Abstract
OBJECTIVE: To develop and validate a clinical prediction model for differentiating non-small cell lung cancer (NSCLC) from benign pulmonary diseases (BPD). METHODS: The retrospective study included 226 participants in the training set (134 with NSCLC and 92 with BPD) and 98 participants in the validation set (62 with NSCLC and 36 with BPD). A logistic regression model was constructed using variables such as sex, hemoptysis, serum biomarkers (carcinoembryonic antigen [CEA] and total protein [TP]), and CT features (volume ratio of solid density region [VR_2A], volume of calcified density area [VR_3], total volume [TV], CT variance [CTV], and maximum surface area [MSA]). The model was validated on an independent cohort, and its performance was evaluated using area under the curve (AUC), accuracy, calibration curves, and decision curves. Additionally, a nomogram was developed for clinical application, and its acceptance and convenience among clinicians were assessed. RESULTS: The prediction model achieved an AUC-ROC of 0.95 in the training set and 0.82 in the validation set. Calibration and decision curves demonstrated that the model had reliable diagnostic performance and good clinical application value. CONCLUSION: The integrated prediction model combining clinical features, laboratory biomarkers, and CT features shows promise for improving the accuracy of differentiating NSCLC from BPD. Further studies are warranted to explore its potential in clinical practice.