基于医学影像的机器学习模型预测胸部肿瘤PD-L1与CD8+TILs表达的性能:系统综述与Meta分析
该系统综述与Meta分析纳入68项研究(其中25项进入Meta分析),评估AI驱动的医学影像预测胸部肿瘤免疫微环境。汇总AUC:PD-L1为0.81,CD8+TIL为0.86;但研究间异质性显著(I²>75%),限制模型推广。
要点
- AI影像模型预测胸部肿瘤PD-L1表达汇总AUC 0.81、敏感度0.77、特异度0.73。
- 预测CD8+TIL汇总AUC 0.86、敏感度0.81、特异度0.81,性能略优于PD-L1。
- 多模态影像与深度学习模型整合可提升预测性能,但研究间异质性I²>75%。
核心结论
- AI驱动医学影像对胸部肿瘤免疫微环境(尤其PD-L1表达与CD8+TIL)表现出较强预测能力,但研究间显著异质性限制当前模型的推广性。
研究设计与样本
- 遵循PRISMA的系统综述与Meta分析,检索PubMed、Cochrane、Embase、Web of Science截至2025年7月1日,共纳入68项研究,其中25项符合Meta分析条件,采用随机效应模型汇总诊断准确性数据。
诊断标准与关键数值
- PD-L1表达预测:汇总AUC 0.81(95%CI 0.78-0.85),敏感度0.77(95%CI 0.74-0.80),特异度0.73(95%CI 0.71-0.76)。
- CD8+TIL预测:汇总AUC 0.86(95%CI 0.82-0.89),敏感度0.81(95%CI 0.72-0.87),特异度0.81(95%CI 0.76-0.84)。
- 亚组分析提示整合多模态影像与深度学习模型可提升预测性能。
临床应用与阅片要点
- 作者认为AI影像模型对胸部肿瘤PD-L1与CD8+TIL具有较强预测能力,但研究间显著异质性限制其推广性。
- 作者建议未来优先开展多中心、大规模研究,标准化影像特征提取,并整合生物学机制以提升临床适用性。
局限性
- 研究间存在显著异质性(I²>75%)。
- 显著的研究间异质性限制当前模型的推广性。
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
RATIONALE AND OBJECTIVES: Medical imaging integrated with artificial intelligence (AI) has demonstrated considerable potential in predicting the tumor immune microenvironment (TIME). However, the robustness of methodologies and predictive performance remains debated. This study systematically reviews and performs a meta-analysis to evaluate the advancements in AI-driven medical imaging for predicting the TIME in thoracic tumors, with a specific focus on Programmed Death-Ligand 1 (PD-L1) expression and CD8+ tumor-infiltrating lymphocytes (TILs). We hypothesize that medical imaging-based AI models can effectively predict key components of the tumor immune microenvironment in thoracic cancers, specifically PD-L1 expression and CD8+ TILs.
METHODS: Following PRISMA guidelines, we systematically searched PubMed, Cochrane, Embase, and Web of Science for studies published up to July 01, 2025. Studies evaluating AI-driven medical imaging for predicting thoracic tumor TIME were included. Diagnostic accuracy data were extracted, and a meta-analysis was conducted using a random-effects model to assess the predictive performance of AI models for PD-L1 expression and CD8+ TILs. Heterogeneity analysis and publication bias assessment were also performed.
RESULTS: A total of 68 studies were included, of which 25 were eligible for meta-analysis. The pooled area under the curve (AUC) for AI-driven medical imaging prediction of PD-L1 expression was 0.81 (95% CI: 0.78-0.85), with a sensitivity of 0.77 (95% CI: 0.74-0.80) and a specificity of 0.73 (95% CI: 0.71-0.76). For CD8+ TIL prediction, the pooled AUC was 0.86 (95% CI: 0.82-0.89), with a sensitivity of 0.81 (95% CI: 0.72-0.87) and a specificity of 0.81 (95% CI: 0.76-0.84). Subgroup analysis indicated that integrating multimodal imaging and deep learning models improved predictive performance. Nevertheless, considerable heterogeneity was observed among studies (I² > 75%).
CONCLUSION: AI-driven medical imaging exhibits strong predictive capability for thoracic tumor TIME, particularly in PD-L1 expression and CD8+ TIL prediction. However, significant interstudy heterogeneity limits the generalizability of current models. Future studies should prioritize multicenter, large-scale research, standardization of imaging feature extraction, and integration of biological mechanisms to improve the clinical applicability of AI models.
引用
- Liu M, Gong J, Liu Y, et al. Performance of Machine Learning Models Based on Medical Imaging in Predicting the expression of PD-L1 and CD8+TILs in Thoracic cancer: A Systematic Review and Meta-Analysis. Acad Radiol. 2025. PMID: 41206268. doi:10.1016/j.acra.2025.10.002
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。