Early Seizure Detection by Applying Frequency-Based Algorithm Derived from the Principal Component Analysis.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Intractable epilepsy treatment is advancing with automatic electrical stimulation triggered by early seizure detection.
• Enhancing the accuracy of early seizure detection is critical for the success of these novel treatments.

Purpose of the Study:

• To propose and validate a frequency-based algorithm using principal component analysis (PCA) for improved early seizure detection.
• To evaluate the efficacy of this PCA-derived feature in a pilocarpine-induced epilepsy rat model.

Main Methods:

• Principal component analysis (PCA) was applied to the covariance matrix of electroencephalograph (EEG) frequency band signals from epileptic rats.
• A PCA-based feature derived from the initial 5 seconds of seizure onset was compared against features from the whole seizure segment and six other conventional features.
• Performance was assessed using False Positive (FP), False Negative (FN), and Latency (Lat) metrics on a testing set.

Main Results:

• The PCA-based feature from the initial seizure segment demonstrated superior performance with 1.40% FP, 0% FN, and 0.14 s Latency.
• This method significantly outperformed other tested features in early seizure detection accuracy.
• The proposed feature effectively captured the characteristics of the initial seizure phase.

Conclusions:

• The proposed frequency-based feature derived from PCA is effective for accurate and early seizure detection.
• Applying PCA to the initial 5-second segment of seizure onset in rat EEGs improves detection rates compared to using the whole segment or conventional methods.