Deep learning-based ballistocardiography reconstruction algorithm on the optical fiber sensor.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Optical Sensing

Background:

• Ballistocardiography (BCG) non-invasively measures cardiac activity via body vibrations.
• Optical fiber sensors, particularly Mach-Zehnder interferometers (MZI), are explored for BCG detection.
• Signal fading in MZIs poses a challenge for accurate BCG acquisition.

Purpose of the Study:

• To develop a novel algorithm for reconstructing BCG signals from MZI.
• To address and solve the signal fading issue in optical fiber-based BCG.
• To simplify the hardware complexity and reduce the cost of BCG monitoring systems.

Main Methods:

• A modified generative adversarial network (GAN) was proposed for BCG signal reconstruction.
• The GAN algorithm was designed to overcome signal fading in Mach-Zehnder interferometers (MZI).
• The model was tested on standard and specialized BCG data, including sinus arrhythmia and post-exercise signals.

Main Results:

• The proposed GAN achieved a high correlation of 0.952 between reconstructed and reference BCG signals.
• The method successfully reconstructed BCG signals despite MZI-induced fading.
• The approach eliminated the need for additional modulators and demodulators in the MZI setup.

Conclusions:

• A novel BCG reconstruction algorithm effectively solves signal fading in optical fiber interferometers.
• The developed method significantly simplifies BCG monitoring systems.
• This innovation holds promise for more accessible and cost-effective cardiac monitoring.