Department of Ultrasound, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China.
Br J Radiol. 2023 Sep;96(1149):20221120. doi: 10.1259/bjr.20221120. Epub 2023 Jul 5.
The objective of this study was to establish a multimodality ultrasound prediction model based on conventional ultrasound (Con-US), shear wave elastography (SWE), and strain elastography (SE) and contrast-enhanced ultrasound (CEUS) and to explore their diagnostic values for thyroid nodules ≤ 10 mm.
This retrospective study included 198 thyroid nodules (maximum diameter≤10 mm) in 198 thyroid surgery patients who were examined preoperatively with above-mentioned methods. The pathological findings of the thyroid nodules were used as the gold standard, and there were 72 benign nodules and 126 malignant nodules. The multimodal ultrasound prediction models were developed by logistic regression analysis based on the ultrasound image appearances. The diagnostic efficacy of these prediction models was then compared and internally cross-validated in a fivefold manner.
The specific features on CEUS (enhancement boundary, enhancement direction and decreased nodule area) and the parenchyma-to-nodule strain ratio (PNSR) on SE and SWE ratio were included in the prediction model. The Model one combining American College of Radiology Thyroid Imaging Reporting and Data Systems (ACR TI-RADS) score with PNSR and SWE ratio had the highest sensitivity (92.8%), while the Model three combining TI-RADS score with PNSR, SWE ratio and specific CEUS indicators had the highest specificity, accuracy, and AUC (90.2%,91.4%, and 0.958, respectively).
The multimodality ultrasound predictive models effectively improved the differential diagnosis of thyroid nodules smaller than 10 mm.
For the differential diagnosis of thyroid nodules ≤ 10 mm, both ultrasound elastography and CEUS could be effective complements to ACR TI-RADS.
本研究旨在建立一种基于常规超声（Con-US）、剪切波弹性成像（SWE）、应变弹性成像（SE）和对比增强超声（CEUS）的多模态超声预测模型，并探讨其对直径≤10mm 的甲状腺结节的诊断价值。
本回顾性研究纳入了 198 例（最大直径≤10mm）甲状腺手术患者的 198 个甲状腺结节，术前均采用上述方法进行检查。以甲状腺结节的病理结果为金标准，其中良性结节 72 个，恶性结节 126 个。基于超声图像特征，采用逻辑回归分析建立多模态超声预测模型。然后，以 5 折交叉验证的方式比较和内部验证这些预测模型的诊断效能。
CEUS 的特定特征（增强边界、增强方向和结节面积减小）和 SE 及 SWE 比值的实质-结节应变比（PNSR）被纳入预测模型。结合 ACR TI-RADS 评分、PNSR 和 SWE 比值的模型一具有最高的敏感性（92.8%），而结合 TI-RADS 评分、PNSR、SWE 比值和特定 CEUS 指标的模型三具有最高的特异性、准确性和 AUC（分别为 90.2%、91.4%和 0.958）。
多模态超声预测模型可有效提高直径小于 10mm 的甲状腺结节的鉴别诊断效能。
对于直径≤10mm 的甲状腺结节的鉴别诊断，超声弹性成像和 CEUS 均可以作为 ACR TI-RADS 的有效补充。
