Lossless EEG data source coding for seizure prone activity

Autonomous sensor networks that provide patient monitoring are growing in popularity due to the prospects of lower cost, and the ease of supervision by the physician. Physiological signals monitoring could result in large volume of data being either transmitted or stored which can then be directly related with the energy consumption for the system. This paper presents an EEG compression scheme that is aimed at real-time patient monitoring. It is lossless and incorporates well-known techniques that are computationally easy. A segmentation process that takes advantage of the 50 Hz mains signal is introduced in this work to reduce the entropy of the data stream. High compression gains of 60-66% for both seizure and non-seizure activity are obtained, and a comparison with other high performance lossless EEG compression strategies are presented. The results show that the proposed method performs 2-6% better than a method which directly applies Huffman coding to a DPCM EEG signal.