A combined model using pre-treatment CT radiomics and clinicopathological features of non-small cell lung cancer to predict major pathological responses after neoadjuvant chemoimmunotherapy

Abstract

Abstract Objective: To explore the correlation between clinicopathology, CT radiomics and major pathologic response of NSCLC after neoadjuvant chemoimmunotherapy, and to establish a combined model to predict MPR. Methods: The clinicopathological characteristics of NSCLC patients who received neoadjuvant chemoimmunotherapy in our hospital from January 2019 to April 2021 were retrospectively collected, and pre-treatment CT Radscore was calculated through standard radiomics workflow. Afterward, independent factors were screened, odds ratios were calculated, and a nomogram for predicting MPR was constructed. ROC curves were plotted for models. Finally, the three models were compared using Delong's test. Results: 211 NSCLC patients were enrolled in this study. Radscore and RECIST assessment results were independent factors for MPR using multivariate logistic regression analysis. The AUC of the combined model was 0.76 (95% CI: 0.68-0.84) in training group, and 0.80 (95% CI: 0.67-0.92) in validation group. Delong's test showed that the AUC of the combined model was significantly different from the radiomics model alone in the training group (p = 0.0067) and also in the validation group (P=0.0009). The calibration curve showed that the predicted MPR was significantly close to the actual MPR of the patient, and the clinical decision curve indicated that the combined model had a higher overall net benefit than the radiomics model alone. Conclusions: The combined model based on pre-treatment CT radiomics and clinicopathological features showed better predictive power than the independent radiomics model or the independent clinicopathological features, which may better guide the personalized neoadjuvant chemoimmunotherapy treatment strategy.