Clinical application of a template-guided automated planning routine.

PURPOSE: Determine the dosimetric quality and the planning time reduction when utilizing a template-based automated planning application. METHODS: A software application integrated through the treatment planning system application programing interface, QuickPlan, was developed to facilitate automated planning using configurable templates for contouring, knowledge-based planning structure matching, field design, and algorithm settings. Validations are performed at various levels of the planning procedure and assist in the evaluation of readiness of the CT image, structure set, and plan layout for automated planning. QuickPlan is evaluated dosimetrically against 22 hippocampal-avoidance whole brain radiotherapy patients. The required times to treatment plan generation are compared for the validations set as well as 10 prospective patients whose plans have been automated by QuickPlan. RESULTS: The generations of 22 automated treatment plans are compared against a manual replanning using an identical process, resulting in dosimetric differences of minor clinical significance. The target dose to 2% volume and homogeneity index result in significantly decreased values for automated plans, whereas other dose metric evaluations are nonsignificant. The time to generate the treatment plans is reduced for all automated plans with a median difference of 9' 50″ ± 4' 33″. CONCLUSIONS: Template-based automated planning allows for reduced treatment planning time with consistent optimization structure creation, treatment field creation, plan optimization, and dose calculation with similar dosimetric quality. This process has potential expansion to numerous disease sites.

Determination of the electronic portal imaging device pixel-sensitivity-map for quality assurance applications. Part 1: Comparison of methods.

PURPOSE: Calibration of a radiotherapy electronic portal imaging device (EPID) using the pixel-sensitivity-map (PSM) in place of the flood field correction improves the utility of the EPID for quality assurance applications. Multiple methods are available for determining the PSM and this study provides an evaluation to inform on which is superior. METHODS: Three different empirical methods ("Calvary Mater Newcastle" [CMN], "Varian," and "WashU") and a Monte Carlo-based method of PSM determination were investigated on a single Varian TrueBeam STx linear accelerator (linac) with an aS1200 EPID panel. PSM measurements were performed for each empirical method three successive times using the 6 MV beam. The resulting PSM from each method was compared to the Monte Carlo method as a reference using 2D percentage deviation maps and histograms plus crossplane profiles. The repeatability of generated PSMs was also assessed via 2D standard deviation (SD) maps and histograms. Additionally, the Beam-Response generated by removal of the PSM from a raw EPID image for each method was visually contrasted. Finally, the practicality of each method was assessed qualitatively and via the measured time required to acquire and export the required images. RESULTS: The median pixel-by-pixel percentage deviation between each of the empirical PSM methods and the Monte Carlo PSM was -0.36%, 0.24%, and 0.74% for the CMN, Varian, and WashU methods, respectively. Ninety-five percent of pixels were found to be repeatable to within -0.21%, 0.08%, 0.19%, and 0.35% (1 SD) for the CMN, Monte Carlo, Varian, and WashU methods, respectively. The WashU method was found to be quickest for data acquisition and export and the CMN the slowest. CONCLUSION: For the first time four methods of generating the EPID PSM have been compared in detail and strengths and weaknesses of each method have been identified. All methods are considered likely to be clinically acceptable and with similar practical requirements.

Determination of the electronic portal imaging device pixel-sensitivity-map for quality assurance applications. Part 2: Photon beam dependence.

PURPOSE: The EPID PSM is a useful EPID calibration method for QA applications. The dependence of the EPID PSM on the photon beam used to acquire it has been investigated in this study for the four available PSM methods. The aim is to inform upon the viability of applying a single PSM for all available photon beams to simplify PSM implementation and maintenance. METHODS: Four methods of PSM determination were each measured once in a single session on a single TrueBeam ® STx linac using 6 MV, 10 MV, 6 MV Flattening-Filter-Free (FFF), and 10 MV FFF photon beams. The resultant PSM was assessed for both intra- and inter-method beam dependence via comparison between PSM of the same method compared to the 6 MV PSM and via comparison between PSM of the same beam with the corresponding Monte Carlo PSM. Comparisons were performed via 2D percentage deviation plots with associated histograms, 1D crossplane profiles, and via mean, median, and standard deviation percentage deviation statistics. Generated beam-response was compared qualitatively via 1D crossplane profile comparison and quantitatively via symmetry assessment with comparison to the IC profiler device. RESULTS: The Varian method provided the most consistent PSM with varying photon beam, with median percent deviation from the 6 MV PSM within 0.14% for all other beams. Qualitatively, each method provided similar beam-response profiles. The measured beam-response symmetry agreed to within 0.2% between the Calvary Mater Newcastle (CMN) method and IC profiler, but agreement reduced to within 0.9% and 2.2% for the Varian and WashU methods. PSM percent deviation with Monte Carlo PSM was within 0.75% for all methods and beams. CONCLUSION: Results suggest that the PSM may be independent of photon beam to clinically relevant levels. The Varian method of PSM determination introduces the least beam dependence into the measured PSM.

Dosimetric impact of range uncertainty in passive scattering proton therapy.

PURPOSE: The objective of this study was to investigate the dosimetric impact of range uncertainty in a large cohort of patients receiving passive scatter proton therapy. METHODS: A cohort of 120 patients were reviewed in this study retrospectively, of which 61 were brain, 39 lung, and 20 prostate patients. Range uncertainties of ±3.5% (overshooting and undershooting by 3.5%, respectively) were added and recalculated on the original plans, which had been planned according to our clinical planning protocol while keeping beamlines, apertures, compensators, and dose grids intact. Changes in the coverage on CTV and DVH for critical organs were compared and analyzed. Correlation between dose change and minimal distance between CTV and critical organs were also investigated. RESULTS: Although CTV coverages and maximum dose to critical organs were largely maintained for most brain patients, large variations over 5% were still observed sporadically. Critical organs, such as brainstem and chiasm, could still be affected by range uncertainty at 4 cm away from CTV. Coverage and OARs in lung and prostate patients were less likely to be affected by range uncertainty with very few exceptions. CONCLUSION: The margin recipe in modern TPS leads to clinically acceptable OAR doses in the presence of range uncertainties. However, range uncertainties still pose a noticeable challenge for small but critical serial organs near tumors, and occasionally for large parallel organs that are located distal to incident proton beams.

Application programming interface guided QA plan generation and analysis automation.

PURPOSE: Linear accelerator quality assurance (QA) in radiation therapy is a time consuming but fundamental part of ensuring the performance characteristics of radiation delivering machines. The goal of this work is to develop an automated and standardized QA plan generation and analysis system in the Oncology Information System (OIS) to streamline the QA process. METHODS: Automating the QA process includes two software components: the AutoQA Builder to generate daily, monthly, quarterly, and miscellaneous periodic linear accelerator QA plans within the Treatment Planning System (TPS) and the AutoQA Analysis to analyze images collected on the Electronic Portal Imaging Device (EPID) allowing for a rapid analysis of the acquired QA images. To verify the results of the automated QA analysis, results were compared to the current standard for QA assessment for the jaw junction, light-radiation coincidence, picket fence, and volumetric modulated arc therapy (VMAT) QA plans across three linacs and over a 6-month period. RESULTS: The AutoQA Builder application has been utilized clinically 322 times to create QA patients, construct phantom images, and deploy common periodic QA tests across multiple institutions, linear accelerators, and physicists. Comparing the AutoQA Analysis results with our current institutional QA standard the mean difference of the ratio of intensity values within the field-matched junction and ball-bearing position detection was 0.012 ± 0.053 (P = 0.159) and is 0.011 ± 0.224 mm (P = 0.355), respectively. Analysis of VMAT QA plans resulted in a maximum percentage difference of 0.3%. CONCLUSION: The automated creation and analysis of quality assurance plans using multiple APIs can be of immediate benefit to linear accelerator quality assurance efficiency and standardization. QA plan creation can be done without following tedious procedures through API assistance, and analysis can be performed inside of the clinical OIS in an automated fashion.

Sensitivity analysis of Monte Carlo model of a gantry-mounted passively scattered proton system.

PURPOSE: This study aimed to present guidance on the correlation between treatment nozzle and proton source parameters, and dose distribution of a passive double scattering compact proton therapy unit, known as Mevion S250. METHODS: All 24 beam options were modeled using the MCNPX MC code. The calculated physical dose for pristine peak, profiles, and spread out Bragg peak (SOBP) were benchmarked with the measured data. Track-averaged LET (LETt ) and dose-averaged LET (LETd ) distributions were also calculated. For the sensitivity investigations, proton beam line parameters including Average Energy (AE), Energy Spread (ES), Spot Size (SS), Beam Angle (BA), Beam Offset (OA), and Second scatter Offset (SO) from central Axis, and also First Scatter (FS) thickness were simulated in different stages to obtain the uncertainty of the derived results on the physical dose and LET distribution in a water phantom. RESULTS: For the physical dose distribution, the MCNPX MC model matched measurements data for all the options to within 2 mm and 2% criterion. The Mevion S250 was found to have a LETt between 0.46 and 8.76 keV.µm-1 and a corresponding LETd between 0.84 and 15.91 keV.µm-1 . For all the options, the AE and ES had the greatest effect on the resulting depth of pristine peak and peak-to-plateau ratio respectively. BA, OA, and SO significantly decreased the flatness and symmetry of the profiles. The LETs were found to be sensitive to the AE, ES, and SS, especially in the peak region. CONCLUSIONS: This study revealed the importance of considering detailed beam parameters, and identifying those that resulted in large effects on the physical dose distribution and LETs for a compact proton therapy machine.

A Monte Carlo based analytic model of the in-room neutron ambient dose equivalent for a Mevion gantry-mounted passively scattered proton system.

The goal of this study was to develop a Monte Carlo (MC)-based analytical model that can predict the in-room ambient dose equivalent from a Mevion gantry-mounted passively scattered proton system. The Mevion S250 and treatment vault were simulated using the MCNPX MC code. The results of the in-room neutron dose measurements, using an FHT 762 WENDI-II detector, were employed to benchmark the MC-derived values. After tuning the MCNPX MC code, for the same beam delivery parameters, the code was used to calculate the neutron spectra and ambient dose equivalent in the vault and at varying angles from the isocenter. Then, based on the calculations, an analytical model was reconstructed and data were fitted to derive the model parameters at 95% confidence intervals (CI). The MCNPX codes were tuned to within about 19% of the measured values for most of the measurements in the vault. For the maze, up to 0.08 mSv Gy-1 discrepancies were found between the experimental measurements and MCNPX calculated results. The analytical model showed up to 18% discrepancy for distances between 100 and 600 cm from the isocenter compared to the MC calculations. The model may underestimate the neutron ambient dose equivalent up to 21% for distances less than 100 cm from the isocenter. The proposed analytical model can be used to estimate the contribution of the secondary neutron dose from the Mevion S250 for the design of local shielding inside the proton therapy treatment vault.

Normalize the response of EPID in pursuit of linear accelerator dosimetry standardization.

Normalize the response of electronic portal imaging device (EPID) is the first step toward an EPID-based standardization of Linear Accelerator (linac) dosimetry quality assurance. In this study, we described an approach to generate two-dimensional (2D) pixel sensitivity maps (PSM) for EPIDs response normalization utilizing an alternative beam and dark-field (ABDF) image acquisition technique and large overlapping field irradiations. The automated image acquisition was performed by XML-controlled machine operation and the PSM was generated based on a recursive calculation algorithm for Varian linacs equipped with aS1000 and aS1200 imager panels. Cross-comparisons of normalized beam profiles and 1.5%/1.5 mm 1D Gamma analysis was adopted to quantify the improvement of beam profile matching before and after PSM corrections. PSMs were derived for both photon (6, 10, 15 MV) and electron (6, 20 MeV) beams via proposed method. The PSM-corrected images reproduced a horn-shaped profile for photon beams and a relative uniform profiles for electrons. For dosimetrically matched linacs equipped with aS1000 panels, PSM-corrected images showed increased 1D-Gamma passing rates for all energies, with an average 10.5% improvement for crossline and 37% for inline beam profiles. Similar improvements in the phantom study were observed with a maximum improvement of 32% for 15 MV and 22% for 20 MeV. The PSM value showed no significant change for all energies over a 3-month period. In conclusion, the proposed approach correct EPID response for both aS1000 and aS1200 panels. This strategy enables the possibility to standardize linac dosimetry QA and to benchmark linac performance utilizing EPID as the common detector.

A method to reconstruct and apply 3D primary fluence for treatment delivery verification.

MOTIVATION: In this study, a method is reported to perform IMRT and VMAT treatment delivery verification using 3D volumetric primary beam fluences reconstructed directly from planned beam parameters and treatment delivery records. The goals of this paper are to demonstrate that 1) 3D beam fluences can be reconstructed efficiently, 2) quality assurance (QA) based on the reconstructed 3D fluences is capable of detecting additional treatment delivery errors, particularly for VMAT plans, beyond those identifiable by other existing treatment delivery verification methods, and 3) QA results based on 3D fluence calculation (3DFC) are correlated with QA results based on physical phantom measurements and radiation dose recalculations. METHODS: Using beam parameters extracted from DICOM plan files and treatment delivery log files, 3D volumetric primary fluences are reconstructed by forward-projecting the beam apertures, defined by the MLC leaf positions and modulated by beam MU values, at all gantry angles using first-order ray tracing. Treatment delivery verifications are performed by comparing 3D fluences reconstructed using beam parameters in delivery log files against those reconstructed from treatment plans. Passing rates are then determined using both voxel intensity differences and a 3D gamma analysis. QA sensitivity to various sources of errors is defined as the observed differences in passing rates. Correlations between passing rates obtained from QA derived from both 3D fluence calculations and physical measurements are investigated prospectively using 20 clinical treatment plans with artificially introduced machine delivery errors. RESULTS: Studies with artificially introduced errors show that common treatment delivery problems including gantry angle errors, MU errors, jaw position errors, collimator rotation errors, and MLC leaf position errors were detectable at less than normal machine tolerances. The reported 3DFC QA method has greater sensitivity than measurement-based QA methods. Statistical analysis-based Spearman's correlations shows that the 3DFC QA passing rates are significantly correlated with passing rates of physical phantom measurement-based QA methods. CONCLUSION: Among measurement-less treatment delivery verification methods, the reported 3DFC method is less demanding than those based on full dose re-calculations, and more comprehensive than those that solely checks beam parameters in treatment log files. With QA passing rates correlating to measurement-based passing rates, the 3DFC QA results could be useful for complementing the physical phantom measurements, or verifying treatment deliveries when physical measurements are not available. For the past 4+ years, the reported method has been implemented at authors' institution 1) as a complementary metric to physical phantom measurements for pretreatment, patient-specific QA of IMRT and VMAT plans, and 2) as an important part of the log file-based automated verification of daily patient treatment deliveries. It has been demonstrated to be useful in catching both treatment plan data transfer errors and treatment delivery problems.

Clinical implementation of multisequence MRI-based adaptive intracavitary brachytherapy for cervix cancer.

The purpose of this study was to describe the clinical implementation of a magnetic resonance image (MRI)-based approach for adaptive intracavitary brachytherapy (ICBT) of cervix cancer patients. Patients were implanted with titanium tandem and colpostats. MR imaging was performed on a 1.5-T Philips scanner using T2-weighted (T2W), proton-density weighted (PDW), and diffusion-weighted (DW) imaging sequences. Apparent diffusion coefficient (ADC) maps were generated from the DW images. All images were fused. T2W images were used for the definition of organs at risk (OARs) and dose points. ADC maps in conjunction with T2W images were used for target delineation. PDW images were used for applicator definition. Forward treatment planning was performed using standard source distribution rules normalized to Point A. Point doses and dose-volume parameters for the tumor and OARs were exported to an automated dose-tracking application. Brachytherapy doses were adapted for tumor shrinkage and OAR variations during the course of therapy. The MRI-based ICBT approach described here has been clinically implemented and is carried out for each brachytherapy fraction. Total procedure time from patient preparation to delivery of treatment is typically 2 hrs. Implementation of our tech-nique for structure delineation, applicator definition, dose tracking, and adaptation is demonstrated using treated patient examples. Based on published recommendations and our clinical experience in the radiation treatment of cervix cancer patients, we have refined our standard approach to ICBT by 1) incorporating a multisequence MRI technique for improved visualization of the target, OARs, and applicator, and by 2) implementing dose adaptation by use of automated dose tracking tools.

Commissioning and initial experience with the first clinical gantry-mounted proton therapy system.

The purpose of this study is to describe the comprehensive commissioning process and initial clinical experience of the Mevion S250 proton therapy system, a gantry-mounted, single-room proton therapy platform clinically implemented in the S. Lee Kling Proton Therapy Center at Barnes-Jewish Hospital in St. Louis, MO, USA. The Mevion S250 system integrates a compact synchrocyclotron with a C-inner gantry, an image guidance system and a 6D robotic couch into a beam delivery platform. We present our commissioning process and initial clinical experience, including i) CT calibration; ii) beam data acquisition and machine characteristics; iii) dosimetric commissioning of the treatment planning system; iv) validation through the Imaging and Radiation Oncology Core credentialing process, including irradiations on the spine, prostate, brain, and lung phantoms; v) evaluation of localization accuracy of the image guidance system; and vi) initial clinical experience. Clinically, the system operates well and has provided an excellent platform for the treatment of diseases with protons.

Use of diverging apertures to minimize the edge scatter in passive scattering proton therapy.

The purpose of this study was to evaluate the use of diverging-cut aperture to minimize collimator contamination in proton therapy. Two sets of apertures with nondivergent and divergent edge were fabricated to produce a 10 cm × 10 cm field at the radiation isocenter of a single-room proton therapy unit. Transverse profiles were acquired in a scanning water tank with both aperture sets. Up to 9.5% extra dose was observed from aperture scattering near the field edges with the nondivergent aperture set at 2 cm above the water surface and remained 3.0% at depth of 10 cm. For the divergent set, the contamination was reduced to less than 3.5% and 1.3%, respectively. Our study demonstrated that scattering from apertures contaminated the dose distribution near the field edge at shallow depth. A diverging-cut aperture was capable of reducing the contamination and is recommended for use in passive scattering proton therapy, especially when critical organs are lateral and proximal to the target at shallow depth.

A pretreatment multiparametric MRI-based radiomics-clinical machine learning model for predicting radiation-induced temporal lobe injury in patients with nasopharyngeal carcinoma.

BACKGROUND: To establish and validate a machine learning model using pretreatment multiparametric magnetic resonance imaging-based radiomics data with clinical data to predict radiation-induced temporal lobe injury (RTLI) in patients with nasopharyngeal carcinoma (NPC) after intensity-modulated radiotherapy (IMRT). METHODS: Data from 230 patients with NPC who received IMRT (130 with RTLI and 130 without) were randomly divided into the training (n = 161) and validation cohort (n = 69) with a ratio of 7:3. Radiomics features were extracted from pretreatment apparent diffusion coefficient (ADC) map, T2-weighted imaging (T2WI), and CE-T1-weighted imaging (CE-T1WI). T-test, spearman rank correlation, and least absolute shrinkage and selection operator (LASSO) algorithm were employed to identify significant radiomics features. Clinical features were selected with univariate and multivariate analyses. Radiomics and clinical models were constructed using multiple machine learning classifiers, and a clinical-radiomics nomogram that combined clinical with radiomics features was developed. Receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA) were drawn to compare and verify the predictive performances of the clinical model, radiomics model, and clinical-radiomics nomogram. RESULTS: A total of 5064 radiomics features were extracted, from which 52 radiomics features were selected to construct the radiomics signature. The AUC of the radiomics signature based on multiparametric MRI was 0.980 in the training cohort and 0.969 in the validation cohort, outperforming the radiomics signature only based on T2WI and CE-T1WI (p < 0.05), which highlighted the significance of the DWI sequence in the prediction of temporal lobe injury. The area under the curve (AUC) of the clinical model was 0.895 in the training cohort and 0.905 in the validation cohort. The nomogram, which integrated radiomics and clinical features, demonstrated an impressive AUC value of 0.984 in the validation set; however, no statistically significant difference was observed compared to the radiomics model. The calibration curve and decision curve analysis of the nomogram demonstrated excellent predictive performance and clinical feasibility. CONCLUSIONS: The clinical-radiomics nomogram, integrating clinical features with radiomics features derived from pretreatment multiparametric MRI, exhibits compelling predictive performance for RTLI in patients diagnosed with NPC.

Simulation-Free Radiation Therapy: An Emerging Form of Treatment Planning to Expedite Plan Generation for Patients Receiving Palliative Radiation Therapy.

Purpose: Herein we report the clinical and dosimetric experience for patients with metastases treated with palliative simulation-free radiation therapy (SFRT) at a single institution. Methods and Materials: SFRT was performed at a single institution. Multiple fractionation regimens were used. Diagnostic imaging was used for treatment planning. Patient characteristics as well as planning and treatment time points were collected. A matched cohort of patients with conventional computed tomography simulation radiation therapy (CTRT) was acquired to evaluate for differences in planning and treatment time. SFRT dosimetry was evaluated to determine the fidelity of SFRT. Descriptive statistics were calculated on all variables and statistical significance was evaluated using the Wilcoxon signed rank test and t test methods. Results: Thirty sessions of SFRT were performed and matched with 30 sessions of CTRT. Seventy percent of SFRT and 63% of CTRT treatments were single fraction. The median time to plan generation was 0.88 days (0.19-1.47) for SFRT and 1.90 days (0.39-5.23) for CTRT (P = .02). The total treatment time was 41 minutes (28-64) for SFRT and 30 minutes (21-45) for CTRT (P = .02). In the SFRT courses, the maximum and mean deviations in the actual delivered dose from the approved plans for the maximum dose were 4.1% and 0.07%, respectively. All deliveries were within a 5% threshold and deemed clinically acceptable. Conclusions: Palliative SFRT is an emerging technique that allowed for a statistically significant lower time to plan generation and was dosimetrically acceptable. This benefit must be weighed against increased total treatment time for patients receiving SFRT compared with CTRT, and appropriate patient selection is critical.

Remote Global Radiation Oncology Education and Training: A Pathway to Increase Access to High-Quality Radiation Therapy Services in Low- and Middle-Income Countries.

Purpose: There is a vital need to train radiation therapy professionals in low- and middle-income countries (LMICs) to develop sustainable cancer treatment capacity and infrastructure. LMICs have started to introduce intensity modulated radiation therapy (IMRT), which is the standard of care in high-income countries, because of improved outcomes and reduced toxicities. This work reports the efficacy of a complementary asynchronous plus synchronous virtual-training approach on improving radiation therapy professions' self-confidence levels and evaluating participants' attitudes toward asynchronous and synchronous didactic hands-on learning in 3 LMICs. Methods and Materials: Training was provided to 37 participants from Uganda, Guatemala, and Mongolia, which included 4 theoretical lectures, 4 hands-on sessions, and 8 self-guided online videos. The 36-day training focused on IMRT contouring, site-specific target/organ definition, planning/optimization, and quality assurance. Participants completed pre- and postsession confidence surveys on a 0 to 10 scale, which was converted to a 5-point Likert rating scale to evaluate the training outcomes. The pros and cons of the 3 different training formats were compared. Results: The participants included 15 (40.5%) radiation oncologists, 11 (29.7%) medical physicists, 6 (16.2%) radiation therapists, and 5 (13.5%) dosimetrists. Approximately 50% had more than 10 years of radiation therapy experience, 70.8% had no formal IMRT training, and only 25% had IMRT at their institutions. The average experience and confidence levels in using IMRT at baseline were 3.2 and 2.9, which increased to 5.2 and 4.9 (P < .001) after the theoretical training. After the hands-on training, the experience and confidence levels further improved to 5.4 and 5.5 (P < .001). After the self-guided training, the confidence levels increased further to 6.9 (P < .01). Among the 3 different training sessions, hands-on trainings (58.3%) were most helpful for the development of participants' IMRT skills, followed by theoretical sessions with 25%. Conclusions: After completing the training sessions, Uganda and Mongolia started IMRT treatments. Remote training provides an excellent and feasible e-learning platform to train radiation therapy professionals in LMICs. The training program improved the IMRT confidence levels and treatment delivery. The hands-on trainings were most preferred.

Evaluation of the efficiency and effectiveness of independent dose calculation followed by machine log file analysis against conventional measurement based IMRT QA.

Experimental methods are commonly used for patient-specific IMRT delivery verification. There are a variety of IMRT QA techniques which have been proposed and clinically used with a common understanding that not one single method can detect all possible errors. The aim of this work was to compare the efficiency and effectiveness of independent dose calculation followed by machine log file analysis to conventional measurement-based methods in detecting errors in IMRT delivery. Sixteen IMRT treatment plans (5 head-and-neck, 3 rectum, 3 breast, and 5 prostate plans) created with a commercial treatment planning system (TPS) were recalculated on a QA phantom. All treatment plans underwent ion chamber (IC) and 2D diode array measurements. The same set of plans was also recomputed with another commercial treatment planning system and the two sets of calculations were compared. The deviations between dosimetric measurements and independent dose calculation were evaluated. The comparisons included evaluations of DVHs and point doses calculated by the two TPS systems. Machine log files were captured during pretreatment composite point dose measurements and analyzed to verify data transfer and performance of the delivery machine. Average deviation between IC measurements and point dose calculations with the two TPSs for head-and-neck plans were 1.2 ± 1.3% and 1.4 ± 1.6%, respectively. For 2D diode array measurements, the mean gamma value with 3% dose difference and 3 mm distance-to-agreement was within 1.5% for 13 of 16 plans. The mean 3D dose differences calculated from two TPSs were within 3% for head-and-neck cases and within 2% for other plans. The machine log file analysis showed that the gantry angle, jaw position, collimator angle, and MUs were consistent as planned, and maximal MLC position error was less than 0.5 mm. The independent dose calculation followed by the machine log analysis takes an average 47 ± 6 minutes, while the experimental approach (using IC and 2D diode array measurements) takes an average about 2 hours in our clinic. Independent dose calculation followed by machine log file analysis can be a reliable tool to verify IMRT treatments. Additionally, independent dose calculations have the potential to identify several problems (heterogeneity calculations, data corruptions, system failures) with the primary TPS, which generally are not identifiable with a measurement-based approach. Additionally, machine log file analysis can identify many problems (gantry, collimator, jaw setting) which also may not be detected with a measurement-based approach. Machine log file analysis could also detect performance problems for individual MLC leaves which could be masked in the analysis of a measured fluence.

Improvement of IMRT QA prediction using imaging-based neural architecture search.

PURPOSE: Machine learning (ML) has been used to predict the gamma passing rate (GPR) of intensity-modulated radiation therapy (IMRT) QA results. In this work, we applied a novel neural architecture search to automatically tune and search for the best deep neural networks instead of using hand-designed deep learning architectures. METHOD AND MATERIALS: One hundred and eighty-two IMRT plans were created and delivered with portal dosimetry. A total of 1497 fields for multiple treatment sites were delivered and measured by portal imagers. Gamma criteria of 2%/2 mm with a 5% threshold were used. Fluence maps calculated for each plan were used as inputs to a convolution neural network (CNN). Auto-Keras was implemented to search for the best CNN architecture for fluence image regression. The network morphism was adopted in the searching process, in which the base models were ResNet and DenseNet. The performance of this CNN approach was compared with tree-based ML models previously developed for this application, using the same dataset. RESULTS: The deep-learning-based approach had 98.3% of predictions within 3% of the measured 2%/2-mm GPRs with a maximum error of 3.1% and a mean absolute error of less than 1%. Our results show that this novel architecture search approach achieves comparable performance to the machine-learning-based approaches with handcrafted features. CONCLUSIONS: We implemented a novel CNN model using imaging-based neural architecture for IMRT QA prediction. The imaging-based deep-learning method does not require a manual extraction of relevant features and is able to automatically select the best network architecture.

Synchronized high-speed scintillation imaging of proton beams, generated by a gantry-mounted synchrocyclotron, on a pulse-by-pulse basis.

BACKGROUND: With the emergence of more complex and novel proton delivery techniques, there is a need for quality assurance tools with high spatiotemporal resolution to conveniently measure the spatial and temporal properties of the beam. In this context, scintillation-based dosimeters, if synchronized with the radiation beam and corrected for ionization quenching, are appealing. PURPOSE: To develop a synchronized high-speed scintillation imaging system for characterization and verification of the proton therapy beams on a pulse-by-pulse basis. MATERIALS AND METHODS: A 30 cm × 30 cm × 5 cm block of BC-408 plastic scintillator placed in a light-tight housing was irradiated by proton beams generated by a Mevion S250 proton therapy synchrocyclotron. A high-speed camera system, placed perpendicular to the beam direction and facing the scintillator, was synchronized to the accelerator's pulses to capture images. Opening and closing of the camera's shutter was controlled by setting a proper time delay and exposure time, respectively. The scintillation signal was recorded as a set of two-dimensional (2D) images. Empirical correction factors were applied to the images to correct for the nonuniformity of the pixel sensitivity and quenching of the scintillator. Proton range and modulation were obtained from the corrected images. RESULTS: The camera system was able to capture all data on a pulse-by-pulse basis at a rate of ∼504 frames per second. The applied empirical correction method for ionization quenching was effective and the corrected composite image provided a 2D map of dose distribution. The measured range (depth of distal 90%) through scintillation imaging agreed within 1.2 mm with that obtained from ionization chamber measurement. CONCLUSION: A high-speed camera system capable of capturing scintillation signals from individual proton pulses was developed. The scintillation imaging system is promising for rapid proton beam characterization and verification.