Early classification of pathological heartbeats on wireless body sensor nodes.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Smart Wireless Body Sensor Nodes (WBSNs) enable continuous bio-signal monitoring but face computational and bandwidth limitations.
• Efficient real-time analysis of electrocardiogram (ECG) signals is crucial for diagnosing heart conditions on resource-constrained devices.
• Identifying critical ECG segments for diagnosis can optimize computational load and energy consumption.

Purpose of the Study:

• To develop a framework for real-time automatic classification of normal and abnormal heartbeats for WBSNs.
• To compare dimensionality reduction strategies for ECG representation to minimize computational effort.
• To integrate these strategies with a neuro-fuzzy classifier for efficient, low-overhead analysis.

Main Methods:

• Comparative analysis of dimensionality reduction techniques for ECG heartbeat representation.
• Implementation of a neuro-fuzzy classification strategy for distinguishing normal and abnormal heartbeats.
• Evaluation of energy consumption by performing detailed analysis only on classifier-identified abnormal heartbeats.

Main Results:

• The neuro-fuzzy classifier effectively discerns normal and pathological heartbeats with minimal runtime and memory overhead.
• Significant energy savings achieved: up to 60% in signal processing and 63% in wireless transmission.
• The proposed framework demonstrates substantial energy reduction by selectively analyzing abnormal heartbeats.

Conclusions:

• The developed framework offers an energy-efficient solution for real-time ECG analysis on WBSNs.
• Combining dimensionality reduction with neuro-fuzzy classification is highly effective for embedded cardiac monitoring.
• This approach significantly reduces the overall energy footprint of WBSN systems for arrhythmia detection.