Diagnosing Dissatisfaction in the Medical Profession (STS Research Paper)
Lee, Sharon, School of Engineering and Applied Science, University of Virginia
Odumosu, Toluwalogo, EN-Comp Science Dept, University of Virginia
Allen, Timothy, EN-Biomed Engr Dept, University of Virginia
Sleep quality is increasingly recognized as a major health issue. However, sleep is poorly understood, and current methods for sleep tracking are either too costly or inaccurate for long-term use. We evaluate a method for sleep stage classification using only signals accessible to mobile health devices, specifically cardiac, respiratory, and movement signals. Using polysomnography data from a set of 20 patients, we evaluate three classifiers: a random forest, a support vector machine, and a recurrent neural network. Using random forest and support vector classification, we achieve accuracies of 87.51 % and 85.48%, respectively. Post-processing smooths the predicted classifications to increase accuracy. Including a time-based feature to account for sleep architecture also improves accuracy. Our results support the efficacy of mobile health devices for sleep tracking.
This paper is designed to analyze which actors form and impact the US medical profession as well as how they do this. Most importantly, I will attempt to elucidate the forces that shape doctors' expectations of the profession in order to determine where and why these expectations have not been met. Both of these efforts implicitly utilize an actor network perspective and additionally are inspired by the co-productionist work. Leading scholars in STS, such as Sheila Jasanoff, Michel Callon, and Michael Lynch, have utilized the framework of coproduction to illuminate "how cognitive understandings of the world we live in are tied at many points to social means of intervening in or coping with the world," (Jasanoff, 2004). First, I will begin with a brief history of how the profession has changed in order to identify the actors that impact and define the profession. Second, I will examine how doctors' initial expectations are formed and the ways in which this perspective changes negatively as these expectations are failed. Finally, I will propose suggestions for change and present further questions that have arose from my research.
BS (Bachelor of Science)
machine learning, sleep stage classification, random forest, support vector machine, polysomnography, mobile health.
Other Title: Evaluation Machine Learning Methods for Sleep Stage Classification using Non-invasive Biometrics (Technical report);
School of Engineering and Applied Science; Bachelor of Science in Biomedical Engineering; Technical Advisor: Timothy Allen; STS Advisor: Toluwalogo B. Odumosu.
All rights reserved (no additional license for public reuse)