Stability analysis and recognition of paroxysmal atrial fibrillation signals
10.3969/j.issn.1005-202X.2025.09.014
- VernacularTitle:阵发性房颤信号的稳定性分析与识别
- Author:
Song LIU
1
;
Donghui LIU
;
Qinghua MENG
;
Dehua HE
Author Information
1. 南宁师范大学广西信息功能材料与智能信息处理重点实验室,广西 南宁 530001;南宁师范大学物理与电子学院,广西 南宁 530001
- Publication Type:Journal Article
- Keywords:
paroxysmal atrial fibrillation detection;
stability analysis;
dynamic mode;
feature extraction
- From:
Chinese Journal of Medical Physics
2025;42(9):1221-1228
- CountryChina
- Language:Chinese
-
Abstract:
The clinical detection of paroxysmal atrial fibrillation(PAF)remains challenging due to its transient and stochastic characteristics,and existing dynamic mode decomposition methods have limitations in modal redundancy reduction and feature extraction when processing single-channel noisy electrocardiogram(ECG)signals.Therefore,a signal analysis method based on high-order dynamic mode decomposition is proposed.It captures high-order correlations within ECG signals through tensor decomposition techniques and decomposes complex signals into physically interpretable dynamic modes.A stability evaluation framework for signal subsystem is established based on modal interaction relationships.By incorporating quantitative indicators including proportion of modes reflecting system instability,modal distribution entropy,and eigenvalue spectrum divergence,a feature discrimination model for PAF is developed.Experimental validation using the MIT-BIH atrial fibrillation database reveals statistically significant differences(P<0.05)in stability-related features between PAF episodes and normal sinus rhythms.The classification model based on support vector machine achieves an average recognition accuracy of 96.15%.These results demonstrate that the proposed method can effectively analyze nonlinear dynamic characteristics in noisy single-lead ECG signals,thereby establishing a novel quantitative analytical framework for early detection and accurate diagnosis of PAF.
