Deep learning vs. conventional techniques for processing and classifying EEG brain disorders: A survey

Manual analysis of electroencephalogram (EEG) data, to diagnose brain-related disorders has many deficiencies such as the time and effort needed to process this data. Moreover, highly trained clinicians or neurophysiologists are needed to interpret It. In contrast, automated analysis of this data through machine-assisted detection of brain disorders can mitigate such deficiencies. This paper gives a survey on the various machine learning techniques proposed to process EEG data for the purpose of classification through automated recognition of patterns embedded in brain signals extracted from the features. It outlines deep learning models used in classification of brain disorders based on EEG, and discusses their advantages, drawbacks, limitations, and challenges compared to conventional signal processing approaches. It also sheds light on the high variability in the datasets utilized in this respect.