Applying a computer algorithm to classify epileptic activity provides critical and reliable information related to localising seizure origin in patients with severe epilepsy beyond that provided by previous approaches, according to a study published in Neuroimage.
Many patients who have severe drug-resistant epilepsy display epileptic spikes on their EEG – a device which gathers information about brain activity and seizures using electrodes placed on the scalp. These spikes represent brief neuronal discharges and appear as sharp waves on the EEG. Neuronal discharges are more frequent than seizures and can come from the same location in the brain.
EEGs provide valuable information about the epileptiform network and are helpful during the pre-surgical evaluation of patients with severe epilepsy.
Lead author Professor Louis Lemieux from UCL’s Institute of Neurology said: “Studying spikes is advantageous as they are more common than seizures and in some patients can point to the region where seizures originate, based on the EEG and clinical manifestation at the start of the seizure.
“If they are confirmed to originate from the same place as the seizures, it can be a promising sign that they are coming from a small localised area of brain, which is very useful for planning the surgical strategy.”
Currently the detection and classification of epileptiform signals collected on an EEG relies on expert observers. This is a very time consuming procedure, which can also lead to variability.
In this study 76 patients undergoing pre-surgical evaluation were recruited and recordings were taken using functional Magnetic Resonance Imaging (fMRI) – a form of brain scanning capable of detecting and localising changes in brain activity.
The team, part-funded by the BRC, used an automatic spike sorting algorithm called Wave_clus to classify epileptic activity visually identified in eight of the patients in whom at least 200 neuronal discharges were recorded.
Wave_clus was found to be an efficient tool. The algorithm identified 29 classes of epileptic activity in the form of neuronal discharges compared to 26 that had been based on visual identification. There was a full match of EEG patterns in two cases, additional EEG information in two cases and covering EEG patterns of the same areas as expert classification in seven of the eight cases.
Professor Lemieux said: “We used the algorithm to encode and make more objective the classification of spikes in EEGs. Compared to the visual subjective classification by experts, there are differences suggestive of a significant improvement.
“The key clinical implication in terms of how EEGs are interpreted more objectively will require validation in time. As processes become more automated, the expert becomes more expert at designing such processes”.
To read Classification of EEG abnormalities in partial epilepsy with simultaneous EEG–fMRI recordings in full click here.
