Critical slowing as a biomarker for seizure susceptibility

Abstract

words)The human brain has the capacity to rapidly change state, and in epilepsy these state changes can be catastrophic, resulting in loss of consciousness, injury and even death. Theoretical interpretations considering the brain as a dynamical system would suggest that prior to a seizure recorded brain signals may exhibit critical slowing, a warning signal preceding many critical transitions in dynamical systems. Using long-term intracranial electroencephalography (iEEG) recordings from fourteen patients with focal epilepsy, we found key signatures of critical slowing prior to seizures. Signals related to a critically slowing process fluctuated over temporally long scales (hours to days), longer than would be detectable in standard clinical evaluation settings. Seizure risk was associated with a combination of these signals together with epileptiform discharges. These results provide strong validation of theoretical models and demonstrate that critical slowing is a reliable indicator that could be used in seizure forecasting algorithms.

Spatiotemporal patterns of high-frequency activity (80-170 Hz) in long-term intracranial EEG

“…Comparing the circadian patterns of HFA and spike rates to seizures based on the individual patient and individual electrodes extends our knowledge of their relationships. In contrast to previous reports, 22,42,43 our study found a striking difference in circadian patterns of HFA and spike rates across electrodes within patients. This spatial variable behaviors of HFA and spike rates could help reconcile the discrepancies in the literature regarding how spikes and HFOs change prior to seizures.…”

Forecasting cycles of seizure likelihood

“…Similarly, physiological signals recorded from wearable devices could improve forecasts of seizure cycles by providing a continuous measure of underlying rhythms, rather than discrete samples (seizure times). We have shown that fast and slow cycles of brain activity can be measured from continuous EEG across diverse frequencies and regions of cortex 18 . These continuous cycles provided the most accurate estimate of seizure likelihood to date.…”

“…However, in eventual real‐world applications, a can be used to iteratively calibrate the forecast based on past performance. High‐ and low‐risk warning thresholds were computed using a pseudoprospective brute force optimization that maximized the time spent in low‐risk periods and number of seizures classified in high‐risk periods 18 …”

“…A key development has been the understanding that seizure onset is modulated by patient‐specific, cyclic patterns 14‐16 . The rhythmic nature of epilepsy has been well documented for centuries; however, it has only recently become clear that an individual's seizure cycles can be used to develop personalized forecasts of future seizure likelihood 13‐14,17,18 . Although changes in EEG signal characteristics provide the clearest biomarker to track cycles of seizure likelihood, 16,18,19 it is possible to measure individual cycles using only self‐reported seizure diaries 15 .…”

“…The rhythmic nature of epilepsy has been well documented for centuries; however, it has only recently become clear that an individual's seizure cycles can be used to develop personalized forecasts of future seizure likelihood 13‐14,17,18 . Although changes in EEG signal characteristics provide the clearest biomarker to track cycles of seizure likelihood, 16,18,19 it is possible to measure individual cycles using only self‐reported seizure diaries 15 . Therefore, for some people, seizure diaries alone may provide a clinically useful forecast of future high or low seizure risk periods.…”