Prediction model for growth trends of subpleural subsolid nodules using CT radiomics and radiological features.

Subpleural subsolid nodules (SSNs) pose challenges in early malignant transformation risk stratification, leading to over-surveillance or delayed treatment. To develop a radiological-radiomics combined model for predicting growth of subpleural SSNs and optimizing individualized follow-up strategies. This retrospective study included 494 subpleural SSNs (training set: 345; test set: 149) with ≥ 3 years follow-up CT. Radiological features (nodule type, morphology, pleural retraction) and radiomics features were analyzed. A radiomics score (Radscore) was developed using least absolute shrinkage and selection operator (LASSO) regression, and a combined model integrating radiological and radiomics predictors was constructed. Model performance was evaluated via the area under the curve (AUC), calibration curves, and decision curve analysis (DCA). The combined model demonstrated favorable performance in both training (AUC 0.896; 95% CI 0.8505-0.9425) and test sets (AUC 0.842; 95% CI 0.7185-0.9600), outperforming radiological model (train: 0.896 vs. 0.716; test: 0.842 vs. 0.741, all P < 0.05) and showed similar performance to the radiomics model (train: 0.896 vs. 0.857, p = 0.047; test: 0.842 vs. 0.840, p = 0.936). Key predictors included part-solid nodule (PSNs) type (OR 2.359, P = 0.009), irregular morphology (OR 2.917, P = 0.001), and pleural retraction (OR 2.227, P = 0.014). Notably, DCA indicated that the combined model had better clinical utility across a range of decision thresholds (10-90%), offering a higher net benefit for guiding interventions. The combined model effectively predicts subpleural SSNs growth, enabling risk stratification to reduce unnecessary follow-up and prioritize early intervention for high-risk nodules.