Toward Online Data Reduction for Portable Electroencephalography Systems in Epilepsy

Portable EEG units are key tools in epilepsy diagnosis. Current systems could be made physically smaller and longer lasting by the inclusion of online data reduction methods to reduce the power required for storage or transmission of the EEG data. This paper presents a real-time data reduction algorithm based upon the discontinuous recording of the EEG: noninteresting background sections of EEG are discarded online, with only potentially diagnostically interesting sections being saved. MATLAB simulations of the algorithm on an EEG dataset containing 982 expert marked events in 4 days of data show that 90% of events can be correctly recorded while achieving a 50% data reduction. The described algorithm is formulated to have a direct, low power, hardware implementation and similar data reduction strategies could be employed in a range of body-area-network-type applications.