Seizure detection with automated EEG analysis: a validation study focusing on periodic patterns

Area of Science:

• Neuroscience
• Medical Technology

Background:

• Automated seizure detection (ASD) algorithms aim to improve EEG analysis.
• Challenging EEG patterns, such as periodic lateralized epileptiform discharges (PLEDs), can complicate automated analysis.

Purpose of the Study:

• To assess the performance of an ASD algorithm in EEGs exhibiting periodic discharges and other difficult patterns.
• To quantify the impact of PLEDs on ASD accuracy.

Main Methods:

• 98 EEG recordings from patients over 1 year old were classified into four groups based on the presence of PLEDs and electrical seizures.
• The Persyst P12 software was used for automated analysis, with results compared to expert neurophysiologist interpretations.
• Key metrics included Positive Percent Agreement (PPA) and False Positive Rate per hour (FPR).

Main Results:

• The ASD algorithm detected 76.1% of 268 seizures across 98 recordings (82.7 hours total).
• Median PPA was 100%, with a median FPR of 0/hour, but performance decreased in recordings with periodic discharges.
• All ictal events were captured in 76.3% of relevant patients.

Conclusions:

• The ASD algorithm shows utility in challenging EEG subsets, but periodic discharges can interfere with its accuracy.
• Further development of ASD techniques is needed to enhance performance and broaden clinical application.