Effective epileptic seizure detection by using level-crossing EEG sampling sub-bands statistical features selection...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neurology

Background:

• Mobile healthcare is advancing with cloud-connected biomedical implants.
• Effective and automated epileptic seizure diagnosis is crucial.

Purpose of the Study:

• To propose an innovative method for automated epileptic seizure diagnosis.
• To reduce computational complexity and transmission activity in real-time.

Main Methods:

• Utilizing a level-crossing analog-to-digital converter (LCADC) for electroencephalogram (EEG) signal acquisition.
• Employing an activity selection algorithm (ASA) for segment selection.
• Implementing adaptive-rate processing including denoising, wavelet decomposition, and dimension reduction.

Main Results:

• Achieved a 4.1-fold and 3.7-fold decline in samples for University of Bonn and Hauz Khas datasets, respectively.
• Reduced computational complexity by over 14-fold.
• Demonstrated 100% classification accuracy in most epilepsy detection cases.

Conclusions:

• The proposed method offers significant reductions in computational load, power, and bandwidth.
• It enables efficient real-time epilepsy diagnosis via cloud-connected implants.
• High accuracy supports its potential in mobile healthcare applications.