Magnetoencephalography for epileptic focus localization based on Tucker decomposition with ripple window

Area of Science:

• Neuroimaging
• Epilepsy Research
• Biomedical Engineering

Background:

• Magnetoencephalography (MEG) is crucial for localizing epileptic sources.
• Current methods face limitations in spatial localization precision.
• Accurate source localization is vital for effective presurgical evaluation.

Purpose of the Study:

• To enhance the spatial localization precision of focal epileptic sources using MEG.
• To introduce a novel tensor-based estimation method for MEG data.
• To compare the proposed method's performance against existing techniques.

Main Methods:

• Applied Tucker decomposition with Higher-order orthogonal iteration (HOOI) to reduce MEG data dimensionality.
• Utilized linearly constrained minimum variance (LCMV) for inverse problem solving.
• Compared the proposed method with DICS, MUSIC, and dipole-fitting on simulated and real clinical MEG data from 31 epilepsy patients, focusing on ripple (80-250 Hz) and spike windows.

Main Results:

• The proposed Tucker estimation method demonstrated superior positional accuracy compared to LCMV, DICS, and MUSIC for simulated data.
• For real clinical data, the proposed method outperformed dipole-fitting for both ripple and spike windows.
• Ripple window analysis yielded higher positional accuracy than spike window analysis, irrespective of the source localization method used.

Conclusions:

• Tucker estimation of MEG data using ripple windows is a promising technique for improving source imaging.
• This approach offers enhanced accuracy for the presurgical evaluation of epilepsy.
• The method shows effective performance for both shallow and deep epileptic sources.