| Name |
Oscar Herreras |
| Country |
Spain |
| Email |
herreras@cajal.csic.es |
| Co-Author(s) |
O. Herreras, V. Makarov, G. Mart\'in V\'azquez, and J. Makarova |
| Submit Time |
2014-03-18 10:41:17 |
Session |
Special Session 77: Theoretical, technical, and experimental challenges in closed-loop approaches in biology
|
|
| Contents |
| The unpredictable discontinuous activation of electrical sources in the brain plus their mixing in the volume constitute major handicaps for analysis based on averaging or frequency-decomposition. Indeed, none captures the natural temporal dynamics of current sources contained in local field potentials (LFPs). Since sources are static, they are amenable to disentanglement through blind-source separation techniques. We earlier combined spatial independent component analysis (sICA) and hierarchical clustering to segregate intracerebral sources into spatially coherent groups that represent pathway-specific activations of target populations. The spatial and temporal mixings influence the efficiency of the separation, which can be strongly optimized by a priori knowledge of the sources features. Thus, we used known mixtures of sources from real experiments to reproduce spatiotemporal fluctuations of LFPs in silico through a realistic multineuronal multicompartmental model. Virtual LFPs were essayed to check the advantages of using PCA prior to ICA. The flexibility of this approach allows the repeated modification of the initial conditions for data choice and pretreatment in an experience-guided cyclic process. We also present hints to increase the relative variance of weak sources and to reduce cross-contamination. |
|