The Human Sleep Project: Polysomnography Recordings for Sleep and Brain Health Research

Approximately 15,000 clinical polysomnography recordings from Massachusetts General Hospital patients form the initial core of this dataset, which is intended to grow to over 200,000 records. The Brain Data Science Platform shares this data to support research on sleep scoring and brain health indicators. The dataset is being used to develop CAISR, a collection of AI tools for better-than-human detection of sleep events and disease risk.

Use Cases

• Training deep neural networks for automated sleep stage scoring based on polysomnography recordings.
• Developing algorithms to detect apnea, hypopnea, and periodic limb movement events from physiological signals.
• Researching 'hidden' information in sleep brain activity to measure risk for neurodegenerative diseases like Alzheimer's.
• Evaluating the impact of lifestyle or pharmacologic interventions on brain health using sleep physiology data.

Strengths

• Initial collection of ~15,000 patient recordings provides a substantial clinical foundation.
• Planned expansion to >200,000 records indicates a large-scale, growing resource.
• Explicitly intended for developing AI tools (CAISR) for advanced sleep analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the current shared summary data is unknown, which may limit suitability assessment.
• Data may reflect clinical bias inherent to the initial hospital source.

Provenance

Source

Brain Data Science Platform, initially with data from Massachusetts General Hospital.

Collection Method

Clinical polysomnography recordings collected from patient evaluations.

Time Range

null

Freshness

Last update date is unknown; freshness unverified.

Geography

null