MER Classification for Deep Brain Stimulation Peter Gemmar, Oliver Gronz, Thorsten Henrichs, Frank Hertel † , Christian Decker ‡ Institute for Innovative Informatics-Applications i3A, University of Applied Sciences (FH) Trier, Germany * † Hospital Idar-Oberstein, Germany ‡ Hospital of the Merciful Brethren, Trier, Germany Preprint: Sixth Heidelberg Innovation Forum, April 15, 2008 Abstract Deep brain stimulation (DBS) of the subthalamic nu- cleus (STN) has become a common procedure to alle- viate the symptoms of advanced Parkinson’s disease. To estimate the optimal site for placement of the defi- nite electrode, up to five microelectrodes are inserted at first and the neuronal activity at the electrode tip is recorded. These microelectrode recordings (MER) are classified to STN or non-STN signals manually by the surgeon, which requires experience and time. A system has been developed for automatic clas- sification of MER signals. The classifier consists of three levels, each of which using a specific criterion to decide whether a MER is STN signal or not. In the first level, the background activity is examined and those recordings showing increased activity are marked. The second level uses the bursty or irreg- ular behavior of typical STN single cell activity for taking decisions. In the last level, the spike rate of duplicated intervals resulting from level 1 and 2 is examined. Results from all levels are combined and thus STN signals are identified. To enhance the evaluation of the different charac- teristics, signal preprocessing is performed in level 2 and level 3. Wavelet transformation is used to re- move background activity (noise) and a multilevel 1- * Contact: Prof. Dr. Peter Gemmar, Fachhochschule Trier, Schneidershof, 54293 Trier, Germany; p.gemmar@fh-trier.de d wavelet decomposition is used to extract certain properties of the signals. The classifier has been tested using 2434 MERs taken from 14 patients. Nearly 95% of the classifi- cations matched with a specialist’s decision. The re- maining differences were mainly due to signals lack- ing distinctive characteristics, especially signals ex- tracted near the border of STN. The classifier will support the surgeon and make the decision process for the final electrode position more reliable and less time consuming. It can easily be adapted for the classification of other functional neural areas than the STN. Keywords: DBS, STN, MER, signal classification, wavelet transformation, Parkinson’s disease 1 Introduction Stereotactic deep brain stimulation is a widespread treatment option for different kinds of neurologi- cal diseases, especially movement disorders, such as Parkinson´s disease (PD), Dystonia, different kinds of tremors, or chronic pain also ([1],[2]). In the treat- ment of advanced PD the STN is considered the most promising target. In this procedure, electrodes are implanted permanently in the patient’s STNs. They emit signals that reduce the effect of the chronic hy- peractivity of STN. Especially for long-term patients, 1