Machine learning predicts seizure activity from high-frequency oscillation rates

The analysis of Patient-3 (good result—TN). (A) Preoperative T1 MRI (left) and postoperative T1 MRI (right). (B) An example of HFO, defined as the co-occurrence of ripple and FR, highlighted in red in their respective filtered range. (C) The distribution of HFO tariffs on the channels (events/minute). Channels whose HFO level exceeds the 95% threshold (black horizontal line) form the HFO zone, highlighted in green. Channels in the resected area are marked in red. The HFO zone was included in the resection which led to good results (TN). Credit: Frontiers of Human Neuroscience (2022). DOI: 10.3389/fnhum.2022.984306

In a groundbreaking study, researchers from HSE University, the RAN Institute of Linguistics and the Pirogov Center measured and analyzed high-frequency oscillations (HFOs) in different regions of the brain. An automated detector predicted seizure outcomes based on HFO levels with an 85% accuracy rate and, by applying machine learning, was able to distinguish between epileptogenic and non-epileptogenic HFOs.

The results of the study are published in Frontiers of Human Neuroscience.

High-frequency oscillations are short-term brain events that, when observed using electroencephalography (EEG), help identify brain regions generating epileptic seizures. Retrospective studies confirm that resection of tissue in these areas can help stop seizures.

However, prospective studies, that is, those conducted to predict surgical outcomes, have reported mixed results. In some cases, resection of tissue in an area with a high number of HFOs detected – and therefore assumed to be epileptogenic – did not cause the seizures to stop.

According to the authors of this study, one of the reasons for the failure to predict surgical results could be the fact that patients were observed during REM sleep or wakefulness. The authors further argue that HFO data from deep sleep (NREM) could significantly improve the prognostic value of HFO levels, but not by much if NREM sleep periods were too short or too few. Another limitation of previous studies was the performance of the detectors used to measure HFO levels.

Researchers from HSE University, the RAN Institute of Linguistics and the Pirogov Center examined differences in HFO amplitude, duration and frequency between healthy and epileptogenic brain tissue. They analyzed HFO levels in the mesial temporal and neocortical regions of patients during NREM sleep using an automatic detector clinically validated in previous studies.

The study predicted seizure outcomes with 85% accuracy. Achieving 100% accuracy would have been impossible, as the detector was unable to distinguish between epileptogenic and healthy HFO levels. This limitation was partly solved by applying machine learning.

The researchers found a marked difference in amplitude between epileptogenic and non-epileptogenic HFO levels in the neocortex (frontal, temporal and parietal lobes). This difference was less pronounced in the mesial temporal regions, where HFO duration was a more important distinction. High-frequency epileptogenic oscillations are approximately the same in all brain regions in terms of amplitude, frequency, shape, and duration. Sound oscillations, however, are very different – mainly in amplitude – in different regions.

“Our results demonstrate that by observing HFO, we can detect epileptogenic areas. This result could be further improved in the future. Machine learning will help distinguish between epileptogenic and healthy oscillations based on their amplitude, frequency and duration,” said Victor Karpychev, research assistant at the HSE Center for Language and the Brain.

This study also indicates that the accuracy of using HFO to identify epileptogenic tissues may be higher if a reliable automatic detector is used during the patient’s NREM sleep phase.

Provided by National Research University Graduate School of Economics