Abstracts

Rationale:
Sharp-wave ripples (SWRs) are known as the most synchronous neuronal activity evoked by the hippocampus. Interictal epileptiform discharges (IED) demonstrate a state of hypersynchronous depolarization of neuronal activity. Both events are thought to be the interaction between glutaminergic and GABAergic neurons. While SWRs are essential for memory consolidation, epileptiform interictal events are pathological and produce clinical symptoms in many cases. These two events may also display similar characteristics in terms of their frequency and temporal patterns, but IEDs can overlap with SWRs in amplitude and duration. The study’s goals were to validate an algorithm to detect transients (putative SWR and IEDs) in intracranial EEG recordings from patients with epilepsy and to investigate whether it is possible to reliably separate the SWRs from IEDs using data pre-processing and a suitable clustering/classification algorithm.
Method:
The recordings were obtained from patients with therapy-resistant epilepsy (TRE) implanted with depth electrodes at our institution for SEEG. The recording included eight sessions of 24 hours of extracellular recordings from two patients for 384 hours. A statistical pre-processing pipeline was employed for detecting the SWRs in the four electrodes located in the hippocampus. As a result, neural events were identified as the most probable SWRs. Each event’s power in 40 different frequency bands was calculated as its identifying features. Furthermore, the Decision tree classifier was used to classify the putative SWRs and Questionable IEDs (QIEDs) from 1444 power features. Then, the k-means clustering methods were used to ensure that extracted observation can be separated into distinct classes of QIEDs and the putative SWRs. In the last step, the Kullback–Lieber divergence method was used to quantify the separability of the two proposed clusters.
Results:
The efficiency of an automatic SWRs detection algorithm and its requirements were examined using the available recordings. Out of all the detected SWRs, 16% were identified as QIEDs, and the remaining as SWRs. The putative QIED class was mostly expected to happen between seizures. However, the putative SWR class in the hippocampus was most likely to occur in the awake state. Furthermore, the correlation between the frequency bands indicates that putative SWRs and putative QIEDs have the most overlap around 75 Hz centre frequency. Figure (1) represents the two distinct clusters of data which were consisting of QIEDs and putative SWRs candidates.
Conclusion:
Our results indicate that the two neural events are prone to appear at different time intervals. Moreover, the clustering shows that the statistical approach can define these events using their frequency and energy patterns. This ratio is subject to change in different sessions and is affected by the number of seizures per day. The finding elaborates that putative SWRS were less heterogeneous compared to QIEDs.
Funding:
:No funding!