Assessment of a computer program to detect epileptiform spikes.

Area of Science:

• Neuroscience
• Medical Technology

Background:

• Automated detection of epileptiform discharges in electroencephalography (EEG) is crucial for clinical analysis.
• Human electroencephalographers (EEGers) vary in expertise, impacting the reliability of manual spike detection.

Purpose of the Study:

• To compare the sensitivity and specificity of an automated spike detection program against a group of human EEGers.
• To evaluate the effectiveness of an automated system as a screening tool for large-scale EEG monitoring data.

Main Methods:

• An automated spike detection program was developed and compared to 6 electroencephalographers.
• An expertise factor was devised to weight EEGer scoring based on experience.
• Event detection performance was analyzed at different amplitude thresholds for both the computer and individual EEGers.

Main Results:

• The automated system demonstrated comparable sensitivity to EEGers, particularly at lower amplitude thresholds.
• Higher-rated EEGers identified more subtle events, while the computer's sensitivity increased linearly with decreased thresholds.
• While false positive detections increased with lower thresholds, they were not found to be significantly distracting.

Conclusions:

• The automated computer system is a reliable screening editor for large volumes of EEG monitoring data.
• The system can achieve comparable sensitivity to human EEGers, though with lower specificity.
• The automated program is more effective than an individual EEGer for the limited purpose of sensitive screening.