增强CT影像组学在鉴别TI-RADS 4A、4B类甲状腺微小结节良恶性的价值

Contrast-Enhanced CT-based Radiomics for Distinguishing Benign from Malignant Thyroid Micro-nodules in TI-RADS Category 4A and 4B

目的探讨基于增强计算机断层扫描(computed tomography,CT)影像组学在鉴别超声甲状腺成像报告和数据系统(thyroid imaging reporting and data system,TI-RADS)诊断为4A、4B类甲状腺微小结节良恶性的价值。方法回顾性入组2018年1月至2022年1月经过术后病理证实甲状腺结节患者300例(115例良性,185例恶性)。从增强CT图像中提取影像组学特征分别构建动脉期、静脉期、延迟期及多期联合(动脉期+静脉期+延迟期)影像组学模型,联合诊断效能最高的影像组学模型和临床特征构建综合诊断模型并绘制诺莫图。使用接受者工作特征曲线下面积(area under the receiver operating characteristic curve,AUC)评估模型的诊断性能,决策曲线分析(decision curve analysis,DCA)评价模型的临床效用。结果多期联合组学模型的诊断表现(训练集AUC vs测试集AUC:0.814 vs 0.718)优于单一期像组学模型(动脉期:0.730 vs 0.601;静脉期:0.794 vs 0.678;延迟期:0.793 vs 0.622);与临床模型(训练集AUC:0.732;测试集AUC:0.766)和多期联合组学模型相比,综合诊断模型在训练集(AUC:0.876)和测试集(AUC:0.813)均具有更好的诊断表现。决策曲线分析表明,该综合诊断模型具有更高的临床应用价值。结论增强CT影像组学能够在术前有效鉴别TI-RADS 4A和4B类甲状腺微小结节的良恶性。

Objective To evaluate the diagnostic value of contrast-enhanced CT-based radiomics model for distinguishing benign from malignant thyroid micro-nodules of thyroid imaging reporting and data system(TI-RADS)category 4A and 4B.Methods This retrospective study included 300 patients with TI-RADS 4A and 4B micro-nodules(115 benign,185 malignant)undergoing preoperative contrast-enhanced CT and surgery from January 2018 to January 2022.We extracted radiomics features from contrast-enhanced CT images to build arterial phase,venous phase,delayed phase and multi-phase combined(arterial phase+venous phase+delayed phase)radiomics model,integrated the most effective radiomics model and the clinical features to construct the synthetic diagnostic model and draw nomogram.The model performance was determined by the area under the receiver operating characteristic curve(AUC).The clinical usefulness was assessed by decision curve analysis(DCA).Results The diagnostic performance of multi-phase combined radiomics model(training set AUC vs test set AUC:0.814 vs 0.718)is better than that of single phase radiomics model(arterial phase:0.730 vs 0.601;venous phase:0.794 vs 0.859;delayed phase:0.793 vs 0.622);compared with the clinical model(training set AUC:0.732;test set AUC:0.766)and multi-phase combined radiomics model,the synthetic diagnostic model had better diagnosis effects in both the training set(AUC:0.876)and the test set(AUC:0.813).The decision curve analysis showed that the combine model had higher clinical application value.Conclusion Contrast-Enhanced CT-based radiomic model had good discrimination for the preoperative prediction of the benign from malignant thyroid micro-nodules in TI-RADS category 4A and 4B.