Comprehensive Sleep Laboratory Data: August - October 2024

Abstract

The Sleep Laboratory at Sunnybrook Health Sciences Centre collects a rich dataset of overnight physiological polysomnography and health questionnaire data from participants who are referred by their physicians for a sleep study. Through the collection of data at the sleep laboratory from clinical diagnostic studies, we have generated, and will continue to collect, a large and diverse dataset. We are prospectively collecting data from all consenting participants who are tested at the sleep laboratory and have created a de-identified dataset that contains raw polysomnography data, aggregated sleep study metrics, medical comorbidities and health questionnaire data, as well as medication information. This extensive sleep and health data will allow for a large variety of research topics that can be explored by future T-CAIREM users, such as predicting disease states like Parkinson’s disease through the analysis of sleep signals.

Background

Sleep disorders affect a large proportion of the population and can significantly impact overall health and quality of life. Additionally, many other disorders such as hypertension, stroke, and cognitive impairment are comorbid with sleep problems, and the treatment of sleep disorders, such as sleep apnea, can improve comorbid health outcomes. Our project aims to collect a large amount of physiological sleep data for AI training and research that can be used, among other purposes, to (i) gain better understanding of how sleep disorders may be risk factors for various medical conditions; (ii) better understand cardiovascular and neurological changes during sleep; (iii) create tools that will facilitate more efficient scoring of polysomnograms through the use of automated sleep scoring algorithms.

Methods

Patients attending the Sunnybrook Sleep Laboratory for a clinical sleep study are given an electronic consent form asking if they consent to having their de-identified data used for research purposes on the Health Data Nexus platform. At night, participants complete a series of questionnaires on a tablet. They then complete an overnight polysomnography (signals described below). The participants will then complete a questionnaire the next morning describing their sleep experience. For clinical purposes, a sleep technologist will score the data for sleep stages, respiratory events, and movements, and the PSG metrics will be aggregated.

Data Description

EDF Files of Polysomnography Data

The Sunnybrook Sleep Laboratory uses a level 1, technologist-monitored in-hospital polysomnography system (Compumedics Neuroscan, Australia, Profusion PSG 5, Grael V2 headbox) which can be set to collect raw PSG signals and export the data as European Data Format (EDF) files. The raw PSG signals include multi-channel electroencephalography (EEG), electrooculography (EOG), electromyography (EMG), electrocardiography (ECG), nasal flow, chest effort, leg movements, body position, and blood oxygen level, which are all routinely collected for clinical care.

Excel Data File

Each row is a unique participant. Please see the attached Questionnaire package and sleep variables for a description of each column. Sleep metrics are scored for clinical purposes by a registered sleep technologist.

The questionnaires included in the battery are as follows:

1. Medical co-morbidities and medications

• Self-administered comorbidity questionnaire
• Medication list
• Lifestyle behaviour and demographics questionnaire (self-developed; asks about demographics, alcohol use, cannabis use, cigarette use, diet and exercise)

2. Screening for sleep disorders

• The STOP-BANG for obstructive sleep apnea risk
• Insomnia severity index
• International Restless legs Syndrome (IRLS) Severity Rating Scale. If patients endorse the first question, they continue the questionnaire, and if not they can skip the remainder of this questionnaire.
• Munich Parasomnia Screening (MUPS). This asks about various parasomnias, or disorders that involve muscle activity during sleep. If patients endorse the first question, they continue the questionnaire, and if not, they can skip the remainder of this questionnaire.

3. Sleep Quality

• Epworth Sleepiness Scale
• Pittsburgh Sleep Quality Index

4. Chronotype (i.e. propensity to sleep earlier or later in the night)

• Horne Ostberg (Morning-Eveningness) Questionnaire.

5. Functional Disability and Quality of Life

• EQ-5D-5L

Usage Notes

Most of the data is available as an excel file. The raw PSG data needs an EDF compatible software to use. In the Python research environment, users can use the PyEDFlib library.

After setting a path to a file in the data folder, you can load in the data with the following command:

from pyedflib import highlevel
import pyedflib as plib
signals, signal_headers, header = highlevel.read_edf(path)

Then signals provides the physiological signals associated with the patient, signal_headers gives a description of each signal, and header provides metadata on the patient. From here, the signals can be analyzed and plotted.

Ethics

REB #6197. This study is approved by the Sunnybrook Research Ethics Board and reviewed by the University of Toronto REB.

Acknowledgements

This project is funded by the T-CAIREM Health Data Nexus Dataset Grants. The study investigators also thank the sleep technologists and administrative staff of the Sunnybrook Sleep Laboratory.

Thank you to the following co-investigators: Dr. Marc Narayansingh, Dr. Karthi Umapathy, Dr. Andrew Lim, Dr. Houman Khosravani, Vikash Nanthakumar

Conflicts of Interest

n/a

References

1. T Kendzerska, BJ Murray, . . . MI Boulos. Polysomnographic Assessment of Sleep Disturbances in Cancer Development: A Historical Multicenter Clinical Cohort Study. Chest. 2023.
2. DR Colelli, GR Dela Cruz, . . . MI Boulos. Impact of sleep chronotype on in-laboratory polysomnography parameters. J Sleep Res. 2023:e13922.