Analysis and Classifikation of multivariate neuronal responses: Spikes and local-field-potentials

H.Haan, Institut für Biophysik, Arbeitsgruppe Hirnforschung
Prof. Dr. J. Krüger, Institut für Biophysik, Arbeitsgruppe Hirnforschung
Dr. J. Timmer, FDM, Fakultät für Physik
Prof. Dr. J. Honerkamp, FDM, Fakultät für Physik

Previous work:

We analysed multivariate pointprocesses - multineuronal response-patterns from visual cortex of anaesthetized monkeys by segmentation, characterization und classifikation of the data with classical methods, Hidden-Markov-Models (HMM) and explorative investigation of structures in the high-dimensional data by multidimensional scaling (MDS). Ws tried to identify parameters of the vis. stimuli by investigation of the path of HMM-transition-matrices (1). Structure in multivariate neuronal response patterns (objects) is hidden in high-dimensional space. The explorative tool MDS is used to visualize this structure by embedding the object's dissimilarity to 3 dimensions, where the significant structure is visible. Different models of MDS are used.

Recent work:

MDS is used to explore structure in spike-pointprocesses. The structure of the scaled data is compared with the topology of the shown vis. stimuli (2). Spikes and local field potentials are recorded at the same electrode. These data are segmented and classified. The work with awake animals enables us to answer interesting questions, but these circumstances make the work and the data analysis difficult. Questions are: how can the eye-saccades be extracted from the data. Is it possible to find and classify suspicious data-segments and can these data segments be connected to visual stimuli or brain states. One animal showed strong oscillations in local field potentials. After segmentation and characterization a relationship to the color of certain vis. stimuli was established (3). The relationship between local field potentials and spikes is investigated.

Publications:

(1)
G. Radons, J. D. Becker, B. Dülfer, J. Krüger:
Analysis, classification and coding of multielectrode spike trains with Hidden Markow Models. Biol. Cybern. 71: 359-373 (1994)
(2)
H. Haan, J. Timmer, J. Krüger:
Multineuronal Spike Patterns: Visualization of relationships by Multidimensional Scaling, in N. Elsner, R. Menzel (Hrsg.) Learning and Memory. Georg Thieme Verlag p. 475 (1995)
(3)
J. Krüger, I. Bondar, H. Haan:
Recording from foveal striate cortex while the monkey is looking at the stimulus. in B. Kappen, S. Gielen (eds.) Neural networks: artificial intelligence and industrial applications. Springer, Berlin Heidelberg New York (1995)


Author: Hubertus Haan, haan@ruf.uni-freiburg.de 10.6.96
back to:
Homepage FDM
Homepage of Brain Research Unit