Christian-Albrechts-Universität zu Kiel

The IEEE Brain Initiative presents The Second Place Award to Team University of Kiel at the IEEE Brain Data Bank Challenge, partnering with IEEE Sensors conference 2019. The interdisciplinary team of re-searchers from University of Kiel, formed by Matthias Schneider, Maria Avitha Francis, Hendrik Lehmann, Igor Barg and Andreas Bahr presented an innovative project on epileptic seizure detection with neural net-works for medical implants.

An epileptic seizure is an abnormal activity of the brain caused by a rapid discharge of large groups of neurons. The idea is to treat a seizure promptly by detecting this and supply local medication in time. A medical implant records the biopotential signals, which are then evaluated with the help of a convolutional neural network (CNN), and in the event of a seizure being detected, an electrochemical pump is activated, which ensures local medication. The CNN is trained offline with approximately 40-50 hours of EEG (electroencephalogram) data and is implemented on a low-power micro-controller that fits into the medical implant. A possible implementation of such a network on a RISC-V microprocessor will also be shown.

This work was supported by DFG research projects Materials for Brain and the CRC 1261. The DFG-funded Research Training Group 2154 ”Materials for Brain” investigates nano- and microscale therapeutically active coatings for their suitability as implants to treat three selected diseases of the brain.