A Deep Learning Strategy for Automatic Sleep Staging Based on Two-Channel EEG Headband Data.

Amelia A Casciola,Sebastiano K Carlucci,Amanda M Punch,Alireza Kazemi,Jason Valerio,Martin J Mckeown,Maryam S Mirian,Daniel Zhou,Michael A Muszynski,Brianne A Kent,Haakon B Nygaard

doi:10.3390/s21103316

Amelia A Casciola, Sebastiano K Carlucci + Show 9 more

Open Access

https://doi.org/10.3390/s21103316

Copy DOI

Abstract

Sleep disturbances are common in Alzheimer’s disease and other neurodegenerative disorders, and together represent a potential therapeutic target for disease modification. A major barrier for studying sleep in patients with dementia is the requirement for overnight polysomnography (PSG) to achieve formal sleep staging. This is not only costly, but also spending a night in a hospital setting is not always advisable in this patient group. As an alternative to PSG, portable electroencephalography (EEG) headbands (HB) have been developed, which reduce cost, increase patient comfort, and allow sleep recordings in a person’s home environment. However, naïve applications of current automated sleep staging systems tend to perform inadequately with HB data, due to their relatively lower quality. Here we present a deep learning (DL) model for automated sleep staging of HB EEG data to overcome these critical limitations. The solution includes a simple band-pass filtering, a data augmentation step, and a model using convolutional (CNN) and long short-term memory (LSTM) layers. With this model, we have achieved 74% (±10%) validation accuracy on low-quality two-channel EEG headband data and 77% (±10%) on gold-standard PSG. Our results suggest that DL approaches achieve robust sleep staging of both portable and in-hospital EEG recordings, and may allow for more widespread use of ambulatory sleep assessments across clinical conditions, including neurodegenerative disorders.

Highlights

We achieve a mean accuracy of 74% when sleep staging HB data, compared to 68% when using a traditional
We have shown that a deep learning (DL) sleep staging model achieves 74% accuracy on low-quality headband EEG data, compared to 77% with gold-standard PSG
The model performs well across all sleep stages, leading to a balanced accuracy of almost 20% more than any machine learning sleep staging method attempted. We show that this model achieves an especially high accuracy for sleep stage N3, with acceptable performance for rapid eye movement (REM) sleep classification, both of which may be highly relevant to the pathophysiology of neurodegenerative disorders [1,6]

Summary

Introduction

The gold standard in sleep assessment is polysomnography (PSG). There are several factors that limit the usefulness of PSG for studying sleep in patients with neurodegenerative diseases. It is relatively expensive, and many PSG studies are statistically under-powered. The unnatural environment and discomfort associated with the numerous electrodes and wires may disturb the subject, and results using PSG may not accurately reflect sleep in the home environment [2]. The proportion of subjects in need of formal sleep assessments outweighs the capacity of accredited PSG sleep laboratories, limiting access to diagnostic services

Methods

Results

Discussion

Conclusion