Short-term treatment response assessment in non-surgical treatment of advanced non-small cell lung cancer based on radiomics of dual-energy CT

Purpose: To build and evaluate a pre-treatment dual-energy CT(DECT)-based clinical-radiomics nomogram for individualized prediction of short-term treatment response to non-surgical treatment in advanced non-small cell lung cancer (NSCLC). Methods: Pre-treatment DECT images were retrospectively collected from 98 pathologically confirmed NSCLC with clinical stage III or IV. Short-term treatment response was determined with follow-up CT of 4-6 courses of treatment. Quantitative radiomics metrics of the lesion were extracted from dual-energy mixed images at venous phase. Least absolute shrinkage and selection operator and correlation analysis were used to select the most relevant radiomics features. Radiomics model, clinical model and clinical-radiomics model were established by multivariate logistic regression. The model with the best prediction performance was visualized as a nomogram, and the consistency between the probability of the actual occurrence of the outcome and the probability predicted by the model was measured by calibration curves. Results: Clinical stage, difference in electron density in arteriovenous phase, difference in slope of energy spectrum in arteriovenous phase, and slope of energy spectrum in venous phase of the tumor were significant clinical predictors of therapy response (P < 0.05). The clinical-radiomics model showed a higher predictive capability (AUC: 0.87 and 0.85 in training and validation sets, respectively) than the radiomics models and the clinical model. The clinical-radiomics nomogram integrating the DECT radiomics signature with clinical stage and spectrum parameters showed good calibration and discrimination. Conclusion: The clinical-radiomics nomogram based on pre-treatment DECT showed good performance in predicting clinical response to non-surgical therapy in NSCLC.