【摘要】 目的 探讨定量PET心肌灌注-代谢不匹配（MIS）能否进一步提升阻塞性冠心病（CAD）诊断的准确性。方法 回顾性收集阻塞性CAD疑似患者97例，将其按照5∶3的比例随机划分为训练集（n=61）和测试集（n=36），并将阻塞性CAD定义为冠脉血管狭窄≥75%。单变量和多变量的Logistic回归分析用于分类模型的训练和性能评估。ROC曲线用于确定诊断阻塞性CAD的最佳截断值以及比较模型间的差异性。结果 单变量Logistic回归模型的AUC值分别为0.735、0.758和0.823,MIS的最佳截断值为9%；多变量Logistic回归模型的AUC值分别为0.766、0.823和0.839，以包含MIS指标的Model＿3最高，且高于对照模型（P=0.0034）；基于Model＿3构建的决策树，在测试集中的准确度最高可达96%。结论 PET MIS是阻塞性CAD的有效预测因子，能有效提升诊断准确性，对疑似病例快速、准确的鉴别诊断具有重要的临床价值。更多还原

【Abstract】 Objective To investigate whether quantitative PET myocardial perfusion-metabolism mismatch can further improve the diagnosis accuracy in predicting obstructive coronary artery disease(CAD). Methods A total of 97 patients with suspected obstructive CAD were retrospectively collected, and randomly divided into training set(61 patients) and validation set(36patients) according to 5∶3. Obstructive CAD was defined as coronary artery≥75% stenosis. Univariate and multivariate logistic regression analysis were used for model training and performance evaluation. The ROC curve analysis was used to determine the best trade-off for diagnosing obstructive CAD and compare the differences between models. Results The areas under the ROC curve(AUC) of three univariate logistic regression models were 0.735, 0.758 and 0.823, respectively, and the best trade-off of MIS was 9%. The AUC values of three multivariate logistic regression models were 0.766, 0.823 and 0.839, respectively. The model including TPD/MBF/MIS achieved the highest AUC and it was significantly higher than the control model including TPD/MBF(P=0.0034). The decision tree based on this model had an accuracy of up to 96% in the validation set. Conclusion PET myocardial perfusion-metabolism mismatch is an effective predictor of obstructive CAD, which can effectively improve the diagnostic accuracy and has important clinical value for the rapid and accurate differential diagnosis of suspected cases.更多还原