Real-Time Heart Arrhythmia Detection Using Apache Spark Structured Streaming

Area of Science:

• Biomedical Engineering
• Computer Science
• Cardiology

Background:

• Cardiac arrhythmias are a leading cause of mortality worldwide.
• Electrocardiogram (ECG) analysis is crucial for diagnosing arrhythmias, but real-time monitoring is needed due to intermittent symptoms.
• Timely detection of arrhythmias through continuous ECG analysis can prevent life-threatening incidents.

Purpose of the Study:

• To implement a machine learning pipeline for real-time cardiac arrhythmia detection using Apache Spark Structured Streaming.
• To evaluate the impact of Structured Streaming on classification performance and detection delay.
• To compare the performance of different machine learning classifiers for arrhythmia detection.

Main Methods:

• Utilized the Apache Spark Structured Streaming module for real-time ECG data analysis.
• Developed and compared three multiclass classifiers: decision trees, random forest, and logistic regression.
• Trained and evaluated models on ECG data from the MIT/BIH database, focusing on normal beats, RBBB, and atrial fibrillation.

Main Results:

• The random forest classifier demonstrated superior performance compared to decision trees and logistic regression.
• The implemented pipeline achieved significant reductions in runtime, especially with an increased number of class labels.
• The study validated previous findings in classification model performance metrics.

Conclusions:

• Apache Spark Structured Streaming offers a viable and efficient platform for real-time cardiac arrhythmia detection.
• The random forest classifier is effective for classifying arrhythmias in real-time ECG data.
• This approach enhances timely arrhythmia detection, potentially improving patient outcomes and reducing healthcare burdens.