Abstract: Background The ultrasonographic presentation of cystic solid thyroid cancer is different solid thyroid cancer，and sonographers currently have insufficient knowledge of the sonograms of cystic solid thyroid cancer，resulting in a high rate of misdiagnosis. At present，there is no relevant study both at home and abroad which applies artificial intelligence system and contrast-enhanced ultrasound（CEUS）scoring system to the clinical diagnosis of cystic solid thyroid cancer. This study inaugural combine the two new ultrasound technologies mentioned above and study the conventional ultrasound and CEUS features of cystic solid thyroid cancer， which aim to skillfully apply the relevant new technologies to improve the diagnostic accuracy of cystic solid thyroid cancer and can help to improve the prognosis of the disease。Objectives The aim of this study was to investigate the diagnostic value of the contrast-enhanced ultrasound（CEUS）scoring system， artificial intelligence and the American College of Radiology Thyroid Imaging and Reporting Data System（ACR TI-RADS）when used by sonographers individually and in combination for the diagnosis of partial cystic thyroid nodules（PCTNs）. Methods A retrospective analysis of conventional ultrasound and CEUS images of enrolled patients from First Hospital of Shanxi Medical University was performed，and a CEUS scoring system was established. The sensitivity， specificity，and area under the curve of CEUS and artificial intelligence，ACR TI-RADS individually in combination for diagnosis were compared among. Results Nine CEUS features of PCTNs were observed and summarized；Among these，seven features—enhanced ring，island-like enhancement，enhancement duration，washout time，post-enhancement nodular margin，post-enhancement solid component size change，and sparse low/no enhancement—exhibited significant intergroup differences（P<0.05 for all）and were incorporated into the CEUS scoring system. CEUS demonstrated diagnostic sensitivity of 75.44%，specificity of 81.03%，and an area under the ROC curve of 0.849. Artificial intelligence achieved diagnostic sensitivity of 82.46%，specificity of 67.24%，and an area under the ROC curve of 0.755. Physician diagnosis using ACR-TR yielded sensitivity of 84.21%，specificity of 82.76%，and an AUC of 0.880；physician diagnosis augmented by AI achieved sensitivity of 91.23%，specificity of 86.21%，and an AUC of 0.929；Physician diagnosis using ACR-TR combined with CEUS demonstrated sensitivity of 91.23%，specificity of 84.48%，and an AUC of 0.914；AI assisted physician diagnosis combined with CEUS achieved sensitivity of 94.7%，specificity of 87.9%，and an AUC of 0.941，exhibiting optimal diagnostic performance.Conclusions The CEUS scoring system established in this study，It enhances the diagnostic efficacy of PCTNs. The use of CEUS and artificial intelligence can improve the diagnostic accuracy of sonographers and improve the prognosis of PCTN patients.

Key words: Thyroid neoplasms, Partial cystic thyroid nodules, Thyroid nodule, Artificial intelligence, Ultrasonography, Scoring system, American college of radiology thyroid imaging reporting and data system

摘要： 背景 囊实性甲状腺癌的超声表现与实性甲状腺恶性肿瘤不同，目前超声医师对囊实性甲状腺癌的声像图认识不足，误诊率较高。目前国内外尚无相关研究将人工智能系统及超声造影评分系统应用于囊实性甲状腺癌的临床诊断，本研究将上述两超声新技术联合，研究囊实性甲状腺癌的常规超声及超声造影声像图特征，并熟练应用相关新技术提高对囊实性甲状腺癌的诊断准确率，有助于改善其预后。目的 探讨超声造影（CEUS）评分系统、人工智能、超声医师采用美国放射学院甲状腺影像报告和数据系统（ACR TI-RADS）单独及联合对甲状腺囊实性结节（PCTNs）的诊断价值。方法 回顾性选取2013年3月—2020年12月山西医科大学第一医院经病理确诊的PCTNs患者106例，共115枚结节。根据病理学检查结果分为良性组与恶性组，良性组患者53例（50%），58枚结节（50.4%）；恶性组患者53例（50%），57枚结节（49.6%）。分析两组患者的常规超声及CEUS图像，建立CEUS评分系统，计算并对比CEUS、人工智能、超声医师采用ACR TI-RADS三者各自及两两联合诊断、三者联合诊断时的灵敏度、特异度及受试者工作特征（ROC）曲线下面积。结果 良性组和恶性组结节的9种CEUS增强特征中，结节增强环、岛样增强、增强时间、消退时间、增强后结节边界、增强后实性部分大小改变、稀疏低/无增强比较，差异有统计学意义（P<0.05），入选CEUS评分系统。CEUS对于PCTNs的诊断灵敏度为75.44%、特异度为81.03%、ROC曲线下面积为0.849；人工智能诊断灵敏度为82.46%、特异度为67.24%、ROC曲线下面积为0.755；医师采用ACR-TR诊断灵敏度为84.21%、特异度为82.76%、ROC曲线下面积为0.880；人工智能辅助医师诊断灵敏度为91.23%、特异度为86.21%、ROC曲线下面积为0.929；医师采用ACR TR联合CEUS诊断灵敏度为91.23%、特异度为84.48%、ROC曲线下面积为0.914；人工智能辅助医师联合CEUS诊断的灵敏度为94.7%、特异度为87.9%、ROC曲线下面积为0.941，具有最佳诊断效能。结论 将人工智能、CEUS无创、高效的新技术应用于PCTNs的诊断，可提高对PCTNs诊断效能，人工智能辅助医师联合CEUS的诊断效能较高，有望实现医师对PCTNs的精准诊断，改善患者预后。

关键词: 甲状腺肿瘤, 囊实性甲状腺癌, 甲状腺结节, 人工智能, 超声检查, 评分系统, 美国放射学院甲状腺影像报告和数据系统