文献速递胸部DRCT

胸部X线片上间质性肺异常的检测:放射科医师与人工智能的诊断效能

Detection of Interstitial Lung Abnormalities on Chest Radiographs: Diagnostic Performance of Radiologists and Artificial Intelligence.
PubMed 文献 · AJR Am J Roentgenol · 2026-04-01

这项回顾性研究纳入1168名接受胸部CT和X线检查的成年人,评估5名胸部放射科医师和一个人工智能模型在X线片上检测间质性肺异常(ILA)的效能,以CT为参考标准。放射科医师和人工智能敏感度均不理想但特异度高,人工智能对纤维化型ILA的AUC为0.92,放射科医师为0.86。X线片ILA与呼吸系统疾病相关死亡独立相关(校正亚分布HR 8.72),但研究结果不支持用X线筛查ILA。

要点

  • 放射科医师与人工智能在X线片上检测ILA的敏感度均不理想(纤维化型分别为62.7%和68.2%),但特异度高(97.8%和98.7%)。
  • 人工智能对纤维化型ILA的AUC为0.92,放射科医师为0.86(p=.06);对纤维化或非纤维化ILA,人工智能AUC为0.83,放射科医师为0.75(p=.009)。
  • X线片ILA与呼吸系统疾病相关死亡独立相关(校正亚分布HR 8.72,p<.001),但与全因死亡无关(校正HR 1.75,p=.17)。

核心结论

  • 放射科医师和人工智能在X线片上检测ILA的敏感度不理想但特异度高;尽管X线片ILA与临床相关结局相关,研究结果不支持用X线筛查ILA,无论采用放射科医师还是人工智能判读。

研究设计与样本

  • 回顾性研究,纳入2007年1月至2010年12月韩国两家机构健康体检中同时接受胸部CT和X线检查的1168名成年人(中位年龄56岁;男786例,女382例),以CT为纤维化型和非纤维化型ILA的参考标准。

诊断标准与关键数值

  • 5名胸部放射科医师独立用5分Likert量表评估X线片ILA可能性,阳性定义为≥4分。
  • 人工智能模型处理X线片生成网状或网结节状阴影概率(0-1),阳性定义为≥0.4。
  • 纤维化型ILA:放射科医师AUC 0.86、敏感度62.7%、特异度97.8%、准确率97.2%;人工智能AUC 0.92、敏感度68.2%、特异度98.7%、准确率98.1%。
  • 纤维化或非纤维化ILA:放射科医师AUC 0.75、敏感度38.5%、特异度98.0%、准确率95.9%;人工智能AUC 0.83、敏感度41.5%、特异度98.8%、准确率96.8%。

局限性

  • 原文未报告局限性。

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)。

英文原文(摘要)· 点击展开

BACKGROUND. Interstitial lung abnormalities (ILA) on chest CT are receiving growing attention given their association with progression to interstitial lung disease. Radiography's role in ILA detection is not well described. OBJECTIVE. The purpose of this study was to evaluate the diagnostic performance of radiologists and an artificial intelligence (AI) model in detecting ILA on chest radiography using CT as the reference standard. METHODS. This retrospective study included adults who underwent both chest CT and chest radiography as part of health checkup programs at two institutions in Korea between January 2007 and December 2010. Five thoracic radiologists independently assessed the likelihood of ILA on radiography using a 5-point Likert scale (positive, ≥ 4). A previously developed AI model processed radiographs to generate a probability (0-1) of reticular or reticulonodular opacities (positive, ≥ 0.4). CT served as the reference standard for fibrotic and nonfibrotic ILA. Radiologists' diagnostic performance for ILA detection was reported using mean performance metrics and compared with AI performance using generalized estimating equations. Associations of AI-based radiographic ILA, adjusting for age, sex, and smoking status, were assessed with all-cause mortality and respiratory disease-related mortality using Cox proportional hazards and Fine-Gray competing-risk regression models. RESULTS. The analysis included 1168 individuals (median age, 56 years; 786 men, 382 women). Forty-one individuals had ILA on CT (fibrotic, 22; nonfibrotic, 19). For fibrotic ILA, radiologists and AI had AUCs of 0.86 and 0.92 (p = .06), sensitivity of 62.7% and 68.2% (p = .43), specificity of 97.8% and 98.7% (p = .05), and accuracy of 97.2% and 98.1% (p = .04), respectively. For fibrotic or nonfibrotic ILA, radiologists and AI had AUCs of 0.75 and 0.83 (p = .009), sensitivity of 38.5% and 41.5% (p = .48), specificity of 98.0% and 98.8% (p = .05), and accuracy of 95.9% and 96.8% (p = .03), respectively. During a median follow-up of 11.9 years, radiographic ILA was independently associated with respiratory disease-related mortality (adjusted subdistribution HR, 8.72; p < .001) but not overall mortality (adjusted HR, 1.75; p = .17). CONCLUSION. Radiologists and AI achieved suboptimal sensitivity, albeit high specificity, for ILA detection on radiography. CLINICAL IMPACT. Despite association of radiographic ILA with a clinically relevant outcome, the findings do not support radiographic screening for ILA, whether incorporating radiologist or AI interpretation.

来源:AJR Am J Roentgenol,查看原文。版权归原作者所有。

引用

  1. Kim H, Lee JE, Chae KJ, et al. Detection of Interstitial Lung Abnormalities on Chest Radiographs: Diagnostic Performance of Radiologists and Artificial Intelligence. AJR Am J Roentgenol. 2026. PMID: 41919933. doi:10.2214/AJR.26.34686

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。