Multirate Processing with Selective Subbands and Machine Learning for Efficient Arrhythmia Classification.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Wearable gadgets are increasingly used in cloud-based health monitoring.
• Efficient signal compression, computation, and power usage are critical for these systems.
• Cardiovascular disease diagnosis relies heavily on electrocardiogram (ECG) signal analysis.

Purpose of the Study:

• To propose an efficient method for cardiovascular disease diagnosis using ECG signals.
• To enhance signal compression, computational efficiency, and reduce power consumption in wearable health monitoring.
• To improve the accuracy and feasibility of detecting cardiac arrhythmias.

Main Methods:

• A novel method combining multirate processing, wavelet decomposition, and frequency content-based subband coefficient selection.
• Machine learning techniques for classification of ECG signals.
• Utilized the MIT-BIH dataset with five-fold cross-validation (5CV) and a partial blind protocol to assess classifier performance and avoid bias.

Main Results:

• Achieved over a 12-fold computational gain and a 13-fold compression gain compared to fixed-rate methods.
• Ensured appropriate signal reconstruction quality.
• Attained high classification accuracies of 97.06% for 5CV and 92.08% for the partial blind protocol.

Conclusions:

• The proposed method is feasible for detecting cardiac arrhythmias using wearable ECG monitoring.
• The approach significantly reduces computational complexity and data transmission requirements.
• Demonstrates the potential for efficient and accurate cloud-based cardiovascular health monitoring.