Leibniz-ScienceCampus

Menu mobile menu

Computer hardware for analysis and modelling of neuronal data from behaving primates

Recent developments of electrodes, recording systems and modern computers have enabled recordings and population analyses of more than 100 neurons in parallel. The availability of these large parallel recording datasets have strengthened the need for a validated analysis package for spiking data including cutting-edge analysis tools, which currently does not exist. In a collaboration between the Cognitive Neuroscience Laboratory and the Neurobiology Laboratory of the German Primate Center (DPZ), we have started developing and testing such analytical tools with the aim of providing an open-source analysis package for spiking data. Many of the emerging analyses require extensive computation power and ideally a processor and graphics card with large-scale parallel processing capabilities. With this Primate Cognition seed fund, we purchased and established an appropriate computer for this purpose and tested and optimized several analytical tools for parallel processing. The currently developed tools can be grouped into three categories (see figure): (1) semi-automatic spike sorting, allowing for sorting of data from more than 100 electrodes in a realistic amount of time, (2) modern single trial state space analyses, allowing extraction of underlying neuronal states from the high-dimensional neuronal population, and (3) single neuron functional connectivity analysis together with the required statistics for more than 100 parallel recorded neurons.

These tools will be made freely available once ready, and will allow researchers at and outside of the DPZ to have quick and easy access to modern validated analysis tools for parallel neuronal recordings, which is an essential requirement for addressing many modern research questions in the field of systems neurophysiology.

Three important parts of the spiking data analysis tools. (Left) Spike sorting of an example electrode resulting in three clusters, as indicated by different colors. The sorter uses a combination of the raw spike waveforms, principal components, and wavelet coefficients for efficient classification. (Middle) Single trial population trajectories of around 100 neurons recoded in parallel while a macaque monkey performed a grasping task. Neuronal activity was decomposed into 6 independent dimensions, with visual, intentional, and movement related information represented by two dimensions each. (Right) Functional connectivity from around 150 neurons recorded in parallel across three different brain regions of one macaque monkey. (Copyright – Benjamin Dann).

Project leaders

Hans Scherberger +49 551 3851-494 Contact Profile

Suresh Krishna +49 551 3851-354 Contact Profile

Benjamin Dann +49 551 3851-484 Contact Profile