Predictive Value of Ultrasound Microvascular Flow Imaging Radiomics Models for BRAF V600E Mutation Status in Thyroid Nodules

Objective: This study aims to evaluate the clinical value of a radiomics model based on microvascular flow imaging (MFI) of thyroid nodules for predicting BRAF V600E mutation status. Methods: A retrospective analysis was performed on 170 patients with thyroid nodules. Clinical data, pathological results, conventional ultrasound images, MFI images, and BRAF V600E mutation testing results were collected. Patients were randomly divided into training (n = 136) and validation (n = 34) cohorts in a 4:1 ratio. Radiomic features were extracted from MFI images by manually delineating regions of interest. Independent risk factors for BRAF V600E mutation were identified from conventional ultrasound and clinical parameters. Feature selection was conducted using the least absolute shrinkage and selection operator (LASSO) algorithm. Combined predictive models were built using five classifiers: support vector machine (SVM), random forest (RF), decision tree (DT), logistic regression (LR), and gradient boosting decision tree (GBDT). A radiomics model based solely on MFI-derived features was constructed using the best-performing classifier (SVM), and its predictive performance was compared with that of the combined model. Model performance was assessed using receiver operating characteristic curves (ROC), area under the curve (AUC), accuracy, sensitivity, and specificity. Results: Significant differences (P < 0.05) in age, vascularity, Bethesda category, morphology, and echogenicity were observed between BRAF V600E mutation-positive and mutation-negative groups. LASSO selected 18 features (15 radiomics, 2 ultrasound, 1 clinical). The SVM-based combined model achieved the highest predictive performance (validation AUC = 0.854), outperforming the standalone MFI-based radiomics model (AUC = 0.764). Conclusion: The combined model incorporating MFI-based radiomics, ultrasound, and clinical features showed superior predictive performance for BRAF V600E mutation status and could support thyroid nodule diagnosis and management.