Affiliations: [a] System Theory Centre (CNR), Department of Biomedical Engineering - Polytechnic of Milan, Milan, Italy | [b]
Department of Neurology, B.S. Raffaele, Milan, Italy | [c] BioMag Laboratory, Medical Engineering Centre, Helsinki University Central Hospital, Helsinki, Finland
Correspondence:
[*]
Corresponding author: F. Grandori, Centro Teoria dei Sistemi (CNR), Dipartimento di Bioingegneria - Politecnico di Milano, Via Ponzio, 34/5, 1-20133 Milan, Italy. Tel. & fax: +3922399 3345.
Abstract: A method for the analysis of variability of EEG signals is described. We examined simulated signals and real EEGs obtained from a normal subject and two epileptic patients. The first step of the method is based on autoregressive (AR) modelling of short EEG epochs. Prediction coefficients of the AR model were computed as a function of time from partially-overlapping moving windows of 2 s duration. The temporal behaviour of these coefficients was analysed to detect variability: quasi-stationary activity causes only smooth changes in the coefficients while variations in the amplitude and/or the frequency content of the signal are shown to produce sharp changes in the coefficients. A segmentation algorithm was developed to detect and quantify with a numerical value (Difference Measure, DM) the AR coefficients variations.
