Energy Entropy Feature Selection for Cardiac Arrhythmia Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cardiac arrhythmia detection methods are prone to measurement errors due to noisy or low-amplitude sensor signals, leading to ambiguity and inaccuracies in diagnosing the type of arrhythmia, particularly in conditions like atrial fibrillation.
Innovation Solution
A method and system that convert biological signals into energy entropy signals, extract specific statistical features, and selectively combine them for more accurate arrhythmia detection, using preprocessing, feature extraction, and feature selection techniques to improve classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional ECG or halter monitor methods are used for arrhythmia detection, then the detection process is simple and straightforward, but the signals are prone to misinterpretation due to ambiguity and lack of clarity
Solution Approach 1:
The patent segments the complex signal processing task into distinct stages: raw signal acquisition, noise filtering, feature extraction (including time-domain, frequency-domain, and time-frequency domain features), and classification. This segmentation allows each stage to be optimized independently, improving overall measurement precision while managing complexity through structured processing pipelines
Solution Approach 2:
The patent transitions from analyzing signals in a single time domain to multiple dimensions including frequency domain (FFT, wavelet transform) and time-frequency domain (spectrograms, short-time Fourier transform). This dimensional expansion provides richer signal characteristics for arrhythmia classification, resolving the contradiction by adding analytical depth without requiring fundamentally new device architecture
2Measurement precision
If IBI is measured between R-wave peaks to reduce measurement error, then measurement precision improves, but the initiation of QRS complex becomes difficult to locate in noisy or low-amplitude sensor signals
Solution Approach 1:
The patent applies preliminary noise filtering and signal conditioning operations before attempting to detect QRS complexes or R-wave peaks. By pre-processing the signal to enhance signal-to-noise ratio through filtering and amplitude normalization, the subsequent peak detection becomes more reliable, resolving the contradiction between measurement precision and detection difficulty
Solution Approach 2:
The patent introduces intermediary processing steps including signal filtering, envelope detection, and feature extraction as mediators between the raw sensor signal and the final IBI measurement. These intermediary operations transform the noisy raw signal into a cleaner representation where R-wave peaks can be reliably identified, solving the detection difficulty while maintaining measurement precision
3Measurement precision
If statistical features are extracted from energy entropy signal for arrhythmia classification, then classification accuracy improves, but the processing time and computational requirements increase
Solution Approach 1:
The patent extracts only the most discriminative statistical features from the energy entropy signal and other signal representations, rather than processing all possible features. By selecting a subset of key features (such as mean, standard deviation, skewness, kurtosis of specific signal components), the classification accuracy is maintained while significantly reducing computational burden and processing time
Solution Approach 2:
The patent applies partial action by focusing computational resources on the most informative time segments and frequency bands of the signal, rather than uniformly processing the entire signal duration. This selective processing approach extracts critical arrhythmia indicators without the need to analyze every aspect of the signal, thereby reducing overall processing time while maintaining high classification accuracy
Data Source
AI summary
A method of measuring biological signals of a person including obtaining physiological signals; converting the physiological signals to an energy entropy signal, determining one or more statistical features based on the energy entropy signal, and selecting and combining two or more of the features. The features include two or more of: Shannon entropy of power spectral density (PSD) of energy entropy; minimum envelopes of three axis signals; minimum Shannon entropy of PSD of three axis entropies; standard deviation of interbeat intervals (IBIs); differences of median filtered IBIs; determinism of recurrence plot of energy entropy; Kurtosis of energy entropy; minimum peak-to-peak amplitude three axis signals; peak-to-peak amplitude of energy entropy; laminarity of recurrence plot of energy entropy; Shannon entropy of horizontal structures in recurrence plot of energy entropy; standard deviation of time-averaged spectrogram; minimum standard deviation of peak positions in spectrogram; and minimum standard deviation of peak widths in spectrogram.