TU Berlin

Neural Information ProcessingFunctional Imaging Methods: Source Separation Techniques

Neuronale Informationsverarbeitung

Page Content

to Navigation

Functional Imaging Methods: Source Separation Techniques

Lupe

Optical recording techniques have become a widespread technique for measuring the activity of neural populations. Typically, the optical signal is generated by a mixture of sources, where only some of them are related to the optical signal which is correlated with the neural activity of interest. In addition, the signal-to-noise ratio is often very low, in particular, for single condition maps or and / short recording sequences. In this project we investigated whether recently blind source separation techniques (ICA, extended spatial decorrelation, generalized linear models, etc.) can be properly extended to cope with abovementioned challenges. For optical imaging of intrinsic signals we found, that second order methods based on spatial decorrelation algorithms provided the best results. Methods were applied to optical recording data using intrinsic signals as well as to calcium imaging data.

Acknowledgement: Research was funded by DFG, Wellcome Trust, and Technische Universität Berlin.

Selected Publications:

Multi Dimensional ICA to Separate Correlated Sources
Citation key Vollgraf2002
Author Vollgraf, R. and Obermayer, K.
Title of Book Advances in Neural Information Processing Systems 14
Pages 993 – 1000
Year 2002
Address Cambridge, Massachusetts
Publisher MIT Press
Abstract We present a new method for the blind separation of sources, which do not fulfill the independence assumption. In contrast to standard methods we consider groups of neighboring samples (``patches\?\?) within the observed mixtures. First we extract independent features from the observed patches. It turns out that the average dependencies between these features in different sources is in general lower than the dependencies between the amplitudes of different sources. We show that it might be the case that most of the dependencies is carried by only a small number of features. Is this case - provided these features can be identified by some heuristic - we project all patches into the subspace which is orthogonal to the subspace spanned by the ``correlated\?\? features. Standard ICA is then performed on the elements of the transformed patches (for which the independence assumption holds) and robustly yields a good estimate of the mixing matrix.
Bibtex Type of Publication Selected:sources
Link to publication Link to original publication Download Bibtex entry

Navigation

Quick Access

Schnellnavigation zur Seite über Nummerneingabe