Multi-channel EEG epileptic spike detection by a new method of tensor decomposition

Area of Science:

• Neuroscience
• Biomedical Engineering
• Data Science

Background:

• Epilepsy is a common neurological disorder requiring accurate diagnosis through electroencephalography (EEG).
• Observing epileptic spikes in multi-channel EEG data is crucial for diagnosis and treatment.
• EEG data can be represented as multi-way tensors, suggesting tensor decomposition as a potential analysis tool.

Purpose of the Study:

• To propose and evaluate a novel system for automatic epileptic spike detection using tensor decomposition.
• To investigate the application of simultaneous multilinear low-rank approximation of tensors (SMLRAT) for EEG analysis.
• To extract efficient EEG features for distinguishing epileptic spikes from artifacts.

Main Methods:

• Formulated epileptic spike feature extraction as the SMLRAT problem using multi-way tensors.
• Employed nonnegative GSMLRAT to estimate 'eigenspikes' for feature extraction.
• Compared the proposed tensor analysis and feature extraction methods against conventional and state-of-the-art techniques.

Main Results:

• The proposed SMLRAT-based system achieved high accuracy in detecting epileptic spikes.
• Efficient EEG features were successfully extracted using the developed method.
• The method demonstrated superior performance compared to other common tensor methods and state-of-the-art approaches.

Conclusions:

• This study presents the first method for automatic EEG epileptic spike detection based on tensor decomposition.
• The developed technique offers a practical solution for differentiating epileptic spikes from artifacts in real-world EEG datasets.
• Tensor decomposition provides a powerful framework for analyzing complex EEG data in epilepsy research.