Biosignals learning and synthesis using deep neural networks.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Signal Processing

Background:

• Modeling and synthesizing biomedical signals is challenging.
• Deep neural networks offer a promising approach for biosignal analysis.
• Accurate synthesis can aid in understanding complex physiological data.

Purpose of the Study:

• To develop and validate a deep neural network for biosignal synthesis.
• To assess the model's ability to learn and replicate morphological characteristics of physiological signals.
• To explore potential applications in signal reconstruction and source detection.

Main Methods:

• Utilized Gated Recurrent Units (GRU) for training respiration (RESP), electromyograms (EMG), and electrocardiograms (ECG).
• Pre-processed, segmented, and quantized signals before feeding them into the GRU-based network.
• Trained the network to predict signal values based on previous data points for synthesis.

Main Results:

• Generated signals exhibited morphological equivalence to the original physiological signals.
• The model successfully learned basic and cyclic characteristics of the signals.
• Synthesized respiration (RESP) and electrocardiogram (ECG) signals closely matched the training data.

Conclusions:

• The proposed deep learning model effectively synthesizes physiological signals.
• This biosignal synthesis technique demonstrates potential for applications in noisy data reconstruction and source identification.
• The approach shows promise for characterizing signals from various physiological sources.