Inter-ictal spike detection using a database of smart templates

Area of Science:

• Neuroscience
• Medical Technology
• Signal Processing

Background:

• Visual analysis of electroencephalography (EEG) is time-intensive and prone to inter-observer variability.
• Assisted automated analysis can streamline EEG interpretation by providing consistent feedback and summarizing key findings.

Purpose of the Study:

• To design an accurate and robust automated system for detecting inter-ictal epileptiform discharges (IEDs) in scalp EEG.
• To reduce the time and inter-observer variability associated with manual EEG analysis.

Main Methods:

• Extracted IED templates from EEG training data, enabling learning through time-shifted correlation.
• Employed classifiers trained on true and false detections to improve prediction accuracy.
• Developed a detection system where trained templates identify IEDs in new EEG data, grouping overlapping detections and assigning certainty values.

Main Results:

• Achieved sensitivities up to 0.99 with 7.24 false positives per minute using 2160 templates on a dataset of 241 IEDs.
• Higher certainty thresholds resulted in a mean sensitivity of 0.90 with 2.36 false positives per minute, reducing false detections.

Conclusions:

• The template-based approach is robust to variations in spike morphology.
• A certainty value for each detection simplifies the EEG review process and enhances efficiency.
• Automated spike detection can significantly assist in visual EEG interpretation, potentially leading to faster review times.