冠状动脉斑块影像组学表型预测致死性或非致死性心肌梗死:SCOT-HEART试验分析
SCOT-HEART试验的事后分析纳入1750例冠状动脉CT血管造影患者,中位随访8.6年,82例发生致死性或非致死性心肌梗死。在临床模型(含心血管风险评分、Agatston评分及阻塞性冠心病)中加入斑块负荷指标预测性能相似(C统计量0.70 vs 0.70),进一步加入特征影像组学特征后模型性能提升(C统计量0.74),且随访第5年后累积/动态AUC更高。
要点
- 基于冠状动脉CT血管造影的影像组学斑块形态特征可提升长期心肌梗死预测,优于临床因素和斑块负荷。
- 在临床模型中加入斑块负荷指标未改善预测(C统计量0.70 vs 0.70)。
- 进一步加入特征影像组学特征后模型C统计量提升至0.74,随访第5年后累积/动态AUC更高。
核心结论
- 基于影像组学的冠状动脉斑块形态精准表型分析,在临床因素和斑块负荷之上,提升了CT血管造影对长期心肌梗死的预测。
研究设计与样本
- SCOT-HEART临床试验的事后分析,纳入1750例患者(年龄58±9岁,56%男性),分析冠状动脉CT血管造影,计算斑块负荷和描述斑块形态的特征影像组学特征,参考标准为致死性或非致死性心肌梗死。
诊断标准与关键数值
- 单变量分析中15个特征影像组学特征与心肌梗死相关,调整心血管风险评分和斑块负荷指标后8个特征保留关联。
- 临床模型(含心血管风险评分、Agatston评分及阻塞性冠心病)加入斑块负荷指标后C统计量0.70 vs 0.70,预测性能相似。
- 进一步加入特征影像组学特征后模型C统计量提升至0.74。
- 时间分析中,包含特征影像组学特征的模型在随访第5年后累积/动态AUC值更高。
临床应用与阅片要点
- 作者结论认为,基于影像组学的斑块形态精准表型分析可在临床因素和斑块负荷之上提升CT血管造影对长期心肌梗死的预测。
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: Coronary computed tomography (CT) angiography-derived attenuation-based plaque burden assessments can identify patients at risk of myocardial infarction.
OBJECTIVES: This study sought to assess whether more detailed plaque morphology assessment using patient-based radiomic characterization could further enhance the identification of patients at risk of myocardial infarction during long-term follow-up.
METHODS: Post hoc analysis of coronary CT angiography was performed within the SCOT-HEART (Scottish Computed Tomography of the HEART) clinical trial. Coronary plaque segmentations were used to calculate plaque burdens and eigen radiomic features that described plaque morphology. Univariable and multivariable Cox proportional hazard models were used to evaluate the association between clinical and image-based features and fatal or nonfatal myocardial infarction, whereas Harrell's C-statistic and cumulative/dynamic area under the curve (AUC) values with cross-validation were used to evaluate prognostic performance.
RESULTS: Scans from 1,750 patients (aged 58 ± 9 years; 56% male) were analyzed. Over a median of 8.6 years of follow-up, 82 patients had a fatal or nonfatal myocardial infarction. Among the eigen radiomic features, 15 were associated with myocardial infarction in univariable analysis, and 8 features retained their association following adjustment for cardiovascular risk score and plaque burden metrics. Adding plaque burden metrics to a clinical model incorporating cardiovascular risk score, Agatston score and presence of obstructive coronary artery disease had similar prediction performance (C-statistic 0.70 vs 0.70), whereas further addition of eigen radiomic features improved model performance (C-statistic 0.74). In temporal analysis, the model including eigen radiomic features had higher cumulative/dynamic AUC values following the fifth year of follow-up.
CONCLUSIONS: Radiomics-based precision phenotyping of coronary plaque morphology provided improvements to long-term prediction of myocardial infarction by CT angiography over and above clinical factors and plaque burden. (Scottish Computed Tomography of the HEART [SCOT-HEART]; NCT01149590).
引用
- Kolossváry M, Lin A, Kwiecinski J, et al. Coronary Plaque Radiomic Phenotypes Predict Fatal or Nonfatal Myocardial Infarction: Analysis of the SCOT-HEART Trial. JACC Cardiovasc Imaging. 2024. PMID: 39480364. doi:10.1016/j.jcmg.2024.08.012
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。