Dissertation defense (April 30, 2024): Thiago Barral Fernandes Reis
Student: Thiago Barral Fernandes Reis
Title: Detecting Sleep Disorders in Polysomnography Data
Advisors: Felipe da Rocha Henriques (advisor) and Michel Pompeu Tcheou (co-advisor)
Committee: Felipe da Rocha Henriques (Cefet/RJ), Michel Pompeu Tcheou (UERJ), Laura Silva de Assis (Cefet/RJ) and Tadeu Nagashima Ferreira (UFF)
Day/Time: April 30, 2024 / 15 p.m.
Abstract: A good quality sleep means a nonstop sleep and with enough quantity of time of each stage. This quality of sleep provides restoration of physiological, neurological and biological functions. An individual experiencing sleep deprivation may exhibit compromised health. The most common illness related to sleep disorders is sleep apnea that cut off involuntarily the air flow hence waking up the person and decreasing sleep quality. To diagnose sleep disorders the most exam used is the Polysomnography, this exam is basically realized spending a entire night in a lab with a lot of wires monitoring several physiological signs. Currently, due the Internet of Things and wearable gadgets we can monitoring some sings that we can monitoring in Polysomnography exam as well, so this study proposes to predict with a good efficiency if a patient has sleep disorders using machine learning algorithms. In this study four machine learning algorithms were evaluated using data from a real Polysomnography database (derived from thirteen signals). Additionally, the impact of reducing the number of signals on algorithm performance was investigated, with the purpose of facilitating the creation of an initial test (with fewer signals, using wearable devices) to complement the Polysomnography exam. The results indicated that the Random Forest algorithm was the most promising, with satisfactory performance even when considering a smaller number of signals.
