Log in
Forgotten password?
Abstracts
| Home |
Program |
Venue |
Registration |
Contact |
Go back to program.
Ingo Bojak: Building blocks for realistic mean field modeling
Mean field models of neural mass activity are attractive for modeling macroscopic neuroimaging data, since on one hand their level of complexity remains manageable and on the other hand non-invasive recordings detect only the coherent "mass action" of neuronal populations anyhow. However, a range of assumptions and simplifications must be made in order to construct computable models that describe real world data well. I will discuss some of these, mostly by focusing on a large scale model of cortical activity embedded in a realistic head model, which we have been developing. Many of the problems we have encountered seem typical, rather than specific to our model, and hence one can reasonably hope that we are currently creating a wheel that need not be re-invented much in future. I will also touch on some related recent works, and will discuss the case of modeling axonal conduction speeds as a typical example of how the field is developing towards greater realism in its underlying assumptions.
Stephen Coombes: Neural field models
The tools of dynamical systems theory are having an increasing impact on our understanding of patterns of neural activity. In this talk I will describe how to build tractable tissue level models that maintain a strong link with biophysical reality. These models typically take the form of nonlinear integro-differential equations. Their non-local nature has led to the development of a set of analytical and numerical tools for the study of waves, bumps and patterns, based around natural extensions of those used for local differential equation models. Here I will present an overview of these techniques.
Shaul Druckmann: From manual parameter tuning to automated parameter constraining of biophysical models
The existence of a diversity of electrical classes of cortical neurons is widely agreed upon. Classes such as “regular spiking”, “fast-spiking”, “stutterers”, “bursters” were identified based mainly on the response of neurons to application of depolarizing current steps. If this diversity is to be explored with detailed biophysical models, a rapid and robust method of constraining the free parameters of such models must be developed and utilized.
The feature-based MOO (“multiple objective optimization”) method recently developed by us utilizes ‘spiking features’ such as spike width, time-to-first-spike, spike frequency, degree of “burstiness” in order to automatically generate experimentally-constrained compartmental models of the different electrical neuron types. We discuss the implications of being able to automatically constrain models as opposed to manually tuning their parameters.
We analyzed the effect of different stimuli on the parameter constraining process showing that as more stimuli are added to the set of experimental responses used for constraining the model, the average error during model construction (training error) increases yet the average discrepancy between model and experimental response to stimuli not used during the training process (generalization error) decreases. In addition, we examine the relative utility of different stimuli in the parameter constraining process. We show that compartmental models that were constrained only by simple stimuli such as step and ramp currents are capable of predicting the timing of spikes in response to noisy current injections emulating barrages of synaptic activity with a very high degree of accuracy (approximately 95% of experimental reliability). Interestingly, we demonstrate that generalization can be asymmetric - models constrained on one stimulus generalizing very well to another, but not vice versa. We discuss possible reasons and implications of this finding.
Rolf Kötter: The New CoCoMac: A brief history of tracer studies in macaque brains and the use of graphical interactions with connectivity databases
I will highlight the benefits of CoCoMac, our online database of primate brain connectivity (cocomac.org), in the context of research and training in systems biology. CoCoMac was started in 1997 by collation and comparison of published results from the growing body of tracer studies detailing the connectivity in brains of macaques (Stephan et al. 2001; Kötter 2004). Our ultimate aim is to build a large-scale circuit wiring diagram of the primate brain that might successfully predict some aspects of (human) brain activity and the effects of lesions (Sporns et al. 2005).
CoCoMac references NCBI’s PubMed for registered journal publications providing the abstract and full bibliographic information at a button click. Conversely, PubMed users can configure the NCBI interface to display available LinkOut to CoCoMac, which adds value by providing supplementary connectivity and mapping information through online retrieval from CoCoMac.
Bidirectional cross-links have also been implemented with BrainInfo (http://braininfo.rprc.washington.edu/), a comprehensive information system on nomenclature, location and fine structure of mammalian brain regions. A single mouse click in BrainInfo retrieves corresponding wiring information from CoCoMac.
Interfaces with Caret and the SuMS (Van Essen et al. 2001) link brain mapping and connectivity data with their spatial representations in monkey and human hemispheres. WebCaret users have direct access to CoCoMac data relevant to the selected brain regions displayed in separate windows. The newly introduced Regional Map (RM) in CoCoMac (Kötter and Wanke 2005) provides a coarse parcellation scheme of cerebral cortex taking into account a combination of microstructural, functional and topographic features. It provides an intuitive and rather uncontroversial naming scheme applicable to human and macaque cerebral cortex. The coordinate-independent mapping of RM areas to familiar partitioning schemes (e.g., Brodmann, Bonin & Bailey, Van Essen) in CoCoMac allowed us to generate a spatial representation of RM on a cortical surface template using the Caret software. Thereby we link the large body of coordinate-independent tracing results to a spatially registered macaque brain and – using existing spatial deformation tools – to the results of neuroimaging studies in humans.
Most recently we have created a 3D visualization software based on drawings from the stereotaxic macaque atlas published by Paxinos et al. (2000; 2008). The software links the atlas to mapping and connectivity information in CoCoMac and provides guidance for experimentalists including the option to dissect and manipulate individual brain structures in the cerebral cortex, thalamus and amygdala. An online version of this tool can be viewed at http://scalablebrainatlas.incf.org
Most recent literature:
Deco G., Jirsa V.K., McIntosh A.R., Sporns O., Kötter R. (2009) The key role of coupling, delay and noise in resting brain fluctuations. Proc. Natl. Acad. Sci. USA, 106: 10302-10307.
Reid A.T., Krumnack A., Wanke E., Kötter R. (2009) Optimization of cortical hierarchies with continuous scales and ranges. NeuroImage, doi:10.1016/j.neuroimage.2009.04.061.
Bohland J.W., et al. (2009) A proposal for a coordinated effort for the determination of brainwide neuroanatomical connectivity in model organisms at a mesoscopic scale. PLoS Comput. Biol., 5: e1000334.
Anders Lansner: Modularization dramatically increases stability of oscillating attractor networks
Attractor neural networks are thought to underlie working memory functions in the cerebral cortex. Several such models have been proposed that successfully reproduce firing properties of neurons recorded from monkeys performing working memory tasks. However, the regular temporal structure of spike trains in these models is often incompatible with experimental data. Here, we show that the in vivo observations of bistable activity with irregular firing at the single cell level can be achieved in a large-scale network model with a modular structure in terms of hypercolumns. Despite high irregularity of individual spike trains, the model shows population oscillations in the beta and gamma band in ground and active states respectively. Irregular firing typically emerges in a high-conductance regime of balanced excitation and inhibition. Population oscillations can produce such a regime, but in previous models only a non-coding ground state was oscillatory. Due to the modular network structure comprising several connected hypercolumns, the oscillatory regime is much more stable in our network and also the active state is oscillatory. The model therefore maintains oscillatory and irregular firing also in the memory retrieval state without fine-tuning. It provides a novel mechanistic view of how irregular firing emerges in cortical populations as they go from beta to gamma oscillations during memory retrieval.
