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

Title: SU-F-R-46: Predicting Distant Failure in Lung SBRT Using Multi-Objective Radiomics Model

Abstract

Purpose: To predict distant failure in lung stereotactic body radiation therapy (SBRT) in early stage non-small cell lung cancer (NSCLC) by using a new multi-objective radiomics model. Methods: Currently, most available radiomics models use the overall accuracy as the objective function. However, due to data imbalance, a single object may not reflect the performance of a predictive model. Therefore, we developed a multi-objective radiomics model which considers both sensitivity and specificity as the objective functions simultaneously. The new model is used to predict distant failure in lung SBRT using 52 patients treated at our institute. Quantitative imaging features of PET and CT as well as clinical parameters are utilized to build the predictive model. Image features include intensity features (9), textural features (12) and geometric features (8). Clinical parameters for each patient include demographic parameters (4), tumor characteristics (8), treatment faction schemes (4) and pretreatment medicines (6). The modelling procedure consists of two steps: extracting features from segmented tumors in PET and CT; and selecting features and training model parameters based on multi-objective. Support Vector Machine (SVM) is used as the predictive model, while a nondominated sorting-based multi-objective evolutionary computation algorithm II (NSGA-II) is used for solving the multi-objective optimization.more » Results: The accuracy for PET, clinical, CT, PET+clinical, PET+CT, CT+clinical, PET+CT+clinical are 71.15%, 84.62%, 84.62%, 85.54%, 82.69%, 84.62%, 86.54%, respectively. The sensitivities for the above seven combinations are 41.76%, 58.33%, 50.00%, 50.00%, 41.67%, 41.67%, 58.33%, while the specificities are 80.00%, 92.50%, 90.00%, 97.50%, 92.50%, 97.50%, 97.50%. Conclusion: A new multi-objective radiomics model for predicting distant failure in NSCLC treated with SBRT was developed. The experimental results show that the best performance can be obtained by combining all features.« less

Authors:
; ; ; ;  [1]
  1. UT Southwestern Medical Center, Dallas, TX (United States)
Publication Date:
OSTI Identifier:
22626765
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; ACCURACY; ALGORITHMS; BIOMEDICAL RADIOGRAPHY; CALCULATION METHODS; DRUGS; FAILURES; IMAGES; LUNGS; NEOPLASMS; OPTIMIZATION; PATIENTS; PERFORMANCE; POSITRON COMPUTED TOMOGRAPHY; RADIOTHERAPY; SENSITIVITY; SIMULATION; SPECIFICITY; TRAINING

Citation Formats

Zhou, Z, Folkert, M, Iyengar, P, Zhang, Y, and Wang, J. SU-F-R-46: Predicting Distant Failure in Lung SBRT Using Multi-Objective Radiomics Model. United States: N. p., 2016. Web. doi:10.1118/1.4955817.
Zhou, Z, Folkert, M, Iyengar, P, Zhang, Y, & Wang, J. SU-F-R-46: Predicting Distant Failure in Lung SBRT Using Multi-Objective Radiomics Model. United States. doi:10.1118/1.4955817.
Zhou, Z, Folkert, M, Iyengar, P, Zhang, Y, and Wang, J. 2016. "SU-F-R-46: Predicting Distant Failure in Lung SBRT Using Multi-Objective Radiomics Model". United States. doi:10.1118/1.4955817.
@article{osti_22626765,
title = {SU-F-R-46: Predicting Distant Failure in Lung SBRT Using Multi-Objective Radiomics Model},
author = {Zhou, Z and Folkert, M and Iyengar, P and Zhang, Y and Wang, J},
abstractNote = {Purpose: To predict distant failure in lung stereotactic body radiation therapy (SBRT) in early stage non-small cell lung cancer (NSCLC) by using a new multi-objective radiomics model. Methods: Currently, most available radiomics models use the overall accuracy as the objective function. However, due to data imbalance, a single object may not reflect the performance of a predictive model. Therefore, we developed a multi-objective radiomics model which considers both sensitivity and specificity as the objective functions simultaneously. The new model is used to predict distant failure in lung SBRT using 52 patients treated at our institute. Quantitative imaging features of PET and CT as well as clinical parameters are utilized to build the predictive model. Image features include intensity features (9), textural features (12) and geometric features (8). Clinical parameters for each patient include demographic parameters (4), tumor characteristics (8), treatment faction schemes (4) and pretreatment medicines (6). The modelling procedure consists of two steps: extracting features from segmented tumors in PET and CT; and selecting features and training model parameters based on multi-objective. Support Vector Machine (SVM) is used as the predictive model, while a nondominated sorting-based multi-objective evolutionary computation algorithm II (NSGA-II) is used for solving the multi-objective optimization. Results: The accuracy for PET, clinical, CT, PET+clinical, PET+CT, CT+clinical, PET+CT+clinical are 71.15%, 84.62%, 84.62%, 85.54%, 82.69%, 84.62%, 86.54%, respectively. The sensitivities for the above seven combinations are 41.76%, 58.33%, 50.00%, 50.00%, 41.67%, 41.67%, 58.33%, while the specificities are 80.00%, 92.50%, 90.00%, 97.50%, 92.50%, 97.50%, 97.50%. Conclusion: A new multi-objective radiomics model for predicting distant failure in NSCLC treated with SBRT was developed. The experimental results show that the best performance can be obtained by combining all features.},
doi = {10.1118/1.4955817},
journal = {Medical Physics},
number = 6,
volume = 43,
place = {United States},
year = 2016,
month = 6
}
  • Purpose: To develop an evidential reasoning approach for selecting the optimal solution from a Pareto solution set obtained by a multi-objective radiomics model for predicting distant failure in lung SBRT. Methods: In the multi-objective radiomics model, both sensitivity and specificity are considered as the objective functions simultaneously. A Pareto solution set with many feasible solutions will be resulted from the multi-objective optimization. In this work, an optimal solution Selection methodology for Multi-Objective radiomics Learning model using the Evidential Reasoning approach (SMOLER) was proposed to select the optimal solution from the Pareto solution set. The proposed SMOLER method used the evidentialmore » reasoning approach to calculate the utility of each solution based on pre-set optimal solution selection rules. The solution with the highest utility was chosen as the optimal solution. In SMOLER, an optimal learning model coupled with clonal selection algorithm was used to optimize model parameters. In this study, PET, CT image features and clinical parameters were utilized for predicting distant failure in lung SBRT. Results: Total 126 solution sets were generated by adjusting predictive model parameters. Each Pareto set contains 100 feasible solutions. The solution selected by SMOLER within each Pareto set was compared to the manually selected optimal solution. Five-cross-validation was used to evaluate the optimal solution selection accuracy of SMOLER. The selection accuracies for five folds were 80.00%, 69.23%, 84.00%, 84.00%, 80.00%, respectively. Conclusion: An optimal solution selection methodology for multi-objective radiomics learning model using the evidential reasoning approach (SMOLER) was proposed. Experimental results show that the optimal solution can be found in approximately 80% cases.« less
  • Objective: To investigate the prognostic significance of image gradients and in predicting clinical outcomes in a patients with non-small cell lung cancer treated with stereotactic body radiotherapy (SBRT) on 71 patients with 83 treated lesions. Methods: The records of patients treated with lung SBRT were retrospectively reviewed. When applicable, SBRT target volumes were modified to exclude any overlap with pleura, chestwall, or mediastinum. The ITK software package was utilized to generate quantitative measures of image intensity, inhomogeneity, shape morphology and first and second-order CT textures. Multivariate and univariate models containing CT features were generated to assess associations with clinicopathologic factors.more » Results: On univariate analysis, tumor size (HR 0.54, p=0.045) sumHU (HR 0.31, p=0.044) and short run grey level emphasis STD (HR 0.22, p=0.019) were associated with regional failure-free survival; meanHU (HR 0.30, p=0.035), long run emphasis (HR 0.21, p=0.011) and long run low grey level emphasis (HR 0.14, p=0.005) was associated with distant failure-free survival (DFFS). No features were significant on multivariate modeling however long run low grey level emphasis had a hazard ratio of 0.12 (p=0.061) for DFFS. Adenocarcinoma and squamous cell carcinoma differed with respect to long run emphasis STD (p=0.024), short run low grey level emphasis STD (p<0.001), and long run low grey level emphasis STD (p=0.024). Multivariate modeling of texture features associated with tumor histology was used to estimate histologies of 18 lesions treated without histologic confirmation. Of these, MVA suggested the same histology as a prior metachronous lung malignancy in 3/7 patients. Conclusion: Extracting radiomics features on clinical datasets was feasible with the ITK package with minimal effort to identify pre-treatment quantitative CT features with prognostic factors for distant control after lung SBRT.« less
  • Purpose: Early prediction of distant metastasis may provide crucial information for adaptive therapy, subsequently improving patient survival. Radiomic features that extracted from PET and CT images have been used for assessing tumor phenotype and predicting clinical outcomes. This study investigates the values of radiomic features in predicting distant metastasis (DM) in non-small cell lung cancer (NSCLC). Methods: A total of 108 patients with stage II–III lung adenocarcinoma were included in this retrospective study. Twenty radiomic features were selected (10 from CT and 10 from PET). Conventional features (metabolic tumor volume, SUV, volume and diameter) were included for comparison. Concordance indexmore » (CI) was used to evaluate features prognostic value. Noether test was used to compute p-value to consider CI significance from random (CI = 0.5) and were adjusted for multiple testing using false rate discovery (FDR). Results: A total of 70 patients had DM (64.8%) with a median time to event of 8.8 months. The median delivered dose was 60 Gy (range 33–68 Gy). None of the conventional features from PET (CI ranged from 0.51 to 0.56) or CT (CI ranged from 0.57 to 0.58) were significant from random. Five radiomics features were significantly prognostic from random for DM (p-values < 0.05). Four were extracted from CT (CI = 0.61 to 0.63, p-value <0.01) and one from PET which was also the most prognostic (CI = 0.64, p-value <0.001). Conclusion: This study demonstrated significant association between radiomic features and DM for patients with locally advanced lung adenocarcinoma. Moreover, conventional (clinically utilized) metrics were not significantly associated with DM. Radiomics can potentially help classify patients at higher risk of DM, allowing clinicians to individualize treatment, such as intensification of chemotherapy) to reduce the risk of DM and improve survival. R.M. has consulting interests with Amgen.« less
  • Purpose: Local failure after definitive stereotactic body radiation therapy (SBRT) for early-stage non-small cell lung cancer (NSCLC) is uncommon. We report the safety and efficacy of SBRT for salvage of local failure after previous SBRT with a biologically effective dose (BED) of ≥100 Gy{sub 10}. Methods and Materials: Using an institutional review board–approved lung SBRT registry, we identified all patients initially treated for early-stage NSCLC between August 2004 and January 2012 who received salvage SBRT for isolated local failure. Failure was defined radiographically and confirmed histologically unless contraindicated. All patients were treated on a Novalis/BrainLAB system using ExacTrac for imagemore » guidance, and received a BED of ≥100 Gy{sub 10} for each SBRT course. Tumor motion control involved a Bodyfix vacuum system for immobilization along with abdominal compression. Results: Of 436 patients treated from August 2004 through January 2012, we identified 22 patients with isolated local failure, 10 of whom received SBRT for salvage. The median length of follow-up was 13.8 months from salvage SBRT (range 5.3-43.5 months). Median tumor size was 3.4 cm (range 1.7-4.8 cm). Two of the 10 lesions were “central” by proximity to the mediastinum, but were outside the zone of the proximal bronchial tree. Since completing salvage, 3 patients are alive and without evidence of disease. A fourth patient died of medical comorbidities without recurrence 13.0 months after salvage SBRT. Two patients developed distant disease only. Four patients had local failure. Toxicity included grade 1-2 fatigue (3 patients) and grade 1-2 chest wall pain (5 patients). There was no grade 3-5 toxicity. Conclusions: Repeat SBRT with a BED of ≥100 Gy{sub 10} after local failure in patients with early-stage medically inoperable NSCLC was well tolerated in this series and may represent a viable salvage strategy in select patients with peripheral tumors ≤5 cm.« less
  • Purpose: To describe the development of a knowledge-based treatment planning model for lung cancer patients treated with SBRT, and to evaluate the model performance and applicability to different planning techniques and tumor locations. Methods: 105 lung SBRT plans previously treated at our institution were included in the development of the model using Varian’s RapidPlan DVH estimation algorithm. The model was trained with a combination of IMRT, VMAT, and 3D–CRT techniques. Tumor locations encompassed lesions located centrally vs peripherally (43:62), upper vs lower (62:43), and anterior vs posterior lobes (60:45). The model performance was validated with 25 cases independent of themore » training set, for both IMRT and VMAT. Model generated plans were created with only one optimization and no planner intervention. The original, general model was also divided into four separate models according to tumor location. The model was also applied using different beam templates to further improve workflow. Dose differences to targets and organs-at-risk were evaluated. Results: IMRT and VMAT RapidPlan generated plans were comparable to clinical plans with respect to target coverage and several OARs. Spinal cord dose was lowered in the model-based plans by 1Gy compared to the clinical plans, p=0.008. Splitting the model according to tumor location resulted in insignificant differences in DVH estimation. The peripheral model decreased esophagus dose to the central lesions by 0.5Gy compared to the original model, p=0.025, and the posterior model increased dose to the spinal cord by 1Gy compared to the anterior model, p=0.001. All template beam plans met OAR criteria, with 1Gy increases noted in maximum heart dose for the 9-field plans, p=0.04. Conclusion: A RapidPlan knowledge-based model for lung SBRT produces comparable results to clinical plans, with increased consistency and greater efficiency. The model encompasses both IMRT and VMAT techniques, differing tumor locations, and beam arrangements. Research supported in part by a grant from Varian Medical Systems, Palo Alto CA.« less