Understanding High-Dimensional Time-Series: Topological and Visual Analytics for Characterizing Sleep Apnea. (R-12224)

This project studies insight generation and exploration of the rich measurements used to diagnose sleep apnea. The current diagnostic process involves a full-night polysomnography (PSG). PSG records containing hours of measurements are summarized using a single-value score. This score does not provide sufficient information to estimate sleep apnea disease severity reliably. Consequently, there is a need for novel representations that describe sleep apnea. We propose developing topological and visual analytics methods to support clinicians in extracting information from PSG measurements. The goal is to facilitate them in developing novel scores that describe sleep apnea and reduce the scoring time of PSG records leading to quicker diagnoses and better-informed treatment selection. Topological Data Analysis algorithms create a compressed representation of complex data and capture the global structure of the data. Interactive visualizations of these models allow for a qualitative understanding of signal properties without requiring in-depth mathematical knowledge. In addition, interfaces for such models provide state of-the-art visualization challenges because these models create complex interconnected networks. Clinicians are interested in the dynamics of the measures at different resolutions and need to combine information from multiple sensors in their analyses.

Keywords:CLINICAL TRIALS

Disciplines:Data mining, Machine learning and decision making, Decision support and group support systems, Visual data analysis, Biomedical signal processing