An algorithm for seizure detection in rodents

Area of Science:

• Neuroscience
• Artificial Intelligence
• Biomedical Engineering

Background:

• Epilepsy research heavily utilizes long-term intracranial electroencephalographic (iEEG) recordings in animal models.
• Manual inspection of iEEG data for seizure detection is time-consuming and labor-intensive.
• Current methods lack efficiency in analyzing large datasets of rodent epilepsy research.

Purpose of the Study:

• To develop and validate an artificial neural network (ANN) algorithm for the automatic detection of seizures in rodent epilepsy models.
• To provide an efficient and reliable tool for analyzing iEEG data in preclinical epilepsy research.
• To streamline the process of seizure frequency assessment in animal studies.

Main Methods:

• Trained an ANN algorithm on iEEG recordings from three mouse models: pilocarpine-induced, albumin-induced, and synapsin triple knockout (STKO).
• Applied signal filtering, segmentation, and feature extraction to iEEG data for ANN classification.
• Developed a graphical user interface for simplified data analysis and seizure detection, validated on over 2800 hours of recordings from 15 animals.

Main Results:

• The developed ANN system achieved high performance with a sensitivity and positive predictive value exceeding 98%.
• The algorithm demonstrated robust performance across diverse epilepsy models, including status epilepticus and post-traumatic epilepsy.
• The system has been successfully employed since 2010 in numerous studies investigating seizure frequency.

Conclusions:

• The proposed ANN algorithm offers a reliable and efficient method for automated seizure detection in mice and rats.
• This AI-driven approach significantly enhances the analysis of iEEG data in epilepsy research.
• The software facilitates more accurate and scalable studies of seizure phenomena in preclinical models.