Identification of TLE Focus from EEG Signals by Using Deep Learning Approach

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Temporal lobe epilepsy is the most common focal seizure type, necessitating accurate focus localization for treatment.
• Current methods rely on manual EEG analysis, which is time-consuming and subjective.
• Developing automated systems is crucial for efficient and reliable epilepsy diagnosis.

Purpose of the Study:

• To develop and validate a novel deep learning-based computer-aided diagnosis (CAD) system.
• To assist physicians in detecting the epileptic focus from electroencephalogram (EEG) recordings.
• To improve the accuracy and efficiency of epilepsy diagnosis for treatment and surgical planning.

Main Methods:

• Utilized a deep learning framework incorporating Long Short-Term Memory (LSTM) networks.
• Employed discrete wavelet transform (DWT) for extracting EEG subband features.
• Implemented an asymmetry score for epileptic focus identification.
• Validated the algorithm on EEG datasets from Ankara University hospital and the Bonn EEG dataset.

Main Results:

• Achieved high accuracy in classifying ictal and interictal epochs (86.84% on hospital data, 96.67% on Bonn dataset).
• Demonstrated superior performance in epileptic focus identification (96.10% accuracy, 100% sensitivity, 93.80% specificity on hospital data).
• The proposed deep learning algorithm shows significant potential as a medical decision support system.

Conclusions:

• The developed deep learning system effectively detects epileptic epochs and localizes the epileptic focus.
• The CAD system shows promise for clinical applications in epilepsy treatment and surgical planning.
• This technology can serve as a valuable medical decision support tool for neurologists.