.png){kind=icon}
.png){kind=icon}
Return
A hybrid framework for ECG derived breathing pattern classification using signal decomposition method and AI algorithms
.png){kind=icon}
PRE
AI
S
P
D
S
V
DOI：10.1016/j.bspc.2026.111162.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Sleep-disordered breathing has a great impact on overall health, raising the risk of metabolic and cardiovascular disorders. For ongoing home-based monitoring, traditional respiratory monitoring techniques like spirometry and plethysmography though non-invasive, can cause discomfort to the patient and cannot be used for long term observation. This study investigates Electrocardiogram-Derived Respiration (EDR) as a non-invasive and economical respiratory signal extraction with the integration of wearable sensor data processing using signal decomposition and artificial intelligence (AI) algorithms. This is attained by the extraction of electrocardiogram (ECG) derived respiratory signals using Variational Mode Decomposition (VMD). The optimal number of modes are selected using the elbow method. The features extracted from the optimal modes are used to categorize the breathing patterns into three groups: normal breathing, deep breathing, and breath-holding using Random Forest (RF) and 1-D convolutional neural network (1-D CNN). The proposed framework is validated on a cardio-respiratory database containing 41 subjects. The high F1-score of 0.97 obtained by both ML and DL indicates that the proposed hybrid framework integrating signal decomposition as input to AI models on EDR classification outperformed the existing techniques. The outcome of this work has the scope to aid the development of affordable and computationally efficient respiratory monitoring frameworks, with potential applicability to wearable deployment pipelines, subject to further validation on real-world wearable-acquired data under uncontrolled environmental conditions. In addition, this study aligns with SDG 3 (Good health and well-being) by enabling AI driven assessment systems facilitating timely detection of respiratory anomalies using breathing patterns.
Journal
IF:
4.9
Papers:
9.7K
Citations:
2.4W
Organization
Cited Papers