Seizure states identification in experimental epilepsy using gabor atom analysis.

Area of Science:

• Computational Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Epileptic seizures progress through distinct states with changing brain signal characteristics.
• Distinguishing these seizure states from time-series data is challenging due to signal similarities.

Purpose of the Study:

• To develop a classifier for identifying epileptic seizure states.
• To utilize time-frequency features from short signal segments for state identification.

Main Methods:

• Brain signals from intracranial recordings of the Kindling model were analyzed.
• Short signal segments were decomposed into atoms to determine Gabor atom density.
• A classifier was built based on Gabor atom density and epoch energy.

Main Results:

• Short signal segments contain sufficient information for seizure state classification.
• The classifier achieved high sensitivity (0.82-0.97) and AUC (0.5-0.9), particularly for seizures from the same subject.
• Gabor atom density effectively reveals seizure states.

Conclusions:

• Gabor atom density is a viable feature for classifying epileptic seizure states.
• Combining Gabor atom density with epoch energy enhances classifier performance.
• The findings support the potential for automatic seizure state classification.