Análise de estratégias de testes sequenciais e de métodos de descorrelação do EEG multicanal na detecção automática de resposta auditiva

• Ano de defesa: 2023
• Autor(a) principal: Patrícia Nogueira Vaz
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ENG - DEPARTAMENTO DE ENGENHARIA ELÉTRICA; Programa de Pós-Graduação em Engenharia Elétrica; UFMG
• Palavras-chave em Português: Resposta auditiva em regime permanente; Eletroencefalograma (EEG); Estratégia para testes sequenciais; Correção de Cholesky; Audiometria; Engenharia elétrica; Eletroencefalografia; Engenharia biomédica
• Link de acesso: http://hdl.handle.net/1843/64560; https://orcid.org/0000-0003-4503-6288

The Auditory Steady-State Response (ASSR) serves as a crucial tool in determining auditory thresholds in individuals who cannot or are unwilling to cooperate with conventional behavioral testing methods. This study evaluates sequential testing techniques for automated ASSR detection, considering both sensitivity and test time. The sequential testing strategies were subjected to tests using the same database and compared by means of uniform performance measures. Additionally, a multichannel electroencephalogram (EEG) decorrelation analysis was conducted in the temporal and, innovatively, frequency domains to enhance test sensitivity. The Cholesky correction method was employed to adjust critical values based on the correlation between EEG channels. All proposed strategies, both sequential testing and decorrelation, demonstrated the ability to control the increase in false-positive rates, remaining within the 90% confidence interval. Sequential testing strategies, when applied with optimized parameters, showed a significant reduction in test time compared to the single-test approach, in some cases without a significant loss in sensitivity. Notably, the latest critical value correction strategies resulted in a 60% reduction in examination time in the absence of a response. Regarding decorrelation methods, the proposed novel approach achieved similar performance, being on average 27 times faster in terms of computational time needed to estimate the critical value compared to the time-domain method. These results highlight the significant potential of sequential testing to enhance automatic audiometry performance, as well as the utility of the decorrelation method when multichannel ASSR data is available. This work is pioneering in conducting a comprehensive comparison of sequential testing strategies on a single database.