skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: MO-FG-202-09: Virtual IMRT QA Using Machine Learning: A Multi-Institutional Validation

Abstract

Purpose: To validate a machine learning approach to Virtual IMRT QA for accurately predicting gamma passing rates using different QA devices at different institutions. Methods: A Virtual IMRT QA was constructed using a machine learning algorithm based on 416 IMRT plans, in which QA measurements were performed using diode-array detectors and a 3%local/3mm with 10% threshold. An independent set of 139 IMRT measurements from a different institution, with QA data based on portal dosimetry using the same gamma index and 10% threshold, was used to further test the algorithm. Plans were characterized by 90 different complexity metrics. A weighted poison regression with Lasso regularization was trained to predict passing rates using the complexity metrics as input. Results: In addition to predicting passing rates with 3% accuracy for all composite plans using diode-array detectors, passing rates for portal dosimetry on per-beam basis were predicted with an error <3.5% for 120 IMRT measurements. The remaining measurements (19) had large areas of low CU, where portal dosimetry has larger disagreement with the calculated dose and, as such, large errors were expected. These beams need to be further modeled to correct the under-response in low dose regions. Important features selected by Lasso to predictmore » gamma passing rates were: complete irradiated area outline (CIAO) area, jaw position, fraction of MLC leafs with gaps smaller than 20 mm or 5mm, fraction of area receiving less than 50% of the total CU, fraction of the area receiving dose from penumbra, weighted Average Irregularity Factor, duty cycle among others. Conclusion: We have demonstrated that the Virtual IMRT QA can predict passing rates using different QA devices and across multiple institutions. Prediction of QA passing rates could have profound implications on the current IMRT process.« less

Authors:
; ;  [1]; ;  [2]
  1. University of Pennsylvania, Philadelphia, PA (United States)
  2. Memorial Sloan-Kettering Cancer Center, New York, NY (United States)
Publication Date:
OSTI Identifier:
22653880
Resource Type:
Journal Article
Resource Relation:
Journal Name: Medical Physics; Journal Volume: 43; Journal Issue: 6; Other Information: (c) 2016 American Association of Physicists in Medicine; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
60 APPLIED LIFE SCIENCES; 61 RADIATION PROTECTION AND DOSIMETRY; DOSIMETRY; FORECASTING; LEARNING; RADIATION DOSES; RADIOTHERAPY; RATS; VALIDATION

Citation Formats

Valdes, G, Scheuermann, R, Solberg, T, Chan, M, and Deasy, J. MO-FG-202-09: Virtual IMRT QA Using Machine Learning: A Multi-Institutional Validation. United States: N. p., 2016. Web. doi:10.1118/1.4957313.
Valdes, G, Scheuermann, R, Solberg, T, Chan, M, & Deasy, J. MO-FG-202-09: Virtual IMRT QA Using Machine Learning: A Multi-Institutional Validation. United States. doi:10.1118/1.4957313.
Valdes, G, Scheuermann, R, Solberg, T, Chan, M, and Deasy, J. 2016. "MO-FG-202-09: Virtual IMRT QA Using Machine Learning: A Multi-Institutional Validation". United States. doi:10.1118/1.4957313.
@article{osti_22653880,
title = {MO-FG-202-09: Virtual IMRT QA Using Machine Learning: A Multi-Institutional Validation},
author = {Valdes, G and Scheuermann, R and Solberg, T and Chan, M and Deasy, J},
abstractNote = {Purpose: To validate a machine learning approach to Virtual IMRT QA for accurately predicting gamma passing rates using different QA devices at different institutions. Methods: A Virtual IMRT QA was constructed using a machine learning algorithm based on 416 IMRT plans, in which QA measurements were performed using diode-array detectors and a 3%local/3mm with 10% threshold. An independent set of 139 IMRT measurements from a different institution, with QA data based on portal dosimetry using the same gamma index and 10% threshold, was used to further test the algorithm. Plans were characterized by 90 different complexity metrics. A weighted poison regression with Lasso regularization was trained to predict passing rates using the complexity metrics as input. Results: In addition to predicting passing rates with 3% accuracy for all composite plans using diode-array detectors, passing rates for portal dosimetry on per-beam basis were predicted with an error <3.5% for 120 IMRT measurements. The remaining measurements (19) had large areas of low CU, where portal dosimetry has larger disagreement with the calculated dose and, as such, large errors were expected. These beams need to be further modeled to correct the under-response in low dose regions. Important features selected by Lasso to predict gamma passing rates were: complete irradiated area outline (CIAO) area, jaw position, fraction of MLC leafs with gaps smaller than 20 mm or 5mm, fraction of area receiving less than 50% of the total CU, fraction of the area receiving dose from penumbra, weighted Average Irregularity Factor, duty cycle among others. Conclusion: We have demonstrated that the Virtual IMRT QA can predict passing rates using different QA devices and across multiple institutions. Prediction of QA passing rates could have profound implications on the current IMRT process.},
doi = {10.1118/1.4957313},
journal = {Medical Physics},
number = 6,
volume = 43,
place = {United States},
year = 2016,
month = 6
}
  • Purpose: To predict organ-at-risk (OAR) complications as a function of dose-volume (DV) constraint settings without explicit plan computation in a multiplan intensity-modulated radiotherapy (IMRT) framework. Methods and Materials: Several plans were generated by varying the DV constraints (input features) on the OARs (multiplan framework), and the DV levels achieved by the OARs in the plans (plan properties) were modeled as a function of the imposed DV constraint settings. OAR complications were then predicted for each of the plans by using the imposed DV constraints alone (features) or in combination with modeled DV levels (plan properties) as input to machine learningmore » (ML) algorithms. These ML approaches were used to model two OAR complications after head-and-neck and prostate IMRT: xerostomia, and Grade 2 rectal bleeding. Two-fold cross-validation was used for model verification and mean errors are reported. Results: Errors for modeling the achieved DV values as a function of constraint settings were 0-6%. In the head-and-neck case, the mean absolute prediction error of the saliva flow rate normalized to the pretreatment saliva flow rate was 0.42% with a 95% confidence interval of (0.41-0.43%). In the prostate case, an average prediction accuracy of 97.04% with a 95% confidence interval of (96.67-97.41%) was achieved for Grade 2 rectal bleeding complications. Conclusions: ML can be used for predicting OAR complications during treatment planning allowing for alternative DV constraint settings to be assessed within the planning framework.« less
  • Purpose: To ensure plan quality for adaptive IMRT of the prostate, we developed a quantitative evaluation tool using a machine learning approach. This tool generates dose volume histograms (DVHs) of organs-at-risk (OARs) based on prior plans as a reference, to be compared with the adaptive plan derived from fluence map deformation. Methods: Under the same configuration using seven-field 15 MV photon beams, DVHs of OARs (bladder and rectum) were estimated based on anatomical information of the patient and a model learned from a database of high quality prior plans. In this study, the anatomical information was characterized by the organmore » volumes and distance-to-target histogram (DTH). The database consists of 198 high quality prostate plans and was validated with 14 cases outside the training pool. Principal component analysis (PCA) was applied to DVHs and DTHs to quantify their salient features. Then, support vector regression (SVR) was implemented to establish the correlation between the features of the DVH and the anatomical information. Results: DVH/DTH curves could be characterized sufficiently just using only two or three truncated principal components, thus, patient anatomical information was quantified with reduced numbers of variables. The evaluation of the model using the test data set demonstrated its accuracy {approx}80% in prediction and effectiveness in improving ART planning quality. Conclusions: An adaptive IMRT plan quality evaluation tool based on machine learning has been developed, which estimates OAR sparing and provides reference in evaluating ART.« less
  • A machine learning–based framework for modeling the error introduced by surrogate models of parameterized dynamical systems is proposed. The framework entails the use of high-dimensional regression techniques (eg, random forests, and LASSO) to map a large set of inexpensively computed “error indicators” (ie, features) produced by the surrogate model at a given time instance to a prediction of the surrogate-model error in a quantity of interest (QoI). This eliminates the need for the user to hand-select a small number of informative features. The methodology requires a training set of parameter instances at which the time-dependent surrogate-model error is computed bymore » simulating both the high-fidelity and surrogate models. Using these training data, the method first determines regression-model locality (via classification or clustering) and subsequently constructs a “local” regression model to predict the time-instantaneous error within each identified region of feature space. We consider 2 uses for the resulting error model: (1) as a correction to the surrogate-model QoI prediction at each time instance and (2) as a way to statistically model arbitrary functions of the time-dependent surrogate-model error (eg, time-integrated errors). We then apply the proposed framework to model errors in reduced-order models of nonlinear oil-water subsurface flow simulations, with time-varying well-control (bottom-hole pressure) parameters. The reduced-order models used in this work entail application of trajectory piecewise linearization in conjunction with proper orthogonal decomposition. Moreover, when the first use of the method is considered, numerical experiments demonstrate consistent improvement in accuracy in the time-instantaneous QoI prediction relative to the original surrogate model, across a large number of test cases. When the second use is considered, results show that the proposed method provides accurate statistical predictions of the time- and well-averaged errors.« less
  • Purpose: AAPM TG114 does not cover the independent verification for IMRT. We conducted a study of independent dose verification for IMRT in seven institutes to show the feasibility. Methods: 384 IMRT plans in the sites of prostate and head and neck (HN) were collected from the institutes, where the planning was performed using Eclipse and Pinnacle3 with the two techniques of step and shoot (S&S) and sliding window (SW). All of the institutes used a same independent dose verification software program (Simple MU Analysis: SMU, Triangle Product, Ishikawa, JP), which is Clarkson-based and CT images were used to compute radiologicalmore » path length. An ion-chamber measurement in a water-equivalent slab phantom was performed to compare the doses computed using the TPS and an independent dose verification program. Additionally, the agreement in dose computed in patient CT images between using the TPS and using the SMU was assessed. The dose of the composite beams in the plan was evaluated. Results: The agreement between the measurement and the SMU were −2.3±1.9 % and −5.6±3.6 % for prostate and HN sites, respectively. The agreement between the TPSs and the SMU were −2.1±1.9 % and −3.0±3.7 for prostate and HN sites, respectively. There was a negative systematic difference with similar standard deviation and the difference was larger in the HN site. The S&S technique showed a statistically significant difference between the SW. Because the Clarkson-based method in the independent program underestimated (cannot consider) the dose under the MLC. Conclusion: The accuracy would be improved when the Clarkson-based algorithm should be modified for IMRT and the tolerance level would be within 5%.« less
  • Purpose: Knowledge-based Planning (KBP) founded on prior planning experience and Auto-Planning Engine (APE; commercialized in Pinnacle v9.10 TPS) based on progressive optimization algorithm both aim to eliminate the trial-and-error process in radiotherapy inverse planning. This study investigates the performance of the approaches in a multi-institutional setting to evaluate their functionalities in oropharyngeal cancer and offers suggestions how they can be implemented in the clinic. Methods: Radboud University Medical Center (RUMC) provided 35 oropharyngeal cancer patients (SIB-IMRT with two-dose-level prescription: 68 Gy to PTV68 and 50.3 Gy to PTV50.3) with corresponding comparative APE plans. Johns Hopkins University (JHU) contributed to amore » three-dose-level (70 Gy 63 Gy and 58.1 Gy) plan library for RUMC’s patient KBP generation. MedStar Georgetown University Hospital (MGUH) contributed to a KBP approach employing overlap-volume histogram (OVH-KBP) for generating RUMC’s patient KBP plans using JHU’s plan library. Since both approaches need their own user-defined parameters as initial inputs the first 10 patients were set aside as training set to finalize them. Meanwhile cross-institutional comparisons and adjustments were implemented for investigating institutions’ protocol discrepancies and the approaches’ user-defined parameters were updated accordingly. The finalized parameters were then applied to the remaining 25 patients for OVH-KBP and APE generation. A Wilcoxon rank-sum test was used for statistical comparison with significance level of p<0.05. Results: On average PTV68’s V95 was 96.5% in APE plans vs. 97% in OVH-KBP plans (p=0.36); PTV50.3’s V95 in APE plans was 97.8% vs.97.6% in OVH-KBP plans (p=0.6); cord’s D0.1 cc was 38.6 Gy in OVH-KBP plans vs. 43.7 Gy in APE plans (p=0.0001); mean doses to larynxes oral cavities parotids and submandibular glands were similar with p>0.2. Conclusions: The study demonstrates that KBP and APE can generate plans of comparable quality in a multi-institutional setting. Variations in clinical protocols can be effectively addressed for cross-institutional adaptations. Binbin Wu and Todd McNutt are the co-inventors of a patent associated with the proposed knowledge-based planning system which was licensed to Varian Medical Systems in 2015; This research was in part supported by Philips Radiation Oncology Systems.« less