Prediction of radiation pneumonitis after definitive radiotherapy for locally advanced non-small cell lung cancer using multi-region radiomics analysis
Abstract Objective To predict grade 2 radiation pneumonitis (RP) in locally advanced non-small cell lung cancer (NSCLC) patients with high accuracy by using multi-region radiomics analysis.Material and Methods Data from 77 patients with NSCLC who underwent definitive radiotherapy from 2008 to 2018 were analyzed. Radiomic feature extraction from the whole lung (whole-lung radiomics analysis) and imaging- and dosimetric-based segmentation (multi-region radiomics analysis) were performed. The data were split into 2 sets: 54 tumors for model training and 23 tumors for model testing. Patients with RP grade ≥2 or RP grade <2 was classified. Models were built with least absolute shrinkage and selection operator logistic regression and applied to the set of candidate predictors. To build predictive models with clinical features, machine-learning methods with neural network classifiers were used. The precision, accuracy, and sensitivity by generating confusion matrices and the areas under the receiver operating characteristic curve (AUC) for each model were evaluated. Results A total of 49383 radiomics features per patient image were extracted from the radiotherapy planning computed tomography scan. We identified 4 features selected in the whole-lung radiomics analysis and 13 radiomics features selected in the multi-region radiomics analysis for the classification. Out of 13 features, the median and high-dose region were selected from the shape analysis. The features of local intensity roughness and variation were selected from the statistical and texture analysis. The accuracy, specificity, sensitivity, and AUC were 80.1%, 79.2%, 88.9%, and 0.84 for our new method of multi-region radiomics analysis, respectively, which was improved from 60.8%, 64.6%, 53.8%, and 0.62 for whole-lung radiomics analysis. Conclusions The developed multi-region radiomics analysis can help predict grade 2 RP for NSCLC after definitive radiotherapy. The radiomics features in the median- and high-dose regions, and that of local intensity roughness and variation were important factors in predicting grade 2 RP.