Improving the detection of auditory steady-state responses near 80 Hz using multiple magnitude-squared coherence and multichannel electroencephalogram

Abstract

Abstract The detection of auditory steady-state responses (ASSRs) is an important task in hearing sciences. Since the signal-to-noise ratio (SNR) of ASSR is low, only signal processing techniques can reveal the presence of such responses immersed in the background electroencephalogram (EEG). The Objective Response Detection (ORD) techniques are mathematical tools that perform an automatic hypothesis test in frequency domain and the Magnitude-Squared Coherence (MSC) is one of the most efficient univariate ORD technique. For a fixed SNR, the only way of improving the detection power of MSC is by using longer signals. However, recently the Multiple Magnitude-Squared Coherence (MMSC) was developed as a multivariable extension of the MSC technique. This multivariate approach may be used to shorten the data length used for detection by using more EEG channels. Thus, this work aims at investigating the performance of the MMSC in the detection of ASSRs and to compare the results with MSC. The ASSRs were elicited by amplitude modulation pure tones delivered binaurally at 70 dB SPL to 24 volunteers with normal hearing thresholds. The carrier frequencies were 500, 1000, 2000 and 4000 Hz and they were modulated with frequencies near 80 Hz. The maximum detection rate and the minimum average detection time obtained were 95.31% and 146.69 s, respectively, and this result was obtained by MMSC using a set of five electrodes (Fz, Pz, Oz, F4 and T4). This result was significantly better than the best MSC result and represented an overall improvement around 18%. Therefore, it is concluded that the multivariate method MMSC has advantage when applied to the detection of ASSRs evoked near 80 Hz.