Application of SHAP visualization and nomogram based on PET/CT radiomics combined models for risk stratification in high-grade invasive lung adenocarcinoma

This study aimed to develop and validate a PET/CT radiomics combined model integrated with visualization tools for preoperative risk stratification of high-grade invasive lung adenocarcinoma (ILA). A total of 202 ILA patients with 207 pathologically confirmed lesions were retrospectively enrolled and randomly divided into training (70%) and validation (30%) sets. Patients with ≥ 20% micropapillary or solid subtypes were classified as the high-risk group. Multivariate logistic regression was used to develop combined clinical and radiomics models. Model performance was evaluated using ROC, calibration, and decision curve analyses, and visualized with Shapley Additive exPlanations (SHAP) and a nomogram. The combined model demonstrated superior diagnostic performance, with higher AUCs (Delong test, all P < 0.05) and calibration curves (Hosmer-Lemeshow test, all P > 0.05), outperforming the clinical model in both the training and validation datasets. It also provided greater clinical net benefit across the central decision threshold range of 0.15 to 0.85. SHAP analysis identified Radscore and SUVmax (mean SHAP values = 0.35 and 0.2, respectively) as the primary risk factors for high-grade ILA. We validated a PET/CT-based combined model utilizing SHAP analysis and a nomogram for preoperative risk stratification in high-grade ILA. The model exhibits high accuracy, commendable calibration, and clinical benefit. This noninvasive, interpretable instrument facilitates objective risk assessment, personalized decision-making, and precise management for ILA.