Abstracts

Rationale: In clinical practice, the detection of interictal epileptiform discharges in diagnostic overnight EEG is a labor-intensive and time consuming task. In this study, we investigate the usefulness of automatically generated EEG reports that show detected epileptic spikes to aid in the visual clinical interpretation of diagnostic EEG. Methods: Thirty-eight patients had a diagnostic EEG recording of approximately 17h (3pm–8am) at the University Hospital of Geneva, Switzerland. Visual “clinical” interpretation, performed by an expert electrophysiologist (MS), was compared to automatically generated EEG reports (Epilog, Belgium) that visualized the detected spikes in the EEG. The average waveform of the epileptic spike clusters, their topography, the detections over time and 10 examples of single spikes were depicted in the report. A blinded expert electrophysiologist (SV) reviewed the patients solely based on the reports by indicating whether the report showed genuine epileptic spikes in each patient or not. We assessed the correspondence between the labor-intensive visual analysis and the analysis based on the automatically generated reports. Results: Visual interpretation revealed that 12 of the 38 admitted patients had epileptic spikes present in the overnight EEG. The diagnosis based on the report corresponded with the visual analysis in 33 of the 38 patients. The sensitivity of epileptic diagnosis was 75%, the specificity 92%, positive predictive value 82% and negative predictive value 89%, which corresponds to a diagnostic odds ratio of 36. Conclusions: We showed that automatically generated, objective and standardized EEG reports can aid in the visual interpretation of overnight EEG. This can help the neurologists decrease the time spent for visual analysis of the EEG. Adding clinical details to the reports may further increase the sensitivity and specificity to diagnose epilepsy from overnight EEG recordings. Funding: This project received funding from the European Union’s Horizon 2020 research and innovation program under the Marie Sklodowska-Curie grant agreement No 660230.