Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Real-Time Diagnostic Integrity Meets Efficiency: A Novel Platform-Agnostic Architecture for Physiological Signal Compression

Neel R Vora, Amir Hajighasemi, Cody T. Reynolds, Amirmohammad Radmehr, Mohamed Mohamed, Jillur Rahman Saurav, Abdul Aziz, Jai Prakash Veerla, Mohammad S Nasr, Hayden Lotspeich, Partha Sai Guttikonda, Thuong Pham, Aarti Darji, Parisa Boodaghi Malidarreh, Helen H Shang, Jay Harvey, Kan Ding, Phuc Nguyen, Jacob M Luber

TL;DR

The paper addresses the challenge of real-time, energy-efficient transmission of head-mounted physiological signals by introducing a platform-agnostic variational autoencoder (VAE) for spectrogram-based compression. It jointly optimizes on-device encoding and server-side seizure classification via XGBoost, achieving a remarkable $1:293$ compression ratio while preserving discriminative information, evidenced by 91% seizure-detection accuracy. The approach is validated on real patient data and demonstrated on edge devices (ARM Cortex-V8, Jetson Nano, Raspberry Pi) with notable power-energy savings and potential clinical impact for long-term monitoring and implanted devices. This work advances wearable neuro-monitoring by enabling low-latency, low-power processing of multi-channel EEG/EMG/EOG/ECG signals with practical deployment capabilities.

Abstract

Head-based signals such as EEG, EMG, EOG, and ECG collected by wearable systems will play a pivotal role in clinical diagnosis, monitoring, and treatment of important brain disorder diseases. However, the real-time transmission of the significant corpus physiological signals over extended periods consumes substantial power and time, limiting the viability of battery-dependent physiological monitoring wearables. This paper presents a novel deep-learning framework employing a variational autoencoder (VAE) for physiological signal compression to reduce wearables' computational complexity and energy consumption. Our approach achieves an impressive compression ratio of 1:293 specifically for spectrogram data, surpassing state-of-the-art compression techniques such as JPEG2000, H.264, Direct Cosine Transform (DCT), and Huffman Encoding, which do not excel in handling physiological signals. We validate the efficacy of the compressed algorithms using collected physiological signals from real patients in the Hospital and deploy the solution on commonly used embedded AI chips (i.e., ARM Cortex V8 and Jetson Nano). The proposed framework achieves a 91% seizure detection accuracy using XGBoost, confirming the approach's reliability, practicality, and scalability.

Real-Time Diagnostic Integrity Meets Efficiency: A Novel Platform-Agnostic Architecture for Physiological Signal Compression

TL;DR

The paper addresses the challenge of real-time, energy-efficient transmission of head-mounted physiological signals by introducing a platform-agnostic variational autoencoder (VAE) for spectrogram-based compression. It jointly optimizes on-device encoding and server-side seizure classification via XGBoost, achieving a remarkable compression ratio while preserving discriminative information, evidenced by 91% seizure-detection accuracy. The approach is validated on real patient data and demonstrated on edge devices (ARM Cortex-V8, Jetson Nano, Raspberry Pi) with notable power-energy savings and potential clinical impact for long-term monitoring and implanted devices. This work advances wearable neuro-monitoring by enabling low-latency, low-power processing of multi-channel EEG/EMG/EOG/ECG signals with practical deployment capabilities.

Abstract

Head-based signals such as EEG, EMG, EOG, and ECG collected by wearable systems will play a pivotal role in clinical diagnosis, monitoring, and treatment of important brain disorder diseases. However, the real-time transmission of the significant corpus physiological signals over extended periods consumes substantial power and time, limiting the viability of battery-dependent physiological monitoring wearables. This paper presents a novel deep-learning framework employing a variational autoencoder (VAE) for physiological signal compression to reduce wearables' computational complexity and energy consumption. Our approach achieves an impressive compression ratio of 1:293 specifically for spectrogram data, surpassing state-of-the-art compression techniques such as JPEG2000, H.264, Direct Cosine Transform (DCT), and Huffman Encoding, which do not excel in handling physiological signals. We validate the efficacy of the compressed algorithms using collected physiological signals from real patients in the Hospital and deploy the solution on commonly used embedded AI chips (i.e., ARM Cortex V8 and Jetson Nano). The proposed framework achieves a 91% seizure detection accuracy using XGBoost, confirming the approach's reliability, practicality, and scalability.
Paper Structure (20 sections, 5 equations, 10 figures, 1 table)

This paper contains 20 sections, 5 equations, 10 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background and Motivation
  3. System Overview
  4. System Training
  5. On Chip Testing
  6. Variational Auto-Encoder
  7. VAE on the Chip
  8. Architecture
  9. Quantization of VAE
  10. XGBoost
  11. A Real-world Study
  12. Data Acquisition
  13. Hardware/Software Implementation
  14. Performance Evaluation
  15. VAE Performance
  16. ...and 5 more sections

Figures (10)

  • Figure 1: System Overview
  • Figure 2: Detailed architecture of the VAE used in this study. Each input is fed through a ResNet 18 network. This network finds the recognition model that has the highest likelihood of generating the input $x$. Then, the decoder (an inverted ResNet18) tries to reconstruct the input based on the $n$ dimensional latent vector learned before.
  • Figure 3: Four randomly selected samples from the test set and their reconstructed versions when latent size is 64. Note that due to normalization and gray-scale transformation, the samples are not in color.
  • Figure 4: Training and validation loss (KL divergence loss plus reconstruction loss) graph for VAE for 64-dimension latent space over steps (i.e. iteration on different batchs)
  • Figure 5: Hardware setup for power measurement and data transfer. Quantized VAE used on Jetson Nano. Power measure for both compression and transfer of compressed data to the server.
  • ...and 5 more figures