Frequency-domain identification of stereoelectroencephalographic transfer functions for brain tissue classification
Résumé
In drug-resistant epileptic patients, direct electrical brain stimulation sessions may be needed to identify epileptic regions on the brain before surgery. In this context, the classification of the brain tissue in which the electrodes are implanted in is very important for the interpretation of observed responses. This paper presents a new method for tissue classification based on the study of the baseline data with signal processing methods and features estimated from non parametric frequency identification of transfer functions. The method is tested on actual data and compared to visual classification of contacts co-registered on magnetic resonance imaging. The results show that good separability between matters, with up to 80% accuracy, can be achieved when considering the pairing of two consecutive electrodes.