Evaluating the accuracy of monitoring seizure cycles with seizure diaries.

Objective: Epileptic seizures occurring in cyclical patterns is increasingly recognized as a significant opportunity to advance epilepsy management. Current methods for detecting seizure cycles rely on intrusive techniques or specialized biomarkers, thereby limiting their accessibility. This study evaluates a non-invasive seizure cycle detection method using seizure diaries and compares its accuracy with cycles identified from intracranial electroencephalography (iEEG) seizures and interictal epileptiform discharges (IEDs).

Methods: Using data from a previously published first in-human iEEG device trial (n = 10), we analyzed seizure cycles identified through diary reports, iEEG seizures, and IEDs. Cycle similarities across diary reports, iEEG seizures, and IEDs were evaluated at periods of 1 to 45 days using spectral coherence, accuracy, precision, recall, and the false-positive rate.

Results: A spectral coherence analysis of the raw signals showed moderately similar periodic components between diary seizures/day and iEEG seizures/day (median = .43, IQR = .68). In contrast, there was low coherence between diary seizures/day and IEDs/day (median = .11, IQR = .18) and iEEG seizures/day and IEDs/day (median = .12, IQR = .19). Accuracy, precision, recall scores, and false-positive rates of iEEG seizure cycles from diary seizure cycles were significantly higher than chance across all participants (accuracy (mean ± standard deviation): .95 ± .02; precision: .56 ± .19; recall: .56 ± .19; false-positive rate: .02 ± .01). However, accuracy, precision, and recall scores of IED cycles from both diary and iEEG cycles did not perform above chance, on average. Recall scores were compared across good diary reporters, under-reporters, and over-reporters, with recall scores generally performing better in good reporters and under-reporters compared to over-reporters.

Significance: These findings suggest that iEEG seizure cycles can be identified with diary reports, even in individuals who under- and over-report seizures. This approach offers an accessible alternative for monitoring seizure cycles compared to more invasive methods.