Related Experiment Video
Updated: Jun 28, 2025
09:09
Design and Construction of a Cost Effective Headstage for Simultaneous Neural Stimulation and Recording in the Water Maze
Published on: October 13, 2010
10.7K
Design of a Sleep Modulation System with FPGA-Accelerated Deep Learning for Closed-loop Stage-Specific In-Phase
Mingzhe Sun1, Aaron Zhou1, Naize Yang1
1Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON, Canada M5S 2E4.
IEEE International Symposium on Circuits and Systems Proceedings. IEEE International Symposium on Circuits and Systems
|April 16, 2024
Summary
This study introduces a novel closed-loop sleep modulation system using a lightweight deep learning model on an FPGA. It improves sleep stage classification accuracy, overcoming limitations of wired systems and enhancing sleep disorder treatments.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Artificial Intelligence
Background:
- Closed-loop sleep modulation shows promise for treating sleep disorders but faces challenges.
- Wired connections and limited real-time sleep stage classification hinder current systems.
- Developing non-invasive, accurate, and portable sleep modulation technology is crucial.
Purpose of the Study:
- To develop a novel, on-device closed-loop sleep modulation system.
- To overcome limitations of wired instrumentation and improve sleep stage classification accuracy.
- To enable enhanced sleep benefits and treatment for sleep disorders.
Main Methods:
- Developed a lightweight deep learning (DL) model for sleep stage classification using single-channel EEG.
- Accelerated the DL model using a low-power field-programmable gate array (FPGA) for on-device processing.
- Employed convolutional neural networks (CNNs) and a bidirectional long-short-term memory (LSTM) network with 8-bit quantization.
Main Results:
- Achieved state-of-the-art sleep stage classification accuracy of 85.8% and a F1-score of 79% on a public dataset.
- Demonstrated the DL model's potential for generalization across different channels and input data lengths.
- Successfully demonstrated closed-loop in-phase auditory stimulation on a test bench.
Conclusions:
- The developed system overcomes key barriers in closed-loop sleep modulation, enabling portable and effective sleep treatments.
- The lightweight DL model on FPGA offers efficient and accurate real-time sleep stage classification.
- This technology paves the way for advanced, personalized sleep modulation therapies.
