Assessing electrocardiogram changes after ischemic stroke with artificial intelligence.

Area of Science:

• Cardiology
• Neurology
• Artificial Intelligence

Background:

• Ischemic stroke (IS) poses significant health risks, often leading to cerebrocardiac syndrome (CCS) with poor outcomes.
• Electrocardiogram (ECG) monitoring is crucial for assessing cardiac health post-IS.

Purpose of the Study:

• To investigate electrocardiogram (ECG) changes following ischemic stroke (IS) using artificial intelligence (AI).
• To develop and evaluate AI models for detecting abnormal ECG patterns post-IS and assessing their prognostic value.

Main Methods:

• Collected and analyzed ECG data from healthy individuals and IS patients.
• Trained convolutional neural network (CNN), random forest (RF), and support vector machine (SVM) models to detect ECG changes.
• Utilized gradient class activation map (Grad-CAM) to identify abnormal ECG regions and compared CNN scores for prognostic assessment.

Main Results:

• A high percentage (78.84%) of IS patients exhibited abnormal ECGs.
• The CNN model demonstrated superior performance (AUC: 0.88) in distinguishing abnormal ECGs post-IS.
• CNN scores significantly correlated with patient prognosis (mRS scores) and Grad-CAM highlighted the V4 lead as a key area of abnormality.

Conclusions:

• Post-IS ECGs frequently show abnormal changes, underscoring the need for vigilant monitoring.
• The developed CNN model offers a robust tool for systematic assessment of ECG anomalies and prognostication in IS patients.