背景:肺癌是2020年第二大最常见的癌症,也是导致癌症死亡的主要原因。尽管人工智能(AI)辅助诊断技术已显示出前景,并已在近年来的临床实践中使用,我国尚无AI辅助CT诊断技术相关产品获得国家药品监督管理局批准。本文的目的是系统地回顾AI辅助CT诊断技术对肺结节良性或恶性分类的诊断性能,并分析中国医生对该技术的看法。
方法:根据纳入和排除标准,从6个文献数据库中检索和筛选所有相关研究。提取数据,并由两名评审员评估研究质量。估计研究异质性和发表偏倚。在中国9家公立三级医院进行了医生认知的问卷调查。荟萃分析,在系统评价中使用了元回归和单变量逻辑模型,以探讨医师的认知与他们对该技术临床应用的支持率之间的关系。
结果:有27项5,727个肺结节的研究最终纳入荟萃分析。我们发现纳入研究的质量普遍可以接受,AI辅助CT诊断技术对肺结节良恶性分类的合并敏感性和特异性分别为0.90和0.89。合并诊断比值比(DOR)为70.33。大多数接受调查的中国医生认为“减少放射科医生的工作量”和“提高诊断效率”是这项技术的重要好处。此外,诊断准确性(包括误诊)和实践经验与医师是否支持其临床应用显著相关.
结论:在肺癌诊断的背景下,AI辅助CT诊断技术对肺结节的良恶性分类具有良好的诊断性能,但其特异性有待提高。
BACKGROUND: Lung cancer was the second most commonly diagnosed cancer and the leading cause of cancer death in 2020. Although artificial intelligence (AI)-assisted diagnostic technologies have shown promise and has been used in clinical practice in recent years, no products related to AI-assisted CT diagnostic technologies for the classification of pulmonary nodules have been approved by the National Medical Products Administration in China. The objective of this article was to systematically
review the diagnostic performance of AI-assisted CT diagnostic technology for the classification of pulmonary nodules as benign or malignant and to analyze physicians\' perceptions of this technology in China.
METHODS: All relevant studies from 6 literature databases were searched and screened according to the inclusion and exclusion criteria. Data were extracted and the study quality was assessed by two reviewers. The study heterogeneity and publication bias were estimated. A questionnaire survey on the perceptions of physicians was conducted in 9 public tertiary hospitals in China. A meta-analysis, meta-regression and univariate logistic model were used in the systematic
review and to explore the association of physicians\' perceptions with their rate of support for the clinical application of the technology.
RESULTS: Twenty-seven studies with 5,727 pulmonary nodules were finally included in the meta-analysis. We found that the quality of the included studies was generally acceptable and that the pooled sensitivity and specificity of AI-assisted CT diagnostic technology for the classification of pulmonary nodules as benign or malignant were 0.90 and 0.89, respectively. The pooled diagnostic odds ratio (DOR) was 70.33. The majority of the surveyed physicians in China perceived \"reduced workload for radiologists\" and \"improved diagnostic efficiency\" as the important benefits of this technology. In addition, diagnostic accuracy (including misdiagnosis) and practical experience were significantly associated with whether physicians supported its clinical application.
CONCLUSIONS: In the context of lung cancer diagnosis, AI-assisted CT diagnostic technology for the classification of pulmonary nodules as benign or malignant has good diagnostic performance, but its specificity needs to be improved.