Introduction
Blind EEG analysis, especially in the case of suspected nonconvulsive status epilepticus (NCSE), can lead to misdiagnosis. It is important to consider patient demographic and clinical information when interpreting EEG recordings, which constitutes a limitation of automatic seizure detectors. This study aims to quantify the impact of this information on EEG interpretation in cases of suspected NCSE.
Methods
EEG recordings from patients with suspected NCSE were analysed by 2 independent reviewers, first without and subsequently with access to clinical data. Inter-rater agreement between the reviewers and with the reference standard was analysed by calculating Cohen’s kappa. Diagnoses issued in the blind and informed conditions were compared using performance metrics, and statistically using McNemar’s test.
Results
Our study included 132 EEG recordings from 75 patients. Of these, 56 recordings were classified as NCSE by the physician at the time of assessment. Access to patient clinical data prompted a change in the diagnoses issued by reviewers 1 and 2 in 12% and 17% of recordings, respectively, and resulted in a marked increase in both inter-rater agreement and performance metrics, although these changes were not statistically significant.
Conclusions
EEG patterns can be interpreted differently depending on the clinical context. Access to patient demographic and clinical information improves EEG interpretation.
