Arrhythmia Classification via Wavelet and EMD Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for classifying arrhythmia types from electrocardiogram (ECG) signals are inefficient in accurately and quickly distinguishing between Ventricular Fibrillation (VF), Ventricular Tachycardia (VT), Ventricular Flutter (VFL), and Supraventricular tachycardia (SVT), especially when ECG waves become deformed, leading to delayed detection which can be fatal.
Innovation Solution
A system and method that utilize a processor to obtain an electrocardiogram signal, extract feature vectors including threshold crossing sample percent, L-Z complexity, empirical mode decomposition complex number, and wavelet transform energy features, and train a prediction model using classifier models such as random forest, BP neural network, or support vector machine to determine the arrhythmia type based on these features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods extract ECG characteristics (P wave, QRS complex, T wave, RR period, ST segment) to classify arrhythmia types, then the classification process can be performed, but the accuracy deteriorates when ECG waves become deformed and detection speed is insufficient
Solution Approach 1:
The patent transforms the ECG signal from time-domain characteristics (P wave, QRS complex, T wave morphology) to frequency-domain features through wavelet transform and empirical mode decomposition. This parameter transformation enables accurate classification even when traditional time-domain waveforms are deformed, as the frequency-energy distribution remains distinctive for different arrhythmia types
Solution Approach 2:
The patent replaces manual/visual inspection of ECG waveforms with automated signal processing algorithms including wavelet transform, empirical mode decomposition, and machine learning classifiers. This substitution dramatically improves detection speed and eliminates human subjectivity, achieving both rapid processing and high accuracy even with deformed waves
2Reliability
If traditional ECG characteristic extraction methods are used, then the system structure remains simple, but the reliability of arrhythmia type determination deteriorates when waves are deformed
Solution Approach 1:
The patent segments the complex ECG signal analysis into multiple independent feature extraction components: wavelet transform for time-frequency analysis, empirical mode decomposition for intrinsic mode functions, and various entropy calculations. Each segment processes specific aspects of the signal, and their combined results provide robust classification even when individual waveform characteristics are deformed
Solution Approach 2:
The patent combines multiple different feature types (wavelet energy features, empirical mode decomposition features, entropy features, statistical features) to create a composite feature vector for classification. This composite approach ensures that if one feature set is affected by waveform deformation, other feature sets can compensate, maintaining high reliability
3Productivity
If multiple ECG characteristics are extracted and analyzed manually, then comprehensive analysis is achieved, but the productivity and detection speed deteriorate
Solution Approach 1:
The patent replaces manual analysis of multiple ECG characteristics with automated computational algorithms. The wavelet transform, empirical mode decomposition, and machine learning classifiers process all features simultaneously and objectively, achieving high-speed automated detection without the time consumption and subjectivity of manual waveform analysis
Solution Approach 2:
The patent performs preliminary signal preprocessing including noise filtering and baseline correction before feature extraction. This preliminary action ensures that subsequent automated analysis operates on clean, standardized data, improving both processing speed and classification accuracy while reducing the computational burden of handling raw, noisy signals
Data Source
AI summary
Systems and methods for determining an arrhythmia type are provided. A method may include: obtaining an electrocardiogram signal, obtaining a feature vector of the electrocardiogram signal, obtaining a trained prediction model, and determining an arrhythmia type of the electrocardiogram signal based on the trained prediction model and the feature vector. The feature vector includes a threshold crossing sample (TCSC) percent feature, an L-Z complexity feature, an empirical mode decomposition (EMD) complex number feature, a sample entropy feature, and a set of wavelet transform energy features.