EEG based detection of conductive and sensorineural hearing loss using artificial neural networks

Abstract

In this paper, a simple method has been proposed to distinguish the normal and abnormal hearing subjects (conductive or sensorineural hearing loss) using acoustically stimulated EEG signals. Auditory Evoked Potential (AEP) signals are unilaterally recorded with monaural acoustical stimulus from the normal and abnormal hearing subjects with conductive or sensorineural hearing loss. Spectral power and spectral entropy features of gamma rhythms are extracted from the recorded AEP signals. The extracted features are applied to machine-learning algorithms to categorize the AEP signal dynamics into their hearing threshold states (normal hearing, abnormal hearing) of the subjects. Feed forward and feedback neural network models are employed with gamma band features and their performances are analyzed in terms of specificity, sensitivity and classification accuracy for the left and right ears across 9 subjects. The maximum classification accuracy of the developed neural network was observed as 96.75 per cent in discriminating the normal and hearing loss (conductive or sensorineural) subjects. From the neural network models, it has been inferred that network models were able to classify the normal hearing and abnormal hearing subjects with conductive or sensorineural hearing loss. Further, this study proposed a feature band-score index to explore the feasibility of using fewer electrode channels to detect the type of hearing loss for newborns, infants, and multiple handicaps, person who lacks verbal communication and behavioral response to the auditory stimulation.