SpikeGUI: software for rapid interictal discharge annotation via template matching and online machine learning

Area of Science:

• Neuroscience and Biomedical Engineering
• Signal Processing and Machine Learning

Background:

• Accurate detection of interictal discharges in electroencephalograms (EEGs) is vital for epilepsy diagnosis and management.
• Current manual interpretation of EEGs is time-intensive, requiring scarce expert neurologists.
• A significant limitation in developing automated spike detectors is the lack of expert-annotated EEG data.

Purpose of the Study:

• To develop a tool that facilitates rapid annotation of interictal discharges for EEG analysis.
• To address the bottleneck of expert time in creating large datasets for automated epilepsy detection.
• To create a generalizable algorithm for waveform and signal type annotation.

Main Methods:

• Development of a graphical user interface named "SpikeGUI" for EEG viewing and annotation.
• Implementation of a custom algorithm combining template matching and online machine learning techniques.
• Focus on accelerating the annotation of interictal epileptiform discharges.

Main Results:

• SpikeGUI significantly enhances the speed of interictal discharge annotation.
• The developed algorithm effectively utilizes template matching and machine learning for efficient annotation.
• The system demonstrates potential for generalization to other signal types beyond epilepsy.

Conclusions:

• SpikeGUI offers a solution to the time-consuming nature of manual EEG annotation.
• The tool facilitates the creation of essential annotated datasets for developing automated spike detectors.
• The underlying algorithm is adaptable for broader applications in signal analysis.