Seizure prediction and forecasting: a scoping review

Abstract

Purpose of review

This scoping review summarizes key developments in the field of seizure forecasting.

Recent findings

Developments have been made along several modalities of seizure forecasting, including long term intracranial and subcutaneous encephalogram, wearable physiologic monitoring, and seizure diaries. However, clinical translation of these tools is limited by various factors. One is the lack of validation of these tools on an external dataset. Moreover, the widespread practice of comparing models to a chance forecaster may be inadequate. Instead, the model should be able to at least surpass a moving average forecaster, which serves as a ‘napkin test’ (i.e., can be computed on the back of a napkin). The impact of seizure frequency on model performance should also be accounted for when comparing performance across studies. Surprisingly, despite the potential for poor quality forecasts, some individuals with epilepsy still want access to imprecise forecasts and some even alter their behavior based upon them.

Summary

Promising advances have been made in the development of tools for seizure forecasting, but current tools have not yet overcome clinical translation hurdles. Future studies will need to address potentially dangerous patient behaviors as well as account for external validation, the napkin test, seizure frequency dependent metrics.