Português English
Contato

Dissertação de Carlos Fabiel Bublitz


Detalhes do Evento


Aluno: Carlos Fabiel Bublitz
Orientador: Prof. Dr. Juergen Rochol
Coorientador: Prof. Dr. Cristiano Bonato Both

Título: A Non-intrusive OSA Severity Estimation for CPAP Therapy Screening based on Snoring Acoustical Analysis
Linha de Pesquisa: Arquiteturas, Protocolos e Gerência de Redes e Serviço

Data: 24/09/2018
Hora: 10h
Local: Anfiteatro Heitor M C Lima no Prédio 3 da UFCSPA.

Banca Examinadora:
– Prof. Dr. Eduardo Alves do Valle Junior (UNICAMP – por videoconferência)
– Prof. Dr. Denis Martinez (UFRGS)
– Prof. Dr. Weverton Luis da Costa Cordeiro (UFRGS)

Presidente da Banca: Prof. Dr. Juergen Rochol

Abstract: Obstructive Sleep Apnea (OSA) is characterized by repeated episodes of partial (hypopnea) or complete (apnea) obstruction of the upper airway during sleep. The clinical effects of OSA are related to the cumulative effects of exposure to periodic asphyxia and sleep fragmentation caused by apneas and hypopneas, such as an increased risk of hypertension, nocturnal dysrhythmias, ventricular failure, myocardial infarction, and stroke. The current gold standard for diagnosing OSA is the overnight Polysomnography (PSG), which requires a full-night sleep laboratory stay, attached to different biological sensors and under the supervision of a technician. Besides the discomfort caused by the invasive sensors, the necessity of a clinical setting and highly specialized infrastructure results in a long waiting list in sleep laboratories and high costs, thus restricting the access to diagnosis and treatment. To improve monitoring of OSA evolution, access to diagnosis and treatment follow up, we propose a Mobile Health (mHealth) solution to take advantage of the smartphone capabilities to deploy a non-invasive OSA severity estimation. We make use of the audio recorded through a smartphone to automatically detect snoring events throughout the night and through the analysis of such events estimate patient’s necessity for Continuous Positive Airway Pressure (CPAP) therapy. For that, we have divided our solution into two phases: (i) a completely unsupervised solution to automatically detect the snoring events in an uncontrolled environment and (ii) the analysis of acoustical features of the snoring events for OSA severity estimation. In the first phase, we can prove the viability of recording the audio and detect the snoring events using a smartphone under an environment susceptible to random noises. In the second phase, we show that a set of global acoustical features from the snoring events can predict the patient’s need for the CPAP therapy. Our proposed solution was evaluated in an uncontrolled (patient’s home) and controlled (sleep laboratory center) environment, reaching satisfactory results in snoring events detection and patient’s classification according to the need for CPAP therapy.

Keywords: OSA, Acoustical Analysis, Clustering, Digital Signal Processing, Mobile Health.