人工智能相关技术在肺结节诊断中的应用:叙述性综述及临床实践与研究建议
该叙述性综述检索近10年PubMed、Web of Science、Cochrane Library及CNKI文献,分析AI在肺结节影像、病理和基因组诊断中的应用现状。AI可提升早期诊断准确性和个体化决策效率,支持良恶性鉴别、病理类型和基因突变预测,但模型标准化不足、外部验证稀缺,尚需大规模前瞻性研究支持临床应用。
要点
- AI在肺结节影像、病理和基因组诊断中均有应用,可提升早期诊断准确性和个体化决策效率。
- 现有证据支持AI用于良恶性肺结节鉴别、病理类型预测和基因突变预测。
- 模型标准化不足且外部验证稀缺,临床推广仍需大规模前瞻性研究支持。
核心结论
- AI有望在肺结节诊断、影像组学、病理和基因组学中发挥重要作用。
- 现有证据可支持AI用于良恶性肺结节鉴别、提高诊断效率、预测病理类型和基因突变。
- AI模型标准化不足、外部验证稀缺,尚需大规模前瞻性研究支持临床应用。
研究设计与样本
- 叙述性综述,检索PubMed、Web of Science、Cochrane Library和中国知网(CNKI)数据库。
- 检索限定近10年英文和中文文献,对参与者年龄或国家无限制。
- 检索关键词包括肺癌、磨玻璃结节(GGN)、人工智能、深度学习、机器学习、影像组学、肺癌病理和基因组学。
主要发现
- 综述了AI在肺结节影像、病理和遗传学诊断中的应用。
- AI技术近期在医疗领域得到广泛应用,显著提升早期诊断准确性和个体化决策效率。
- AI模型标准化不足、外部验证稀缺,大规模前瞻性研究仍欠缺。
临床怎么用
- 现有证据支持将AI用于良恶性肺结节鉴别以提高诊断效率。
- AI可用于预测肺结节病理类型和基因突变,辅助临床评估与治疗决策。
- 临床应用AI时需注意模型标准化不足和外部验证稀缺的局限。
局限性
- AI模型标准化不足。
- 外部验证稀缺。
- 尚需大规模前瞻性研究支持临床应用。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
J Thorac Dis. 2025 Aug 31;17(8):6326-6338. doi: 10.21037/jtd-2025-1512. Epub 2025 Aug 28. ABSTRACT BACKGROUND AND OBJECTIVE: Lung cancer has the highest global incidence and mortality rate among malignancies, primarily due to the complexity and diversity of early lung nodules and tumor changes, in addition to the involvement of numerous other factors, such as delayed screening for early lung nodules and difficulties in treatment. With the continuous optimization of artificial intelligence (AI) models, novel paradigms are emerging in the detection of malignant pulmonary nodules, histopathologic subtyping, and genomic prediction-advances that increasingly inform clinical assessment and therapeutic decision-making. This article reviews Chinese and English research over the past 10 years and explores the current application status of AI in the diagnosis of pulmonary nodules. METHODS: Literature available primarily from the PubMed, Web of Science, Cochrane Library, and China National Knowledge Infrastructure (CNKI) databases was collected and analyzed to characterize the current state of AI application in the diagnosis of pulmonary nodules across different disciplines. The following keywords were used in the search: "lung cancer", "ground-glass nodules (GGN)", "artificial intelligence", "deep learning", "machine learning", "radiomics", "pathology of lung cancer", and "genomics". The search is limited to English and Chinese publications within the last 10 years, with no restrictions on the age or country of the participants. KEY CONTENT AND FINDINGS: This study reviews the application of AI in pulmonary nodule diagnosis across imaging, pathology, and genetics. Recently, AI technology has found extensive applications in the healthcare sector. AI significantly enhances the accuracy of early diagnosis and the efficiency of individualized decision-making. However, the standardization of models is insufficient, external validation is scarce, and large-scale prospective studies are still needed to support clinical applications. CONCLUSIONS: AI is expected to play a significant role in the diagnosis, radiomics, pathology and genomics of pulmonary nodules in the future. Existing evidence can support the differentiation between benign and malignant pulmonary nodules, improve diagnostic efficiency, and predict pathological types and gene mutations. Thus, AI has a promising application prospect in clinical practice. PMID: 40950910 | PMC: PMC12433047 | DOI: 10.21037/jtd-202
引用
- 指南共识 · 肺结节与肺癌筛查:Narrative review of the application of artificial intelligence-related technologies in the diagnosis of pulmonary nodules with recommendations for clinical practice and future research
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。