of a recurrent neural network for processing image metrics so as to detect delivery errors as early as possible in the treatment.

Materials and methods

Volumetric modulated arc therapy (VMAT) plans of six prostate patients were used to generate sequences of predicted portal images. Errors were introduced into the treatment plans and the modified plans were delivered to a water-equivalent phantom. Four different metrics were used to detect errors. These metrics were applied to a threshold-based method to detect the errors as soon as possible during the delivery, and also to a recurrent neural network consisting of four layers. A leave-two-out approach was used to set thresholds and train the neural network then test the resulting systems.

Results

When using a combination of metrics in conjunction with optimal thresholds, the median segment index at which the errors were detected was 107 out of 180. When using the neural network, the median segment index for error detection was 66 out of 180, with no false positives. The neural network reduced the rate of false negative results from 0.36 to 0.24.

Conclusions

The recurrent neural network allowed the detection of errors around 30% earlier than when using conventional threshold techniques. By appropriate training of the network, false positive alerts could be prevented, thereby avoiding unnecessary disruption to the patient workflow.

Optimisation of a composite difference metric for prompt error detection in real-time portal dosimetry of simulated volumetric modulated arc therapy

OBJECTIVES

In real-time portal dosimetry, thresholds are set for several measures of difference between predicted and measured images, and signals larger than those thresholds signify an error. The aim of this work is to investigate the use of an additional composite difference metric (CDM) for earlier detection of errors.

METHODS

Portal images were predicted for the volumetric modulated arc therapy plans of six prostate patients. Errors in monitor units, aperture opening, aperture position and path length were deliberately introduced into all 180 segments of the treatment plans, and these plans were delivered to a water-equivalent phantom. Four different metrics, consisting of central axis signal, mean image value and two image difference measures, were used to identify errors, and a CDM was added, consisting of a weighted power sum of the individual metrics. To optimise the weights of the CDM and to evaluate the resulting timeliness of error detection, a leave-pair-out strategy was used. For each combination of four patients, the weights of the CDM were determined by an exhaustive search, and the result was evaluated on the remaining two patients.

RESULTS

The median segment index at which the errors were identified was 87 (range 40-130) when using all of the individual metrics separately. Using a CDM as well as multiple separate metrics reduced this to 73 (35-95). The median weighting factors of the four metrics constituting the composite were (0.15, 0.10, 0.15, 0.00). Due to selection of suitable threshold levels, there was only one false positive result in the six patients.

CONCLUSION

This study shows that, in conjunction with appropriate error thresholds, use of a CDM is able to identify increased image differences around 20% earlier than the separate measures.

ADVANCES IN KNOWLEDGE

This study shows the value of combining difference metrics to allow earlier detection of errors during real-time portal dosimetry for volumetric modulated arc therapy treatment.

Dosimetric accuracy of dynamic couch rotation during volumetric modulated arc therapy (DCR-VMAT) for primary brain tumours

Radiotherapy treatment plans using dynamic couch rotation during volumetric modulated arc therapy (DCR-VMAT) reduce the dose to organs at risk (OARs) compared to coplanar VMAT, while maintaining the dose to the planning target volume (PTV). This paper seeks to validate this finding with measurements. DCR-VMAT treatment plans were produced for five patients with primary brain tumours and delivered using a commercial linear accelerator (linac). Dosimetric accuracy was assessed using point dose and radiochromic film measurements. Linac-recorded mechanical errors were assessed by extracting deviations from log files for multi-leaf collimator (MLC), couch, and gantry positions every 20 ms. Dose distributions, reconstructed from every fifth log file sample, were calculated and used to determine deviations from the treatment plans. Median (range) treatment delivery times were 125 s (123-133 s) for DCR-VMAT, compared to 78 s (64-130 s) for coplanar VMAT. Absolute point doses were 0.8% (0.6%-1.7%) higher than prediction. For coronal and sagittal films, respectively, 99.2% (96.7%-100%) and 98.1% (92.9%-99.0%) of pixels above a 20% low dose threshold reported gamma  <1 for 3% and 3 mm criteria. Log file analysis showed similar gantry rotation root-mean-square error (RMSE) for VMAT and DCR-VMAT. Couch rotation RMSE for DCR-VMAT was 0.091° (0.086-0.102°). For delivered dose reconstructions, 100% of pixels above a 5% low dose threshold reported gamma  <1 for 2% and 2 mm criteria in all cases. DCR-VMAT, for the primary brain tumour cases studied, can be delivered accurately using a commercial linac.

Dose prescription with spatial uncertainty for peripheral lung SBRT

Current clinical practice is to prescribe to 95% of the planning target volume (PTV) in 4D stereotactic body radiotherapy (SBRT) for lung. Frequently the PTV margin has a very low physical density so that the internal target volume (ITV) receives an unnecessarily high dose. This study investigates the alternative of prescribing to the ITV while including the effects of positional uncertainties. Five patients were retrospectively studied with volumetric modulated arc therapy treatment plans. Five plans were produced for each patient: a static plan prescribed to PTV D95% , three probabilistic plans prescribed to ITV D95% and a static plan re-prescribed to ITV D95% after inverse planning. For the three probabilistic plans, the scatter kernel in the dose calculation was convolved with a spatial uncertainty distribution consisting of either a uniform distribution extending ±5 mm in the three orthogonal directions, a distribution consisting of delta functions at ±5 mm, or a Gaussian distribution with standard deviation 5 mm. Median ITV D50% is 23% higher than the prescribed dose for static planning and only 10% higher than the prescribed dose for prescription to the ITV. The choice of uncertainty distribution has less than 2% effect on the median ITV dose. Re-prescribing a static plan and evaluating with a probabilistic dose calculation results in a median ITV D95% which is 1.5% higher than when planning probabilistically. This study shows that a robust probabilistic approach to planning SBRT lung treatments results in the ITV receiving a dose closer to the intended prescription. The exact form of the uncertainty distribution is not found to be critical.

Non-coplanar trajectories to improve organ at risk sparing in volumetric modulated arc therapy for primary brain tumors

BACKGROUND AND PURPOSE

To evaluate non-coplanar volumetric modulated arc radiotherapy (VMAT) trajectories for organ at risk (OAR) sparing in primary brain tumor radiotherapy.

MATERIALS AND METHODS

Fifteen patients were planned using coplanar VMAT and compared against non-coplanar VMAT plans for three trajectory optimization techniques. A geometric heuristic technique (GH) combined beam scoring and Dijkstra's algorithm to minimize the importance-weighted sum of OAR volumes irradiated. Fluence optimization was used to perform a local search around coplanar and GH trajectories, producing fluence-based local search (FBLS) and FBLS+GH trajectories respectively.

RESULTS

GH, FBLS, and FBLS+GH trajectories reduced doses to the contralateral globe, optic nerve, hippocampus, temporal lobe, and cochlea. However, FBLS increased dose to the ipsilateral lens, optic nerve and globe. Compared to GH, FBLS+GH increased dose to the ipsilateral temporal lobe and hippocampus, contralateral optics, and the brainstem and body. GH and FBLS+GH trajectories reduced bilateral hippocampi normal tissue complication probability (p=0.028 and p=0.043, respectively). All techniques reduced PTV conformity; GH and FBLS+GH trajectories reduced homogeneity but less so for FBLS+GH.

CONCLUSIONS

The geometric heuristic technique best spared OARs and reduced normal tissue complication probability, however incorporating fluence information into non-coplanar trajectory optimization maintained PTV homogeneity.

Quality of treatment plans and accuracy of in vivo portal dosimetry in hybrid intensity-modulated radiation therapy and volumetric modulated arc therapy for prostate cancer

BACKGROUND AND PURPOSE

Delivering selected parts of volumetric modulated arc therapy (VMAT) plans using step-and-shoot intensity modulated radiotherapy (IMRT) beams has the potential to increase plan quality by allowing specific aperture positioning. This study investigates the quality of treatment plans and the accuracy of in vivo portal dosimetry in such a hybrid approach for the case of prostate radiotherapy.

MATERIAL AND METHODS

Conformal and limited-modulation VMAT plans were produced, together with five hybrid IMRT/VMAT plans, in which 0%, 25%, 50%, 75% or 100% of the segments were sequenced for IMRT, while the remainder were sequenced for VMAT. Integrated portal images were predicted for the plans. The plans were then delivered as a single hybrid beam using an Elekta Synergy accelerator with Agility head to a water-equivalent phantom and treatment time, isocentric dose and portal images were measured.

RESULTS

Increasing the IMRT percentage improves dose uniformity to the planning target volume (p<0.01 for 50% IMRT or more), substantially reduces the volume of rectum irradiated to 65Gy (p=0.02 for 25% IMRT) and increases the monitor units (p<0.001). Delivery time also increases substantially. All plans show accurate delivery of dose and reliable prediction of portal images.

CONCLUSIONS

Hybrid IMRT/VMAT can be efficiently planned and delivered as a single beam sequence. Beyond 25% IMRT, the delivery time becomes unacceptably long, with increased risk of intrafraction motion, but 25% IMRT is an attractive compromise. Integrated portal images can be used to perform in vivo dosimetry for this technique.

Direct aperture optimization using an inverse form of back-projection

Direct aperture optimization (DAO) has been used to produce high dosimetric quality intensity-modulated radiotherapy (IMRT) treatment plans with fast treatment delivery by directly modeling the multileaf collimator segment shapes and weights. To improve plan quality and reduce treatment time for our in-house treatment planning system, we implemented a new DAO approach without using a global objective function (GFO). An index concept is introduced as an inverse form of back-projection used in the CT multiplicative algebraic reconstruction technique (MART). The index, introduced for IMRT optimization in this work, is analogous to the multiplicand in MART. The index is defined as the ratio of the optima over the current. It is assigned to each voxel and beamlet to optimize the fluence map. The indices for beamlets and segments are used to optimize multileaf collimator (MLC) segment shapes and segment weights, respectively. Preliminary data show that without sacrificing dosimetric quality, the implementation of the DAO reduced average IMRT treatment time from 13 min to 8 min for the prostate, and from 15 min to 9 min for the head and neck using our in-house treatment planning system PlanUNC. The DAO approach has also shown promise in optimizing rotational IMRT with burst mode in a head and neck test case.

An experimental evaluation of the Agility MLC for motion-compensated VMAT delivery

An algorithm for dynamic multileaf-collimator (dMLC) tracking of a target performing a known a priori, rigid-body motion during volumetric modulated arc therapy (VMAT), has been experimentally validated and applied to investigate the potential of the Agility (Elekta AB, Stockholm, Sweden) multileaf-collimator (MLC) for use in motion-compensated VMAT delivery. For five VMAT patients, dosimetric measurements were performed using the Delta(4) radiation detector (ScandiDos, Uppsala, Sweden) and the accuracy of dMLC tracking was evaluated using a gamma-analysis, with threshold levels of 3% for dose and 3 mm for distance-to-agreement. For a motion trajectory with components in two orthogonal directions, the mean gamma-analysis pass rate without tracking was found to be 58.0%, 59.0% and 60.9% and was increased to 89.1%, 88.3% and 93.1% with MLC tracking, for time periods of motion of 4 s, 6 s and 10 s respectively. Simulations were performed to compare the efficiency of the Agility MLC with the MLCi MLC when used for motion-compensated VMAT delivery for the same treatment plans and motion trajectories. Delivery time increases from a static-tumour to dMLC-tracking VMAT delivery were observed in the range 0%–20% for the Agility, and 0%–57% with the MLCi, indicating that the increased leaf speed of the Agility MLC is beneficial for MLC tracking during lung radiotherapy.

Trajectory optimization for dynamic couch rotation during volumetric modulated arc radiotherapy

Non-coplanar radiation beams are often used in three-dimensional conformal and intensity modulated radiotherapy to reduce dose to organs at risk (OAR) by geometric avoidance. In volumetric modulated arc radiotherapy (VMAT) non-coplanar geometries are generally achieved by applying patient couch rotations to single or multiple full or partial arcs. This paper presents a trajectory optimization method for a non-coplanar technique, dynamic couch rotation during VMAT (DCR–VMAT), which combines ray tracing with a graph search algorithm. Four clinical test cases (partial breast, brain, prostate only, and prostate and pelvic nodes) were used to evaluate the potential OAR sparing for trajectory-optimized DCR–VMAT plans, compared with standard coplanar VMAT. In each case, ray tracing was performed and a cost map reflecting the number of OAR voxels intersected for each potential source position was generated. The least-cost path through the cost map, corresponding to an optimal DCR–VMAT trajectory, was determined using Dijkstra's algorithm. Results show that trajectory optimization can reduce dose to specified OARs for plans otherwise comparable to conventional coplanar VMAT techniques. For the partial breast case, the mean heart dose was reduced by 53%. In the brain case, the maximum lens doses were reduced by 61% (left) and 77% (right) and the globes by 37% (left) and 40% (right). Bowel mean dose was reduced by 15% in the prostate only case. For the prostate and pelvic nodes case, the bowel V50 Gy and V60 Gy were reduced by 9% and 45% respectively. Future work will involve further development of the algorithm and assessment of its performance over a larger number of cases in site-specific cohorts.