Epilepsy is one of the most common neurological brain disorders affecting approximately 50 million people worldwide.1 Up to 70% of epileptic patients are managed with anti-epileptic drugs (AEDs) while the other 30% cannot be managed with AEDs.2 The most affected age groups are those at the extremes of age, and the disease peaks between the ages of 10 and 20 years.3
EEG recordings interpretation: challenges and opportunities
An electroencephalogram (EEG) is one of the main diagnostic tests for epilepsy. Moreover, EEG is used in the focus localization process for patients who undergo focal resection surgery to delineate the brain area where the seizure focus is located. Because this surgery involves removing a part of the brain, many risks are associated, such as difficulties with speech, vision, memory, and movement. A favorable surgical outcome is highly dependent upon accurate localization of the seizure focus. Diagnosis and focus localization are performed by clinical experts through visual discrimination of EEG patterns. For diagnosis, the ictal (seizure) versus normal EEG patterns are compared whereas discrimination between focal versus non-focal EEG channels are examined for focus localization. This process is difficult, time-consuming, and prone to errors. Consequently, there is a clear demand for the application of powerful artificial intelligence (AI) methods for automatic, fast, and accurate binary EEG classification. The EEG signals from epileptogenic brain areas are more stationary, more nonlinear-dependent, and less random compared to those recorded from non-epileptogenic brain areas.4 Because of this, it is possible to automate the focus localization task by extracting the significant patterns that distinguish focal from non-focal EEG channels.