Forecasting cycles of seizure likelihood

Area of Science:

• Neurology
• Biomedical Engineering
• Digital Health

Background:

• Seizure unpredictability is a major challenge for individuals with epilepsy.
• Seizure occurrences often follow circadian and multiday patterns, influenced by physiological and environmental factors.
• Electronic seizure diaries offer a method for tracking and analyzing seizure patterns.

Purpose of the Study:

• To develop personalized seizure forecasts using a mobile seizure diary application.
• To leverage identified circadian and multiday seizure cycles for predictive modeling.

Main Methods:

• A pseudoprospective study was conducted using data from 50 mobile app users.
• Individual seizure cycles were estimated from self-reported seizure times.
• Forecasting models were validated against electrographic seizures from the Neurovista dataset.

Main Results:

• Seizure likelihood forecasts based on self-reported cycles predicted electrographic seizures in about half the cohort.
• Mobile app forecasts enabled users to spend 67.1% of time in a low-risk state.
• An average of 69.1% of seizures occurred during predicted high-risk periods.

Conclusions:

• Mobile seizure diary apps can generate accurate and clinically relevant personalized seizure forecasts.
• These findings support the potential for prospective clinical trials of seizure forecasting apps.
• Seizure forecasting technology may empower individuals with epilepsy to better manage their condition.