Abstracts

Rationale: Chronic microwire recordings from human epilepsy patients to study single neuron behavior in humans have recently attracted a lot of interest. Spike sorting, the method used to infer the activity of individual neurons from the signal on a microelectrode, is an important intermediate step in the analysis of such data. We observed that signals recorded from commercially available bundles of microwire electrodes may show a large degree of overlap and investigated its impact on spike sorting and downstream analyses. Methods: 7 patients at the Epilepsy Center Freiburg were implanted with macro-micro depth electrode combinations (Ad-Tech) during invasive diagnostic workup in order to determine the seizure onset zone. Microelectrode signals were sampled at 20 kHz with amplifiers (Inomed) working in parallel with the system used for clinical recordings. Signal analysis was carried out in Matlab^®. Results: In several cases, spikes sortable on one of the microwires of an electroce array were detectable also on several other of its wires. Making use of the partial signal overlap on groups of microwires, i.e. treating them as stereotrodes, improved the accuracy of spike sorting. Conclusions: Taking into account signal overlap on bundles of microwires can prevent overestimation of unit numbers and improve the quality of spike sorting, thus improving sensitivity and accuracy of single unit analyses in humans. The origin of the observed effect (comparatively low impedance of the microwires, close proximity of the microwire tips, capacitative coupling along the electrodes and their connections) deserves further investigation. Supported by Deutsche Forschungsgemeinschaft, SFB 780, TP C3, and National Institute of Health, Grant 2R01MH061975-07A2.