Data Mining the Human Sleep Wake Cycle (original) (raw)

Abstract

"""""""Data Science involves the blending of techniques and theories from various disciplines including mathematics, statistics, engineering, and computer science, to conduct data mining, in order to elicit substantial information from datasets. The interdisciplinary nature of data science provides a wide platform for the utilization and development of powerful computational tools to tackle an array of real world data sets. In this talk, I will discuss my data science based approach in processing human polysomnograms (psgs), sleep records. The human psg consists of a collection of bio-physiological events recorded simultaneously over many hours. Multiple non-invasive electrodes, located on different body regions, are used to obtain information regarding physiological state fluctuations of the subject in order to access sleep health. Although, information regarding healthy adult sleep is well characterized the establishment of robust models for sleep in patients suffering from certain sleep disorders and neurological pathologies are still being demystified. The delay in understanding the sleep patterns of these patients can be contributed to the need for efficient approaches to analyze the large, complex, and at times noisy records generated from sleep studies. In order to meet this need, I have used a combination of signal processing, machine learning, and statistical modeling techniques to probe psgs and obtain information regarding the human sleep cycle of patients with sleep disorders and neurological pathologies. More specifically, I focus on pre-processing, data characterization, and post-processing challenges. My investigations indicate that Data Science offers many robust tools to hasten understanding of the human sleep cycle with respect to pathological conditions. TO VIEW VIDEO OF THIS TALK PLEASE GO TO THE LINK: http://www.youtube.com/watch?v=73A2db9g1QY TO VIEW THE ABSTRACT AND BIO FOR THIS TALK PLEASE GO TO THE LINK: https://www.ee.washington.edu/cgi-bin/research/colloquium/display.pl?id=197 """""""

Jacqueline Fairley hasn't uploaded this talk.

Let Jacqueline know you want this talk to be uploaded.

Ask for this talk to be uploaded.