SpindleSphere: A Web-based Platform for Large-scale Sleep Spindle Analysis and Visualization

• Published in: AMIA ... Annual Symposium proceedings Vol. 2017; pp. 1159 - 1168
• Main Authors: Li, Xiaojin, Cui, Licong, Tao, Shiqiang, Zeng, Ningzhou, Zhang, Guo-Qiang
• Format: Journal Article
• Language: English
• Published: United States American Medical Informatics Association 2017
• Subjects: Algorithms; Brain Waves - physiology; Computer Systems; Data Visualization; Electroencephalography; Humans; Internet; Metadata; Polysomnography; Sleep Stages - physiology
• ISSN: 1942-597X; 1559-4076

Sleep spindles are a hallmark of stage 2 non-REM sleep that contain information about heritable traits that play an important role in neurological diseases. One of the key challenges in leveraging spindles for clinical research is the lack of a data processing pipeline and web-based, platform for managing and visualizing spindle-specific data at scale. We propose SpindleSphere, a scalable integrated data management and visualization platform for spindle research. SpindleSphere has several features: (1) standardized, metadata-based, search and query of annotated polysomnography (PSG), the gold, standard for sleep diagnosis: (2) event-specific signal exporting: (3) interface for interactive waveform visualization: (4) multi-scale spindle rendering: and (5) parallel algorithm in MapReduce for detection of spindle segments. SpindleSphere provides real-time visualization of multi-modal signals from National Sleep Research Resource (NSRR) for spindle characterization. Preliminary evaluation of SpindleSphere was performed on the NSRR (130 GB of PSG data from 300 subjects).