机器学习法识别STEMI患者经皮冠状动脉介入治疗后微循环障碍的影像学相关危险因素  

Machine Learning to Identify Imaging Related Risk Factors for Microvascular Dysfunction after Percutaneous Coronary Intervention in Patients with STEMI

作　　者：王浩然 张上鸿 刘洁良 黄明亮 杨峻青[1] 李光[1] 胡天宇 董豪坚[1] WANG Haoran;ZHANG Shanghong;LIU Jieliang;HUANG Mingliang;YANG Junqing;LI Guang;HU Tianyu;DONG Haojian(Guangdong Cardiovascular Institute,Guangdong Provincial People's Hospital(Guangdong Academy of Medical),Guangzhou 510080,China;Department of Cardilogy,Linzhi People's Hospital of Xizang,Linzhi,Xizang 860000,China)

机构地区：[1]广东省心血管病研究所广东省人民医院(广东省医学科学院),广州510080 [2]西藏林芝市人民医院心内科,西藏林芝860000

出　　处：《岭南心血管病杂志》2024年第5期486-495,共10页South China Journal of Cardiovascular Diseases

基　　金：西藏自治区科技计划项目(项目编号:XZ202201ZY0051G);广东省基础与应用基础研究基金项目(项目编号:2024A1515012943);急诊再灌注治疗策略优化研究(项目编号:2016YFC1301202)。

摘　　要：目的探究急性ST段抬高型心肌梗死(ST-segment elevation myocardial infarction,STEMI)患者在接受急诊经皮冠状动脉介入(percutaneous coronary intervention,PCI)治疗后发生冠状动脉微循环障碍(coronary microvascular dysfunction,CMD)的影像学相关危险因素。方法回顾性纳入广东省人民医院2016年至2022年期间STEMI发病24 h内接受急诊PCI治疗的患者的临床资料。以血管造影衍生的微循环阻力诊断CMD,依据纳入的影像学指标,采用k-means聚类分析法计算最佳聚类数,通过Logistic回归及主成分分析法确定不同影像学指标对CMD的贡献度大小。结果共纳入1045例患者,880例男性,占84.2%;年龄为(60.9±12.6)岁,其中506例患者(48.4%)发生了微循环障碍。基于冠状动脉影像学指标的聚类分析识别了两个不同的心肌梗死患者群体,两个群体术后患CMD的风险差异具有统计学意义(P<0.001)。Logistic回归及主成分分析结果显示,术前指标按贡献百分比排序为:心肌梗死溶栓试验(thrombolysis in myocardial infarction,TIMI)血流分级、心肌呈色分级(myocardial blush grade,MBG)、“罪犯”血管闭塞、“罪犯”血管定位,均与CMD患病风险显著相关(P<0.05)。结论机器学习法发现,术前评估TIMI和MBG血流分级、“罪犯”血管闭塞有助于识别急诊PCI治疗后CMD的高风险患者。Objectives To explore the imaging-related risk factors of coronary microvascular dysfunction(CMD)in patients with acute ST-segment elevation myocardial infarction(STEMI)after emergency percutaneous coronary interven‐tion(PCI).Methods A retrospective study was conducted on STEMI patients who received emergency PCI treatment within 24 hours of onset at Guangdong Provincial People's Hospital from 2016 to 2022.CMD was diagnosed using angiography-derived microcirculatory resistance.K-means clustering analysis was employed to calculate the optimal number of clusters based on included imaging indicators,and Logistic regression and principal component analysis were used to determine the contribution of different imaging indicators to CMD.Results A total of 1045 patients[880 males,accounting for 84.2%;average age(60.9±12.6)years]were included,among which 506 patients(48.4%)developed microvascular dysfunction.Clustering analysis based on coronary imaging indicators identified two distinct groups of myo‐cardial infarction patients,and the risk difference of CMD between the two groups was statistically significant(P<0.001).Logistic regression and principal component analysis results showed that the preoperative indicators ranked by contribution percentage were thrombolysis in myocardial infarction(TIMI)grade,myocardial blush grade(MBG)grade,culprit vessel occlusion,and culprit vessel location,all significantly associated with the risk of CMD(P<0.05).Conclusions Machine learning methods reveal that preoperative assessment of TIMI and MBG blood flow grading and culprit vessel occlusion helps identify high-risk patients for CMD after emergency PCI.

关 键 词：心肌梗死 经皮冠状动脉介入治疗 冠状动脉微循环障碍 机器学习法 

分 类 号：R542.22[医药卫生—心血管疾病]