Author: Chen, Shuyang
Title: Optical fiber sensing with deep learning for biomedical application
Advisors: Yu, Changyuan (EIE)
Degree: Ph.D.
Year: 2022
Subject: Optical fiber detectors
Biomedical engineering
Hong Kong Polytechnic University -- Dissertations
Department: Department of Electronic and Information Engineering
Pages: xxiii, 134 pages : color illustrations
Language: English
Abstract: In the modern aging society, health care has been a major concern among people. Ballistocardiography (BCG) is a vibration signal related to cardiac activity, which can be obtained in a non-invasive way by optical fiber sensors. In this work, an optical fiber interferometer-based BCG monitoring system has been developed. Several deep learning models are explored to optimize this monitoring system.
Firstly, a BCG monitoring system based on the optical fiber sensor is proposed. A moving-coil transducer-based new phase modulation method is developed to address the signal fading problem in the optical fiber interferometer, which can keep the output of the system in quadrature by a closed-loop controller. As a result, a BCG signal without baseline drift can be obtained. This optical fiber interferometer-based BCG monitoring system offers the benefits of being tiny, low-cost, portable, and user-friendly.
Secondly, we study the individual heartbeat waveform detection algorithm based on BCG signals. A convolutional neural network (CNN) is first built to classify the IJK-complex, background, and body movement signals. Since this CNN model needs a series of time-consuming pre-processing works, we propose an end-to-end modified U-net to improve the individual heartbeat waveform detection algorithm. This network has demonstrated its capacity to segment the IJK complex and body movement in the BCG signal with great accuracy.
Then, a modified generative adversarial network (GAN) is presented to reconstruct BCG signals in an optical fiber interferometer with the intensity interrogation mode. This method eliminates the need for extra modulators and demodulators in the interferometer, lowering costs and simplifying the hardware. The results show that the algorithm can reconstruct the BCG signal successfully. To further test the model performance, we have analyzed the reconstruction results based on the collected data on sinus arrhythmia and post-exercise cardiac activities. In conclusion, this signal reconstruction algorithm simplifies the BCG monitoring system by solving the signal fading problem in the optical fiber interferometer in a novel way.
Finally, a compressed sensing (CS) framework is built for the BCG signal based on the optical fiber sensing system. Four types of CS reconstruction algorithms, Basis Pursuit (BP), orthogonal matching pursuit (OMP), and two block sparse Bayesian learning (BSBL) algorithms are used to verify the reconstruction performance in BCG signals under different CRs. The performance of two BSBL algorithms outperforms the other two algorithms. Traditional reconstruction algorithms perform poorly when CR is greater than 90%. Therefore, an end-to-end deep learning model is developed to reconstruct BCG. The performance of the model is good when CR increases from 50% to 90%. For the high CR over 90%, though the performance is slightly degraded, the IJK complex in the BCG can be recovered, and the MAE of the HR is low than 1 bpm when the CR is below 95%.
Rights: All rights reserved
Access: open access

Files in This Item:
File Description SizeFormat 
6596.pdfFor All Users4.82 MBAdobe PDFView/Open

Copyright Undertaking

As a bona fide Library user, I declare that:

  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.

By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

Show full item record

Please use this identifier to cite or link to this item: