中国实用外科杂志

中国实用外科杂志 ›› 2022, Vol. 42 ›› Issue (06): 680-684.DOI: 10.19538/j.cjps.issn1005-2208.2022.06.17

• 论蓍 • 上一篇    下一篇

动态人工智能超声辅助诊断系统对甲状腺结节诊断价值研究

王    冰1a，张明博1b，万    政1a，姚    京1a，公凤霞1a，李    晨1a，张淋淋1a，苗    欣1a，吴高松2，田    文1a   

  1. 1 中国人民解放军总医院第一医学中心  a.普通外科医学部甲状腺（疝）外科  b.超声诊断科，北京100853；2 武汉大学中南医院甲状腺乳腺外科，湖北武汉 430071
  • 出版日期:2022-06-01 发布日期:2022-06-12

  • Online:2022-06-01 Published:2022-06-12

摘要： 目的    探讨动态人工智能（AI）超声辅助诊断系统（以下简称动态AI）对甲状腺结节良恶性鉴别诊断价值及对治疗策略的指导意义。方法    选取2021年11月1日至2021年12月31日于中国人民解放军总医院第一医学中心甲状腺（疝）外科接受手术治疗的193例（共346枚结节）甲状腺结节病人，应用动态AI对结节进行良恶性鉴别，对比不同性别、年龄的病人其不同大小、性质的结节的动态AI检查结果、术前细针穿刺细胞学检查结果及术后病理学检查结果，评价动态AI诊断效能。结果    动态AI诊断甲状腺结节的灵敏度为88.5%，特异度为 76.5%，准确率为 85.0%，与术后病理学检查结果有较高一致性（Kappa＝0.643，P＜0.01）。对比动态AI检查和术前穿刺活检的灵敏度（χ2=0.210，P=0.647）和准确率（χ2=1.699，P=0.192）差异无统计学意义，特异度差异有统计学意义（χ2=7.238，P=0.007）。动态AI检查在不同性别（χ2=0.002，P=0.968）及年龄（χ2=3.913，P=0.41）人群中准确率差异无统计学意义，稳定性较好。良性结节的准确率（76.5%）较恶性结节的准确率（88.5%）低（χ2=8.184，P=0.004）。动态AI检查准确率随着结节大小增加有上升趋势（χ2=7.357，P=0.007），直径＞1 cm结节准确率可达94.4%。结论    动态AI检查对甲状腺结节的良恶性有较高的诊断价值，且无创可重复、高效便捷、安全可靠，可减少不必要的穿刺，节约医疗资源及社会成本，可有效辅助外科医师为病人制定科学合理的个体化诊治策略。

关键词: 甲状腺结节, 人工智能, 超声检查, 精准诊断

Abstract: The diagnostic value of dynamic AI intelligent assistant diagnostic system for thyroid nodules        WANG Bing*,ZHANG Ming-bo,WAN Zheng,et al. *Department of Thyroid & Hernia Surgery,Medical Department of General Surgery,the First Medical Center of Chinese PLA General Hospital,Beijing 100853,China
Corresponding author：TIAN Wen,E-mail:tianwen301_cta01@163.com
Abstract    Objective    To explore the value of dynamic AI intelligent assistant diagnostic system in differentiating benign and malignant thyroid nodules and its guiding significance for treatment strategy. Methods    193 patients with 346 thyroid nodules in total who received surgical treatment in the Department of Thyroid & Hernia Surgery of the First Medical Center of Chinese PLA General Hospital from November 1，2021 to December 31，2021 were selected. Dynamic AI was used to differentiate benign and malignant nodules. The results of dynamic AI examination，preoperative fine needle aspiration and postoperative pathological results of nodules of different sizes and nature in patients of different genders and ages were compared to evaluate the diagnostic efficacy of dynamic AI. Results    The sensitivity，specificity and accuracy of dynamic AI in diagnosing thyroid nodules were 88.5%，76.5% and 85.0%，which were consistent with postoperative pathological results. There was no statistical difference in sensitivity and accuracy between dynamic AI examination and preoperative needle biopsy，but there was statistical difference in specificity. There was no statistical difference in the accuracy of dynamic AI test in different genders and ages，showing good stability. The accuracy of benign nodules was lower than that of malignant nodules.The accuracy of dynamic AI examination increased with the increase of nodule size. The accuracy rate of nodules ＞1 cm was 94.4%. Conclusion    This research shows that dynamic AI intelligent assistant diagnostic system has high diagnostic value for thyroid nodules，which is noninvasive，repeatable，efficient，convenient and safe，needs short learning curve，reduces unnecessary puncture，and saves medical resources and the social costs. Thus it can assist surgeons to develop scientific and rational individualized treatment strategies for inpatients and outpatients.

Key words: thyroid nodule, artificial intelligence, ultrasonography, accurate diagnosis