冠状动脉周围脂肪组学预测MACE的系统综述与Meta分析
单纯组学模型AUC为0.78,三联组合模型AUC最高为0.87。提示PCAT放射组学可能较传统临床与影像模型提供增量预测价值,但仍需标准化与多中心前瞻性验证。
要点
- 单纯放射组学模型敏感度0.70、特异度0.74、AUC 0.78,效能中等
- 放射组学+临床+影像三联模型预测效能最高,AUC达0.87
- 放射组学模型AUC较临床模型高0.05、较影像模型高0.18
核心结论
- 基于CCTA的PCAT放射组学对冠心病患者MACE具有中等预测效能,并可能较传统临床与影像模型提供增量价值
- 三联组合模型即放射组学+临床+影像表现最优,AUC为0.87,LnDOR为4.05
- 证据总体确定性为中等,临床转化前需要标准化影像组学流程及多中心前瞻性验证
研究设计与样本
- 按分类器、感兴趣区和随访时长进行亚组分析,并评估方法学质量与证据确定性
关键影像表现/征象
- PCAT放射组学特征可反映冠状动脉局部炎症与微结构变化,在CCTA上表现为脂肪组织异质性改变
- 单纯放射组学模型敏感度0.70、特异度0.74,整体区分能力中等
诊断标准与关键数值
- 单纯放射组学模型:敏感度0.70,特异度0.74,AUC 0.78
- 放射组学+临床模型:AUC 0.80
- 放射组学+影像模型:敏感度0.89,LnDOR 2.93
- 三联组合模型即放射组学+临床+影像:AUC 0.87,LnDOR 4.05
- 放射组学模型较临床模型ΔAUC=0.05,较影像模型ΔAUC=0.18;三联模型较基础模型ΔAUC=0.06,ΔLnDOR=2.21
临床应用与阅片要点
- 对行CCTA评估冠心病的患者,PCAT放射组学可作为传统危险因素与斑块指标之外的定量补充信息
- 联合临床变量可小幅改善预测,联合影像信息改善更大,报告时可关注多模型联合判别思路
- 当前不宜单独依据PCAT放射组学作出临床决策,应结合常规临床与影像评估
- 阅片时需注意不同后处理流程与感兴趣区勾画方式可能影响组学特征稳定性
局限性
- 缺乏标准化放射组学流程与多中心前瞻性外部验证
- 方法学质量评分平均仅18/36
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: Pericoronary adipose tissue (PCAT) reflects local coronary inflammation and microstructural changes and may predict major adverse cardiac events (MACE). Radiomics extracts high-dimensional features from PCAT on coronary CT angiography, capturing tissue heterogeneity beyond conventional risk factors or plaque metrics. This study evaluated the diagnostic performance of PCAT radiomics for MACE prediction.
MATERIALS AND METHODS: PubMed, Scopus, and Web of Science were searched from inception to October 2025. Ten retrospective studies were included. Diagnostic metrics were extracted and pooled using random-effects models. Subgroup analyses were performed by classifier, region of interest, and follow-up duration. Methodological quality and certainty of evidence were assessed.
RESULTS: Radiomics-only models showed moderate performance (sensitivity 0.70, specificity 0.74, area under the curve (AUC) 0.78). Combined models improved discrimination, with radiomics + clinical (AUC 0.80) and radiomics + imaging (sensitivity 0.89; diagnostic odds ratio (LnDOR) 2.93). Triple-combination models achieved the highest performance (AUC 0.87; LnDOR 4.05). Radiomics models showed higher AUC than clinical (ΔAUC = 0.05) and imaging models (ΔAUC = 0.18), with inconsistent sensitivity and specificity differences. Adding clinical variables provided modest improvement, whereas imaging integration yielded greater gains. Triple models showed the largest improvement (ΔAUC = 0.06; ΔLnDOR = 2.21). Mean Radiomics Quality Score was 18/36, and overall evidence certainty was moderate.
CONCLUSION: PCAT radiomics derived from CCTA shows moderate predictive performance for MACE in patients with coronary artery disease and may provide incremental value over conventional clinical and imaging models. Standardized radiomics pipelines and multicenter prospective validation are required for clinical translation.
KEY POINTS: Question Can quantitative analysis of pericoronary adipose tissue on coronary computed tomography angiography improve the prediction of major adverse cardiac events beyond clinical risk factors? Findings Radiomics models showed higher AUC than clinical and imaging models, while combined models demonstrated the highest predictive performance across included studies. Clinical relevance Pericoronary adipose tissue radiomics may provide additional quantitative information on coronary inflammation and may offer potential incremental value for risk prediction in patients undergoing coronary computed tomography angiography. However, further external validation and standardization are required before clinical implementation.
引用
- Mirzohreh, Dehghan, Sadeghi, et al. Pericoronary fat radiomics on coronary CT angiography for predicting major adverse cardiac events: a systematic review and meta-analysis. Eur Radiol. 2026. PMID: 42521849. doi:10.1007/s00330-026-12778-z
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。