Machine learning predicts epileptogenic activity from high-frequency oscillations

In an innovative study, researchers from HSE, the RAN Institute of Linguistics and the Pirogov Center measured and analyzed high-frequency oscillations (HFO) in different regions of the brain. An automated detector predicted seizure outcomes based on HFO rates with an accuracy rate of 85%, and by applying machine learning made it possible to distinguish between epileptic and non-epileptic HFO.

The results of the study have been published in Frontiers in human neuroscience.

High-frequency short-range oscillations brain Events that, when observed with an electroencephalogram (EEG), help identify the areas of the brain that trigger epileptic seizures. Retrospective studies confirm that removing tissue in such areas can help stop seizures.

However, prospective studies, that is, those conducted to predict surgical outcome, have reported mixed results. In some cases, resection of tissue in an area with a large amount of HFO detected – and thus presumably causing epilepsy – did not cause the seizures to stop.

According to the authors of this study, one of the reasons for the failure to predict the outcome of surgery may be the fact that patients were observed during REM sleep or wakefulness. The authors also argue that HFO data from deep sleep (NREM) can significantly improve the predictive value of HFO rates, but not much if NREM sleep periods are too short or too little. Another limitation of previous studies was the performance of the detectors used to measure HFO rates.

Researchers from HSE, the RAN Institute of Linguistics and the Pirogov Center examined differences in HFO amplitude, duration and frequency between healthy and epileptic brain tissue. They analyzed HFO rates in the temporal regions and neocortex of patients during NREM sleep using an automated detector that has been clinically validated in previous studies.

The study predicted seizure outcomes with an accuracy of 85%. Achieving 100% accuracy was impossible, as the detector was unable to distinguish between epileptic and healthy HFO rates. This limitation has been partially addressed by the app machine learning.

The authors found a significant difference in amplitude between epileptic-causing and non-epileptic HFO rates in the neocortex (frontal, temporal, and parietal lobes). This difference was less pronounced in the intermediate time regions, where HFO duration was a more important distinction. The high-frequency oscillations causing epilepsy are almost the same across regions of the brain In terms of amplitude, frequency, form and duration. However, healthy oscillations differ greatly – mainly in amplitude – in different regions.

“Our findings show that by monitoring heavy fuel oil, we can detect epilepsy-causing regions. This finding can be further improved in the future. It will enable machine learning to distinguish between epilepsy-causing oscillations and healthy oscillations based on their amplitude, frequency and duration.” Victor Karbyshev, research assistant at the HSE Center for Language and the Brain.

This study also indicates that the accuracy of using HFO to identify epileptogenic tissue could be higher if a reliable autodetector was used during the patient’s NREM sleep phase.

Provided by the National Research University Higher School of Economics