Sleep Electroencaphalography-Based Brain Age Index

Abstract

The human electroencephalogram (EEG) of sleep undergoes profound changes with age. These changes can be conceptualized as "brain age (BA)," which can be compared to chronological age to reflect the degree of deviation from normal aging. Here, we develop an interpretable machine learning model to predict BA based on 2 large sleep EEG data sets: the Massachusetts General Hospital (MGH) sleep lab data set (N = 2532; ages 18-80); and the Sleep Heart Health Study (SHHS, N = 1974; ages 40-80). The model obtains a mean absolute deviation of 7.6 years between BA and chronological age (CA) in healthy participants in the MGH data set. As validation, a subset of SHHS containing longitudinal EEGs 5.2 years apart shows an average of 5.4 years increase in BA. Participants with significant neurological or psychiatric disease exhibit a mean excess BA, or "brain age index" (BAI = BA-CA) of 4 years relative to healthy controls. Participants with hypertension and diabetes have a mean excess BA of 3.5 years. The findings raise the prospect of using the sleep EEG as a potential biomarker for healthy brain aging.

Background

Human sleep undergoes robust and predictable changes with age, reflected in both sleep macrostructure and electroencephalographic (EEG) microstructure. In terms of sleep macrostructure, older participants sleep earlier and wake earlier. They have shorter sleep duration, increased sleep fragmentation, and reduced percentages of rapid eye movement (REM) sleep, as well as (at least in males) deep non-REM (NREM) sleep. In terms of EEG microstructure, older participants exhibit reduced slow waves during deep sleep (Carrier et al., 2001; Larsen et al., 1995), decreased sleep spindle amplitude, density, and duration (Purcell et al., 2017), and less phase coupling between slow oscillations and sleep spindles (Helfrich et al., 2017). Compared to the sleep macrostructure, EEG microstructures have direct neural structural bases and are closely related to brain functions and brain health. Therefore, we propose brain age (BA) based on sleep EEG, and the deviation of brain age from chronological age as brain age index (BAI). We hypothesize that higher BAI is associated with neuropathological conditions such as dementia.

Software Description

An implementation of sleep EEG-based brain age (BA) and brain age index (BAI) can be found on the Luna sleep signal analysis software website, with detailed documentation. An online interactive version is available at Moonlight. Both websites are created by Dr. Shaun Purcell and Dr. Senthil Palanivelu at Brigham Women's Hospital and Harvard Medical School. The original GitHub repo is at https://github.com/bdsp-core/SleepEEGBasedBrainAge. The training set of healthy subjects for training BAI is available in this project, named "BAI_MGH_healthy_BDSP.csv"

Technical Implementation

Internally, the brain age model requires extracting per-epoch spectral features, spindle characteristics, and time-domain statistics from EEGs. Then the features are averaged across epochs from different sleep stages. The averaged features from different sleep stages are concatenated and fed to the brain age model, which is a linear model followed by age-dependent correction. The original implementation requires all six channels. The implementation on Luna requires C3-M2 and C4-M1 channels, with a slight increase in mean absolute error. The EEGs should have an uV unit, notch-filtered to remove line noise, and band-pass filtered at 0.3-35Hz. Removing ECG artifacts is optional but recommended.

Installation and Requirements

Please see here. For Windows, a docker is required.

Usage Notes

A certain level of command-line familiarity is required when using the BAI model in Luna.

Release Notes

The BAI model is included in Luna v0.99 and above.

Ethics

Data collection and sharing for the Sleep Electroencaphalography-Based Brain Age Index (BAI) project is performed under Institutional Review Board (IRB) approvals at Massachusetts General Hospital (MGH), with a waiver of the requirement for informed consent. MGH data is generated as part of usual patient care. All data is de-identified.

Acknowledgements

MBW reports grants from NIH-NINDS (NIH-NINDS 1K23NS090900, 1R01NS102190, 1R01NS102574, 1R01NS107291). MTB has received funding from the Center for Integration of Medicine and Innovative Technology, the Milton Family Foundation, the MGH-MIT Grand Challenge, and the American Sleep Medicine Foundation, and the Department of Neurology.

Conflicts of Interest

MTB was a part of the development team. RJT reports the following: (1) Patent, license and royalties from MyCardio, LLC, for an ECG-based method to phenotype sleep quality and sleep apnea; (2) Grant support, license and intellectual property (patent submitted) from DeVilbiss Healthcare; (3) GLG consulting for general sleep medicine; (4) Intellectual Property (patent) for a device using CO2 for central / complex sleep apnea. This is not an industry supported study, and none of these entities had any role in the study.