BRAF^(V600E)基因联合增强CT的列线图模型对TI-RADS 3类及以上甲状腺结节良恶性鉴别诊断的价值

Value of a column-line diagram model incorporating BRAF^(V600E)gene and enhanced CT in the differential diagnosis of thyroid nodules categorized as TI-RADS 3 and above

目的构建BRAF^(V600E)基因突变和增强CT特征的TI-RADS 3类及以上甲状腺结节良恶性预测模型,并评估其诊断效能。方法回顾性分析2022年10月~2024年2月安徽医科大学附属巢湖医院收治的251例TI-RADS3类及以上甲状腺结节患者的资料,以超声引导下细针穿刺细胞学及术后病理结果为“金标准”,其中良性结节177个,恶性结节74个。通过LASSO回归法筛选变量和选择预测因子,并建立预测模型。结果采用LASSO回归法选取4个变量,分别为年龄、BRAF^(V600E)基因、增强CT中的边界模糊、包膜不连续特征,基于这些变量构建BRAF^(V600E)基因突变联合增强CT预测模型并进一步验证。联合预测模型的AUC为0.816,高于单独增强CT预测模型(AUC=0.755),差异有统计学意义(P<0.05)。联合预测模型的敏感度为88.7%,特异度为63.5%,准确度为81.7%,Hosmer-Lemeshow拟合检验P=0.4564。联合预测模型与单独增强CT预测模型相比的净重新分类指数为0.308(0.151~0.465)(P<0.001),综合判别改善指数为0.114(0.060~0.167)(P<0.001)。决策曲线分析及校正曲线显示联合预测模型有较高的预测性能。结论BRAF^(V600E)基因联合增强CT特征的列线图模型对甲状腺TI-RADS3类及以上良恶性结节的鉴别具有较高的诊断价值。

Objective To construct a thyroid imaging reporting and data system(TI-RADS)benign-malignant prediction model for thyroid nodules categorized as TI-RADS 3 and above,incorporating both BRAF^(V600E)gene mutation status and enhanced CT features,and assess its diagnostic efficacy.Methods A retrospective analysis of data from 251 patients with TI-RADS 3 and above thyroid nodules admitted to Chaohu Hospital of Anhui Medical University from October 2022 to February 2024 were conducted.Ultrasound-guided fine-needle aspiration cytology and postoperative pathology served as the"gold standard",with 177 nodules classified as benign and 74 as malignant.The LASSO regression method was employed for variable and predictor selection,leading to the establishment of a prediction model.Results LASSO regression identified four variables for inclusion in the prediction model:age,BRAF^(V600E)gene mutation status,presence of blurred borders on enhanced CT,and discontinuity of the nodule envelope.A prediction model for BRAF^(V600E)gene mutation status in enhanced CT was developed based on these variables and subsequently validated.The AUC for the combined prediction model was 0.816,surpassing that of the enhanced CT prediction model alone(AUC=0.755)with statistical significance(P&lt;0.05).The joint prediction model demonstrated a sensitivity of 88.7%,specificity of 63.5%,and accuracy of 81.7%,with a Hosmer-Lemeshow fit test yielding P=0.4564.The net reclassification index compared to the enhanced CT prediction model alone was 0.308(0.151-0.465)(P<0.001),and the integrated discrimination improvement index was 0.114(0.060-0.167)(P<0.001).Decision curve analysis and calibration curves confirmed the high predictive performance of the combined prediction model.Conclusion The column-line diagram model combining BRAF^(V600E)gene mutation status with enhanced CT features demonstrates significant diagnostic value in distinguishing between benign and malignant nodules categorized as TI-RADS 3 and above.