Mapeamento do escalpo para detecção de respostas auditivas a partir de detectores objetivos de resposta multivariados

Abstract

This study investigated Auditory Steady-State Responses (ASSRs), which are bioelectrical potentials generated in the brain by auditory stimuli. Used in automated pure-tone audiometry, ASSRs do not require behavioral responses from patients, being especially useful when cooperation is limited, such as in infants. Amplitude-modulated tones are commonly used to evoke ASSRs, and their detection is statistically performed by objective detectors, such as magnitude-squared coherence (MSC), synchronous component measurement (CSM), and the local F-test (STF). Multivariate detectors, such as multivariate magnitude-squared coherence (MMSC), multivariate synchronous component measurement (MCSM), and the multivariate local F-test (MLFT), analyze multiple electroencephalogram (EEG) signals, and their effectiveness depends on the signal-to-noise ratio (SNR). To automatically predict auditory thresholds, it is necessary to detect ASSRs with very low SNRs. This project compared the performance of MMSC and MCSM in detecting ASSRs in scalp regions that evoke the highest amplitudes, using a genetic algorithm model to map these areas and a heuristic to optimize electrode combinations. The results highlighted the superiority of MCSM over MMSC in the best electrode combinations, as well as the influence of phase shift and correlation between EEG signals on detector performance.