Right Precordial U-Waves in Severe Aortic Stenosis: "Real World" Prevalence and Diminution by TAVR.
Right Precordial U-Waves in Severe Aortic Stenosis: "Real World" Prevalence and Diminution by TAVR.
复制标题
严重主动脉瓣狭窄中的右心前区 U 波：TAVR 的“真实世界”患病率和减少。
DOI：
10.1016/j.jacc.2022.04.012
复制
发表时间：
2022
影响因子：
24
通讯作者：
Vivek Y. Reddy
中科院分区：
文献类型：
--
作者：
J. Lampert;Matthew Pulaski;M. Miller;W. Whang;J. Koruth;Benjamin S. Glicksberg;Samin K. Sharma;Srinivas R. Dukkipati;V. Fuster;Vivek Y. Reddy
使用卷积神经网络（人工智能算法中常见的深度学习模型），机器学习发现可以仅根据12导联心电图（ECG）识别中度或重度主动脉瓣狭窄（AS）患者。显着性映射是一种技术，它阐明了卷积神经网络在模型预测中使用的特征重要性，从而向人类观察者强调了用于通知模型预测的算法的具体特征。显著性标测将12导联ECG上的右胸U波（U波）确定为识别AS患者的关键特征。3，4由于关于将机器学习识别的ECG特征转化为临床实践的数据很少，我们试图：1）评估严重AS中U波的真实存在; 2）确定经导管主动脉瓣置换术（TAVR）是否对动态U波变化有任何影响。2021年8月至12月期间，共有168名在西奈山医院接受TAVR的连续患者在获得机构审查委员会批准后接受了筛选。如果患者超声心动图显示为重度AS（主动脉瓣面积< 1 cm 2，平均跨瓣压差40 mm Hg，连续波多普勒峰值速度4 m/s），则纳入研究。排除低血流、低梯度AS患者。从电子病历中提取临床、实验室、心电图和超声心动图数据，直至达到50例患者纳入分析。U波患病率也进行了评估，在50名患者的年龄，性别和高血压病史相匹配的内科服务。U波被定义为导联V1至V3中T波之后的正向偏转，在一个不同的TP段内，不是由于伪影。U波减小定义为TAVR前与TAVR后获得的ECG（TAVR后首次ECG或TAVR入院后最后一次可用ECG）之间可见U波振幅的降低。分别使用双尾配对样本Student t检验或卡方检验分析连续变量和分类变量之间的差异，62%（31/50）的重度AS患者在接受TAVR前存在U波（图1A）。在这31例TAVR前U波患者中，87%的患者在TAVR后ECG上显示U波减小（图1B），如下所示：1）7/31例患者（22.5%）的U波振幅减小（但仍可测量）; 2）20/31例患者（64.5%）的U波在瓣膜介入后完全消退。TAVR前平均U波振幅为0.48 mm（95% CI：0.33-0.62 mm），TAVR后为0.12 mm（95% CI：0.04-0.21 mm; P< 0.01）。在19例TAVR前无U波的患者中，TAVR后无U波。基线临床变量（性别、年龄、慢性肾脏疾病、高血压或体重指数）、血清电解质浓度（钠、钾、氯、钙或镁）或围手术期超声心动图室壁厚度无显著差异。尽管TAVR前后的平均收缩压无显著差异，但TAVR后的平均舒张压有统计学显著性降低（70 mm Hg [95% CI：66-75 mm Hg] vs 65 mm Hg [95% CI：62-68 mm Hg]; P= 0.03）。平均主动脉瓣压差从53 mm Hg（95% CI：49-57 mm Hg）降至12 mm Hg（95% CI：10-13 mm Hg; P< 0.01）。至TAVR后ECG的中位时间为2天（IQR：1-4天）。在对照非TAVR人群中，U波的发生率明显较低（50例患者中的3例[6%]; P< 0.001）。3例患者中，U…
Using a convolutional neural network, a common deep learning model in artificial intelligence algorithms, machine learning found that patients with moderate or severe aortic stenosis (AS) can be identified based solely on the 12-lead electrocardiogram (ECG). 1, 2 Saliency mapping is a technique that elucidates feature importance used by convolutional neural networks in model prediction, thereby underscoring to the human observer which specific features the algorithm used to inform model prediction. Saliency mapping identified right precordial U waves (U waves) on the 12-lead ECG as a key feature to identify patients with AS. 3, 4 Because data are scarce regarding the translation of machine learning–identified ECG features into clinical practice, we sought to: 1) assess the real-world presence of U waves in severe AS; and 2) determine whether there is any effect of transcatheter aortic valve replacement (TAVR) on dynamic U-wave changes. Between August and December 2021, a total of 168 consecutive patients undergoing TAVR at Mount Sinai Hospital were screened after approval by the Institutional Review Board. Patients were included if they had echocardiographically severe AS (aortic valve area< 1 cm2, mean gradient $40 mm Hg, and peak continuous wave Doppler velocity $4 m/s). Patients with low-flow, low-gradient AS were excluded. Clinical, laboratory, electrocardiographic, and echocardiographic data were extracted from the electronic medical record until 50 patients were reached for inclusion in the analysis. U-wave prevalence was also assessed in 50 patients matched to age, gender, and hypertension history on an internal medicine service. U waves were defined as a positive deflection in leads V1 to V3 after the T-wave within a distinct TP segment not due to artifact. U-wave diminution was defined as a reduction in amplitude of visible U waves between the ECGs obtained pre-TAVR vs post-TAVR (first ECG post-TAVR or last available ECG from TAVR admission). Differences between continuous and categorical variables were analyzed by using a 2-tailed paired sample Student’s t-test or chi-square test, respectively.U waves were present in 62%(31 of 50) of patients with severe AS before undergoing TAVR (Figure 1A). Among these 31 patients with pre-TAVR U waves, 87% exhibited U-wave diminution (Figure 1B) on the post-TAVR ECG, as follows: 1) in 7 (22.5%) of 31 patients, the U-wave amplitude diminished (but was still measurable); and 2) in 20 (64.5%) of 31 patients, the U wave resolved completely after valve intervention. The mean U-wave amplitude was 0.48 mm before TAVR (95% CI: 0.33-0.62 mm) and 0.12 mm (95% CI: 0.04-0.21 mm; P< 0.01) after TAVR. Of 19 patients without U waves pre-TAVR, none developed U waves after TAVR. There were no significant differences in baseline clinical variables (sex, age, chronic kidney disease, hypertension, or body mass index), serum electrolyte concentrations (sodium, potassium, chloride, calcium, or magnesium), or echocardiographic wall thickness perioperatively. Although there was no significant difference between mean systolic blood pressure before and after TAVR, there was a statistically significant decrease in mean diastolic blood pressure after TAVR (70 mm Hg [95% CI: 66-75 mm Hg] vs 65 mm Hg [95% CI: 62-68 mm Hg]; P= 0.03). The mean aortic valve gradient decreased from 53 mm Hg (95% CI: 49-57 mm Hg) to 12 mm Hg (95% CI: 10-13 mm Hg; P< 0.01). The median time to the post-TAVR ECG was 2 days (IQR: 1-4 days). U waves were significantly less prevalent (3 of 50 patients [6%]; P< 0.001) in a control non-TAVR population. Of the 3 patients with U …