Towards Real-Time Heartbeat Classification: Evaluation of Nonlinear Morphological Features and Voting Method

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Abnormal heart rhythms pose a global health challenge.
• Current manual classification of heartbeats is time-consuming, error-prone, and can delay patient care.
• Real-time detection of rare abnormal heartbeats is difficult due to underrepresentation in datasets.

Purpose of the Study:

• To develop an automated system for accurate heartbeat classification, particularly for rare morphologies.
• To enable real-time detection and prompt intervention for cardiac conditions.
• To create an algorithm suitable for implementation on portable Field-Programmable Gate Array (FPGA) devices.

Main Methods:

• Utilized nonlinear morphological features for heartbeat classification.
• Implemented a voting scheme to enhance the recognition of rare heartbeat types.
• Tested the algorithm on the Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) database according to Association for the Advancement of Medical Instrumentation (AAMI) standards.

Main Results:

• The proposed automated method demonstrated superior performance, especially for minority classes.
• Achieved 90.4% accuracy for the fusion class and 100% accuracy for the unknown class.
• The algorithm shows significant potential for real-time application on portable devices.

Conclusions:

• The developed automated heartbeat classification system effectively addresses the challenge of rare morphologies.
• This method offers a promising advancement for real-time cardiac monitoring and diagnosis.
• The algorithm's design facilitates deployment on resource-constrained hardware like FPGAs for accessible healthcare solutions.