Predicting BRAFV600E Mutation in Papillary Thyroid Carcinoma via dual-energy CT-based Radiomics

Yi Rongqigang Xiewuli TangKang Li^*

• Chongqing General Hospital, Chongqing, China

Introduction: This study focused on the creation and validation of a radiomics model that employs dual-energy computed tomography (DECT) for the preoperative prediction of the BRAF V600 mutation in patients with papillary thyroid carcinoma (PTC).We retrospectively recruited 138 individuals diagnosed with PTC who had undergone preoperative DECT and obtained their pathological findings. On the basis of the gene sequencing results, the patients were categorized into two groups, namely, patients with the BRAF V600E mutation (n=93) and those with the wild-type BRAF gene (n=45). The patients were subsequently allocated randomly to a training cohort (n=97) and a test cohort (n=41). Rad-scores were constructed by utilizing 3D-SLICER to extract radiomic characteristics from both arterial-and venous-phase tumor images. In the training cohort, the random forest algorithm combined with 10-fold cross-validation was used for feature selection, and logistic regression was applied to build models on the basis of spectral parameters (Model 1) and radiomic features (Model 2), as well as a joint model (Model 3). The model underwent 5-fold cross-validation in the training cohort, followed by final validation in the test cohort.The effectiveness of the model was evaluated through receiver operating : characteristic (ROC) curve analysis, decision curve analysis (DCA), and calibration curve analysis. Additionally, SHapley Additive exPlanations (SHAP) was used to interpret the model's predictions. Results: Model 1 was constructed on the basis of the standardized iodine concentration in the venous phase (nICVP), whereas Model 2 was constructed on the basis of the Rad scores in the arterial and venous phases. Model 3 demonstrated superior performance in predicting the BRAF V600E mutation, achieving an area under the curve (AUC) of 0.95 in the test cohort. Calibration curve analyses and DCA further confirmed the robust efficacy of the combined model. A SHAP summary plot illustrated the impact of each feature on the model, whereas a SHAP force plot showed the integration of feature effects on individual responses. Discussion: The DECT-based radiomic model has good predictive value for the BRAF V600E mutation in PTC patients.