基于冠脉CTA的多模态特征(解剖学与血流动力学)与阻塞性冠心病的相关性及诊断预测价值研究

Correlation and Diagnostic Predictive Value of Multimodal Features (Anatomical and Hemodynamic) Derived from Coronary CTA in Obstructive Coronary Artery Disease

摘要: 目的：探讨基于冠状动脉CT血管造影(coronary computed tomography angiography, CCTA)提取的高级多模态衍生指标(血流动力学、局部炎症及解剖形态)，对传统解剖学定义的中重度阻塞性冠心病(狭窄 ≥ 50%)的综合评估与识别价值。方法：回顾性纳入疑似冠状动脉粥样硬化性心脏病(coronary atherosclerotic heart disease, CAD)并接受CCTA检查的患者，依据冠脉任何一支主支狭窄是否≥ 50%分为阻塞性CAD组与非阻塞性CAD组。利用后处理工作站提取目标病变的心外膜脂肪组织(EAT)形态学指标、血管周围脂肪衰减指数(FAI)、冠脉钙化总积分(总CACS)、斑块负荷(Plaque burden)以及无创血流储备分数(CT-FFR)。采用独立样本t检验、Mann-Whitney U检验和卡方检验比较两组临床与影像基线差异，构建向后逐步多因素Logistic回归模型分析独立关联因子，并通过受试者工作特征(ROC)曲线比较不同联合模型的诊断效能。结果：最终纳入有效样本77例(非阻塞性CAD组16例，阻塞性CAD组61例)。结果显示，两组在性别比例、血压、心率、EAT收缩期体积(收缩期EAT体积)、FAI平均值(平均FAI)及对数斑块体积(对数斑块体积)上均无显著统计学差异(均P > 0.05)。然而，阻塞性组的平均年龄显著偏高(65.75 ± 10.37岁)，且高血压、糖尿病患病率及基础服用他汀(93.4%)和阿司匹林(83.6%)的比例极显著升高。在影像学指标上，两组在总CACS、CT-FFR及斑块负荷上存在极显著差异(均P ≤ 0.005)。多因素向后逐步Logistic回归分析表明，CT-FFR是识别阻塞性CAD的强独立关联因素(OR = 0.187, P = 0.007)。ROC曲线分析显示，单纯包含局部脂肪、早期形态及功能学指标(收缩期EAT体积、平均FAI、对数斑块体积、CT-FFR)的基础模型A AUC为0.807；在此基础上引入常规斑块负荷的模型C未见显著诊断增益(AUC = 0.804)；而引入总CACS构建的联合模型B诊断效能达到最优，AUC跃升至0.891 (95% CI: 0.822~0.970, P < 0.001)。结论：在评估中晚期阻塞性冠心病时，单独的局部脂肪形态与炎症指标(EAT/FAI)诊断价值有限。将反映血流动力学损害的FFR与反映解剖学硬化终态的CACS相联合，实现了多维度的正交互补，可显著提升阻塞性CAD的综合识别效能，为临床无创危险分层提供了更优的影像学策略。

Abstract: Objective: To evaluate the comprehensive diagnostic and identification value of advanced multimodal derived parameters (hemodynamics, local inflammation, and anatomical morphology) extracted from coronary computed tomography angiography (CCTA) for anatomically defined moderate-to-severe obstructive coronary artery disease (stenosis ≥ 50%). Methods: Patients with suspected coronary atherosclerotic heart disease (CAD) who underwent CCTA were retrospectively enrolled. Based on whether the stenosis of any major coronary branch was ≥ 50%, patients were divided into an obstructive CAD group and a non-obstructive CAD group. Morphological metrics of epicardial adipose tissue (EAT), perivascular fat attenuation index (FAI), total coronary artery calcium score (CACS), plaque burden, and non-invasive fractional flow reserve (CT-FFR) of the target lesions were extracted. Independent associated factors were analyzed using a backward stepwise multivariate logistic regression model, and the diagnostic efficacies of different combined models were compared using receiver operating characteristic (ROC) curves. Results: A total of 77 valid samples were included (16 in the non-obstructive CAD group and 61 in the obstructive CAD group). There were no significant statistical differences in sex, blood pressure, heart rate, EAT systolic volume, mean FAI, and log-transformed plaque volume between the two groups (all P > 0.05). However, the obstructive group showed a significantly older mean age, higher prevalence of hypertension and diabetes, and significantly higher proportions of baseline statin (93.4%) and aspirin (83.6%) usage. Radiologically, highly significant differences were observed in total CACS, CT-FFR, and plaque burden between the two groups (all P ≤ 0.005). Multivariate logistic regression indicated that CT-FFR was a strong independent associated factor for identifying obstructive CAD (OR = 0.187, P = 0.007). ROC curve analysis showed that the basic Model A (incorporating EAT volume, mean FAI, log-plaque volume, and CT-FFR) yielded an AUC of 0.807. The introduction of conventional plaque burden (Model C) provided no significant diagnostic increment (AUC = 0.804). In contrast, the combined Model B, which incorporated total CACS into Model A, demonstrated the optimal diagnostic efficacy, with the AUC significantly increasing to 0.891 (95% CI: 0.822~0.970, P < 0.001). Conclusion: In the assessment of moderate-to-advanced obstructive CAD, the diagnostic value of isolated local fat morphology and inflammation indices (EAT/FAI) is limited. The combination of CT-FFR (reflecting hemodynamic impairment) and CACS (reflecting the anatomical end-stage of sclerosis) achieves multi-dimensional orthogonal complementarity. This combination significantly enhances the comprehensive identification efficacy of obstructive CAD, providing an optimal non-invasive imaging strategy for clinical risk stratification.