In Search of an Optimal Subset of ECG Features to Augment the Diagnosis of Acute Coronary Syndrome at the Emergency Department

Abstract

Background Classical ST‐T waveform changes on standard 12‐lead ECG have limited sensitivity in detecting acute coronary syndrome (ACS) in the emergency department. Numerous novel ECG features have been previously proposed to augment clinicians' decision during patient evaluation, yet their clinical utility remains unclear. Methods and Results This was an observational study of consecutive patients evaluated for suspected ACS (Cohort 1 n=745, age 59±17, 42% female, 15% ACS; Cohort 2 n=499, age 59±16, 49% female, 18% ACS). Out of 554 temporal‐spatial ECG waveform features, we used domain knowledge to select a subset of 65 physiology‐driven features that are mechanistically linked to myocardial ischemia and compared their performance to a subset of 229 data‐driven features selected by multiple machine learning algorithms. We then used random forest to select a final subset of 73 most important ECG features that had both data‐ and physiology‐driven basis to ACS prediction and compared their performance to clinical experts. On testing set, a regularized logistic regression classifier based on the 73 hybrid features yielded a stable model that outperformed clinical experts in predicting ACS, with 10% to 29% of cases reclassified correctly. Metrics of nondipolar electrical dispersion (ie, circumferential ischemia), ventricular activation time (ie, transmural conduction delays), QRS and T axes and angles (ie, global remodeling), and principal component analysis ratio of ECG waveforms (ie, regional heterogeneity) played an important role in the improved reclassification performance. Conclusions We identified a subset of novel ECG features predictive of ACS with a fully interpretable model highly adaptable to clinical decision support applications. Registration URL: https://www.clinicaltrials.gov ; Unique Identifier: NCT04237688.

Machine Learning for the ECG Diagnosis and Risk Stratification of Occlusion Myocardial Infarction at First Medical Contact

“…ECG featurization was described in detail elsewhere. 19 Brie y, ECG signal preprocessing and feature extraction were performed using a manufacturer-speci c software (Philips DXL diagnostic 12/16 lead ECG analysis program). ECG signals were rst preprocessed to remove noise, artifacts, and baseline wander.…”

“…Next, we identi ed an optimal parsimonious list of the most important ECG features that are mechanistically linked to cardiac ischemia as described in detail elsewhere. 19 Brie y, to prevent omittedfeature bias, we used a hybrid approach that combines domain knowledge with a data-driven strategy.…”

“…We selected 73 morphological ECG features out of 554 temporal-spatial metrics using a hybrid datadriven and domain expertise approach. 19 Using these features, ten classi ers were trained to learn ischemic patterns between ACS and non-ACS groups and to estimate the probability of OMI: regularized logistic regression, linear discriminant analysis, support vector machine, Gaussian Naïve Bayes, random forest, gradient boosting machine, extreme gradient boosting, stochastic gradient descent logistic regression, k-nearest neighbors, and arti cial neural networks. We chose these classi ers because they learn different mathematical representations in the data, maximizing the chance of nding the best modeling approach for relating complex ECG data to underlying physiology.…”

“…[15][16][17][18] This excess risk can be mitigated with enhanced diagnostic criteria. Although important ECG signatures of OMI are frequently described in the literature, [19][20][21][22] they are subtle, involve the entire QRST complex, and are spatial in nature (i.e., changes diluted across multiple leads). [23][24][25][26] Visual inspection of ECG images by clinical experts, thus, is suboptimal and leads to a high degree of variability in ECG interpretation.…”

Different Routes or Methods of Application for Dimensionality Reduction in Multicenter Studies Databases

“…To further optimize the number of components, a third verification (Table 3) is performed with random forest, [33] by Gini index and information value by weight of evidence (WOE) [34,35] to ascertain whether it is possible to effectively reduce dimensional space by calculating the importance of variables, based on the data-variables explored. Based on the outputs, the conclusion coincided with that of the PCA analysis.…”