Work

CT-based explainable machine learning for predicting benign and malignant thyroid nodules: a multi-center study

Year: 2025

Type: article

Abstract: Objective: This study intends to construct a CT-based explainable machine learning model for preoperative prediction of thyroid nodule benignity or malignancy, aiming to provide a more accurate tool for clinical decision-making and management. Materials and methods: 41). Radiomics features were extracted from preoperative CT images, and the optimal features were selected to construct a radiomics score (Rad_Score). Clinical risk factors were identified using univariate and multivariate logistic regression. LR and SVM algorithms were used to establish three models: a clinical model, an imaging model, and a combined model (integrating clinical factors and Rad_Score). The combined model was visualized using SHAP (SHapley Additive exPlanations) analysis. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). Results: A total of 17 features were ultimately selected for Rad_Score calculation. The combined model demonstrated the best performance, with the LR combined model achieving AUC values of 0.962, 0.913, and 0.914 in the training set, internal validation set, and external validation set, respectively, all higher than the LR clinical model and LR radiomics model; and the LR combined model outperforms the SVM combined model (0.953, 0.885, and 0.842). The SHAP analysis revealed the relative importance of the key feature (Rad_score) in model prediction, enhancing model transparency. Conclusion: The combined model performs better under the LR algorithm. Combined with SHAP explainable analysis, it provides a non-invasive, efficient, and transparent tool for preoperative differentiation of benign and malignant thyroid nodules, potentially optimizing individualized clinical management.

Source: Frontiers in Oncology

DOI: 10.3389/fonc.2025.1675943