Improving Auditory Steady-State Response Detection Using Multichannel EEG Signal Processing

The ability to hear and process sounds is crucial. For adults, the inevitable ongoing aging process reduces the quality of the speech and sounds one perceives. If this effect is allowed to evolve too far, social isolation may occur. For infants, a disability in processing sounds results in an inappropriate development of speech, language, and cognitive abilities. To reduce the handicap of hearing loss in children, it is important to detect the hearing loss early and to provide effective rehabilitation. As a result, hearing of all newborns needs to be screened. If the outcome of the screening does not indicate normal hearing, more detailed hearing assessment is required. However, standard behavioral testing is not possible, so that assessment has to rely on objective physiological techniques that are not influenced by sleep or sedation. The last few decades, the use of auditory steady-state responses (ASSRs) has been investigated as an objective technique to assess hearing thresholds at different frequencies. In this research project, we focus on reducing the required recording time of the ASSR technique and on improving its robustness against unwanted artifacts, generated by e.g. muscle activity, eye blinks, and electrode cable movement. This objective is achieved by processing multichannel electroencephalogram (EEG) recordings. First, we build a setup that allows us to apply custom made stimuli and to record multichannel EEG. Second, the effect of two multichannel processing techniques applied on these data is investigated. Both an independent component analysis (ICA) based technique and a multichannel Wiener filter (MWF) based approach show that a significant measurement time reduction is possible when compared with standard single channel recordings. Afterwards, the ICA- and MWF-based approaches are incorporated into a procedural multichannel framework that is constructed from elements of detection theory. It is shown that this detection theory based approach increases the number of detections significantly when compared with a noise-weighted single channel technique, in the case of artifact-rich EEG. Finally, the optimal electrode positions are determined for the recording of ASSRs originating mainly from the brainstem (and the auditory cortex). After processing with the multichannel EEG processing techniques presented in this work, these positions guarantee a close-to-optimal assessment of the subject’s hearing thresholds.