Regensburg 2016 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

BP: Fachverband Biologische Physik

BP 22: Posters - Neurosciences

BP 22.3: Poster

Montag, 7. März 2016, 17:30–19:30, Poster C

Analyzing and modeling dynamics of cortical steady state responses to long lasting stimuli — •Patrick Krauss^1,2, Achim Schilling¹, Konstantin Tziridis¹, Claus Metzner², and Holger Schulze¹ — ¹Experimental Otolaryngology, ENT-Hospital, University of Erlangen — ²Department of Physics, Biophysics Group, University of Erlangen

We present a novel method for analyzing and modeling high-dimensional data such as multichannel cortical recordings, which is derived from multidimensional scaling (MDS). A fundamental shortcoming of classical MDS is the impossibility of assigning coordinates in target space to new points without re-running the entire scaling procedure. To overcome this problem we construct a mapping matrix M from high-dimensional state space to target space while preserving all mutual Euclidean distances. We use our method to reveal the relation between auditory perception and neuronal activity. The temporal development of the spatial activity pattern across the recording channels corresponds to a trajectory in a high-dimensional state space. Projecting trajectories with the matrix M reveals attractor-like dynamics. Remarkably, this finding remains undiscovered when performing other dimensionality reduction methods such as PCA or ICA. In addition, we use an animal model to induce tinnitus. Our method enables inferring the pitch of the tinnitus precept from recorded neuronal data, which we validate using a behavioral tinnitus assessment paradigm. Finally, inverting the matrix M results in a simple generator model of stimulus specific attractor dynamics.