The Boston Children's Hospital Sleep Corpus: A Collection of 15,695 Annotated Pediatric Polysomnograms

Comprehensive Summary

This study helps address the lack of polysomnography datasets in pediatric sleep research by creating the most diverse set to date called the Boston Children's Hospital (BCH) Sleep Corpus. The dataset is comprised of millions of sleep stage annotations, breathing, waking, and limb movement occurrences, and thousands of diagnoses using patient health records. Each of the polysomnographies had a median sleeping duration of 8.9 hours, and overall totaled 139,208 hours in EEG data. The datasets showed some trends like the decrease of REM sleep from 33.5% of neonates to 16.3% of teenagers. N2 sleep inreased from 21.7% to 35.4%. Apenas declined with age progression. The datasets also showed age- and region- specific EEG trends across pediatric patients.

Outcomes and Implications

As the largest pediatric polysomnography dataset, the BCH Sleep Corpus holds immense potential to improve treatments for various sleeping disorders by allowing physicians to analyze trends and linkages between data sets. The work will help biomedical technology developers to create artificial intelligence systems that will quickly analyze this dataset and help physicians provide the best possible care for their patients by working in tandem with them to diagnose pediatric patients more accurately and efficiently.