Dataset Information

Radiomics Model to Predict Early Progression of Nonmetastatic Nasopharyngeal Carcinoma after Intensity Modulation Radiation Therapy: A Multicenter Study.

ABSTRACT:

Purpose

To examine the prognostic value of a machine learning model trained with pretreatment MRI radiomic features in the assessment of patients with nonmetastatic nasopharyngeal carcinoma (NPC) who are at risk for 3-year disease progression after intensity-modulated radiation therapy and to explain the radiomics features in the model.

Materials and methods

A total of 277 patients with nonmetastatic NPC admitted between March 2008 and December 2014 at two imaging centers were retrospectively reviewed. Patients were allocated to a discovery or validation cohort based on where they underwent MRI (discovery cohort, n = 217; validation cohort, n = 60). A total of 525 radiomics features extracted from contrast material-enhanced T1- or T2-weighted MRI studies and five clinical features were subjected to radiomic machine learning modeling to predict 3-year disease progression. Feature selection was performed by analyzing robustness to resampling, reproducibility between observers, and redundancy. Features for the final model were selected with Kaplan-Meier analysis and the log-rank test. A support vector machine was used as the classifier for the model. To interpret the pattern learned from the model, Shapley additive explanations (SHAP) was applied.

Results

The final model yielded an area under the receiver operating characteristic curve of 0.80 in both the discovery (95% bootstrap confidence interval: 0.80, 0.81) and independent validation (95% bootstrap confidence interval: 0.73, 0.89) cohorts. Analysis with SHAP revealed that tumor shape sphericity, first-order mean absolute deviation, T stage, and overall stage were important factors in 3-year disease progression.

Conclusion

These results add to the growing evidence of the role of radiomics in the assessment of NPC. By using explanatory techniques, such as SHAP, the complex interaction of features learned by the model may be understood.© RSNA, 2019Supplemental material is available for this article.

SUBMITTER: Du R

PROVIDER: S-EPMC8017427 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:ImportancePatients with nonmetastatic nasopharyngeal carcinoma (NPC) are primarily treated by radiotherapy with curative intent with or without chemotherapy and often experience substantial treatment-related toxic effects even with modern radiation techniques, such as intensity-modulated radiation therapy (IMRT). Intensity-modulated proton therapy (IMPT) may improve the toxicity profile; however, there is a paucity of data given the limited availability of IMPT in regions with endemic NPC.ObjectiveTo compare toxic effects and oncologic outcomes among patients with newly diagnosed nonmetastatic NPC when treated with IMPT vs IMRT with or without chemotherapy.Design, setting, and participantsThis retrospective cohort study included 77 patients with newly diagnosed nonmetastatic NPC who received curative-intent radiotherapy with IMPT or IMRT at a tertiary academic cancer center from January 1, 2016, to December 31, 2019. Forty-eight patients with Epstein-Barr virus (EBV)-positive tumors were included in a 1:1 propensity score-matched analysis for survival outcomes. The end of the follow-up period was March 31, 2021.ExposuresIMPT vs IMRT with or without chemotherapy.Main outcomes and measuresThe main outcomes were the incidence of acute and chronic treatment-related adverse events (AEs) and oncologic outcomes, including locoregional failure-free survival (LRFS), progression-free survival (PFS), and overall survival (OS).ResultsWe identified 77 patients (25 [32.5%] women; 52 [67.5%] men; median [interquartile range] age, 48.7 [42.2-60.3] years), among whom 28 (36.4%) were treated with IMPT and 49 (63.6%) were treated with IMRT. Median (interquartile range) follow-up was 30.3 (17.9-41.5) months. On multivariable logistic regression analyses, IMPT was associated with lower likelihood of developing grade 2 or higher acute AEs compared with IMRT (odds ratio [OR], 0.15; 95% CI, 0.03-0.60; P = .01). Only 1 case (3.8%) of a chronic grade 3 or higher AE occurred in the IMPT group compared with 8 cases (16.3%) in the IMRT group (OR, 0.21; 95% CI, 0.01-1.21; P = .15). Propensity score matching generated a balanced cohort of 48 patients (24 IMPT vs 24 IMRT) and found similar PFS in the IMPT and IMRT groups (2-year PFS, 95.7% [95% CI, 87.7%-100%] vs 76.7% [95% CI, 60.7%-97.0%]; hazard ratio [HR], 0.31; 95% CI, 0.07-1.47; P = .14). No locoregional recurrence or death was observed in the IMPT group from the matched cohort. Two-year LRFS was 100% (95% CI, 100%-100%) in the IMPT group and 86.2% (95% CI, 72.8%-100%) in the IMRT group (P = .08). Three-year OS was 100% (95% CI, 100%-100%) in the IMPT group and 94.1% (95% CI, 83.6%-100%) in the IMRT group (P = .42). Smoking history was the only clinical factor significantly associated with both poor LRFS (HR, 63.37; 95% CI, 3.25-1236.13; P = .006) and poor PFS (HR, 6.33; 95% CI, 1.16-34.57; P = .03) on multivariable analyses.Conclusions and relevanceIn this study, curative-intent radiotherapy with IMPT for nonmetastatic NPC was associated with significantly reduced acute toxicity burden in comparison with IMRT, with rare late complications and excellent oncologic outcomes, including 100% locoregional control at 2 years. Prospective trials are warranted to direct the optimal patient selection for IMPT as the primary radiotherapy modality for nonmetastatic NPC.

Project description:Background and purpose: Adaptive radiotherapy (ART) can compensate for the dosimetric impacts induced by anatomic and geometric variations in patients with nasopharyngeal carcinoma (NPC); Yet, the need for ART can only be assessed during the radiation treatment and the implementation of ART is resource intensive. Therefore, we aimed to determine tumoral biomarkers using pre-treatment MR images for predicting ART eligibility in NPC patients prior to the start of treatment. Methods: Seventy patients with biopsy-proven NPC (Stage II-IVB) in 2015 were enrolled into this retrospective study. Pre-treatment contrast-enhanced T1-w (CET1-w), T2-w MR images were processed and filtered using Laplacian of Gaussian (LoG) filter before radiomic features extraction. A total of 479 radiomics features, including the first-order (n = 90), shape (n = 14), and texture features (n = 375), were initially extracted from Gross-Tumor-Volume of primary tumor (GTVnp) using CET1-w, T2-w MR images. Patients were randomly divided into a training set (n = 51) and testing set (n = 19). The least absolute shrinkage and selection operator (LASSO) logistic regression model was applied for radiomic model construction in training set to select the most predictive features to predict patients who were replanned and assessed in the testing set. A double cross-validation approach of 100 resampled iterations with 3-fold nested cross-validation was employed in LASSO during model construction. The predictive performance of each model was evaluated using the area under the receiver operator characteristic (ROC) curve (AUC). Results: In the present cohort, 13 of 70 patients (18.6%) underwent ART. Average AUCs in training and testing sets were 0.962 (95%CI: 0.961-0.963) and 0.852 (95%CI: 0.847-0.857) with 8 selected features for CET1-w model; 0.895 (95%CI: 0.893-0.896) and 0.750 (95%CI: 0.745-0.755) with 6 selected features for T2-w model; and 0.984 (95%CI: 0.983-0.984) and 0.930 (95%CI: 0.928-0.933) with 6 selected features for joint T1-T2 model, respectively. In general, the joint T1-T2 model outperformed either CET1-w or T2-w model alone. Conclusions: Our study successfully showed promising capability of MRI-based radiomics features for pre-treatment identification of ART eligibility in NPC patients.

Project description:ImportanceBecause of tumor heterogeneity, overall survival (OS) differs significantly among individuals with nasopharyngeal carcinoma (NPC), even among those with the same clinical stage. Relying solely on TNM staging to guide treatment remains imperfect.ObjectivesTo establish a comprehensive nomogram to estimate individualized OS and to explore stratified treatment regimens for risk subgroups in nonmetastatic NPC.Design, setting, and participantsThis cohort study included 8093 patients diagnosed with NPC at a single center in China from April 2009 to December 2015. The sample was split into a training cohort (5398 participants [66.7%]) and validation cohort (2695 [33.3%]). Data were analyzed in May 2020.ExposuresAge, T stage, N stage, Epstein-Barr virus (EBV) DNA level, serum lactate dehydrogenase (LDH) levels, and albumin (ALB) levels.Main outcomes and measuresThe primary end point was OS. The nomogram for estimating OS was generated based on multivariate Cox proportional hazards regression. The performance of the nomogram was quantified using Harrell concordance index (C index), the area under the curve (AUC) of the receiver operating characteristic curve, and a calibration curve. OS rates were established using the Kaplan-Meier method, and intersubgroup differences were examined by the log-rank test.ResultsAmong the 8093 participants, 5688 (70.3%) were men, and the median age at diagnosis was 45 years (range, 7-85 years). Six variables (age, T stage, N stage, EBV DNA levels, LDH levels, and ALB levels) were identified through multivariate Cox regression and incorporated into a nomogram to estimate OS. The resulting nomogram showed excellent discriminative ability and significantly outperformed the eighth edition of the American Joint Committee on Cancer/Union for International Cancer Control TNM staging system for estimating OS (C index, 0.716 [95% CI, 0.698-0.734] vs 0.643 [95% CI, 0.624-0.661]; P < .001; AUC, 0.717 [95% CI, 0.698-0.737] vs 0.643 [95% CI, 0.623-0.662]; P < .001), and the calibration curves showed satisfactory agreement between the actual and nomogram-estimated OS rates. The validation cohort confirmed the results. Patients were stratified into 4 risk groups based on the 25th, 50th, and 75th percentile score values estimated from the nomogram. The 4 nomogram-defined risk groups demonstrated significantly different intergroup OS (3-year OS rates: risk group 1, 1328 of 1345 [98.7%]; risk group 2, 1289 of 1341 [96.1%]; risk group 3, 1222 of 1321 [92.5%]; risk group 4, 1173 of 1391 [84.3%]; P < .001). These risk groups were associated with the efficacy of different treatment regimens. For example, for risk group 4, induction chemotherapy with concurrent chemoradiotherapy was associated with a significantly better OS than concurrent chemoradiotherapy (log-rank P = .008) and intensity-modulated radiotherapy alone (log-rank P < .001).Conclusions and relevanceIn this study, the proposed nomogram model enabled individualized prognostication of OS and could help to guide risk-adapted treatment for patients with nonmetastatic NPC.