Biomedical subjects

Klaus Obermayer

Publications and source records attributed to Klaus Obermayer.

At least 19 recordsLinked to original sources

PubMed2006

The role of feedback in shaping the extra-classical receptive field of cortical neurons: a recurrent network model.

The responses of neurons in sensory cortices are affected by the spatial context within which stimuli are embedded. In the primary visual cortex (V1), orientation-selective responses to stimuli in the receptive field (RF) center are suppressed by similarly oriented stimuli in the RF surround. Surround suppression, a likely neural correlate of perceptual figure-ground segregation, is traditionally thought to be generated within V1 by long-range horizontal connections. Recently however, it has been shown that these connections are too short and too slow to mediate fast suppression from distant regions of the RF surround. We use an anatomically and physiologically constrained recurrent network model of macaque V1 to show how interareal feedback connections, which are faster and longer-range than horizontal connections, can generate "far" surround suppression. We provide a novel solution to the puzzle of how surround suppression can arise from excitatory feedback axons contacting predominantly excitatory neurons in V1. The basic mechanism involves divergent feedback connections from the far surround targeting pyramidal neurons sending monosynaptic horizontal connections to excitatory and inhibitory neurons in the RF center. One of several predictions of our model is that the "suppressive far surround" is not always suppressive, but can facilitate the response of the RF center, depending on the amount of excitatory drive to the local inhibitors. Our model provides a general mechanism of how top-down feedback signals directly contribute to generating cortical neuron responses to simple sensory stimuli.

Action Potentials↗

PubMed2006

A new summarization method for Affymetrix probe level data.

MOTIVATION: We propose a new model-based technique for summarizing high-density oligonucleotide array data at probe level for Affymetrix GeneChips. The new summarization method is based on a factor analysis model for which a Bayesian maximum a posteriori method optimizes the model parameters under the assumption of Gaussian measurement noise. Thereafter, the RNA concentration is estimated from the model. In contrast to previous methods our new method called 'Factor Analysis for Robust Microarray Summarization (FARMS)' supplies both P-values indicating interesting information and signal intensity values. RESULTS: We compare FARMS on Affymetrix's spike-in and Gene Logic's dilution data to established algorithms like Affymetrix Microarray Suite (MAS) 5.0, Model Based Expression Index (MBEI), Robust Multi-array Average (RMA). Further, we compared FARMS with 43 other methods via the 'Affycomp II' competition. The experimental results show that FARMS with default parameters outperforms previous methods if both sensitivity and specificity are simultaneously considered by the area under the receiver operating curve (AUC). We measured two quantities through the AUC: correctly detected expression changes versus wrongly detected (fold change) and correctly detected significantly different expressed genes in two sets of arrays versus wrongly detected (P-value). Furthermore FARMS is computationally less expensive then RMA, MAS and MBEI. AVAILABILITY: The FARMS R package is available from http://www.bioinf.jku.at/software/farms/farms.html. SUPPLEMENTARY INFORMATION: http://www.bioinf.jku.at/publications/papers/farms/supplementary.ps

Algorithms↗

PubMed2006

A cross-spectrum weighting algorithm for speech enhancement and array processing: combining phase-shift information and stationary signal properties.

In this paper, a gain function for noise cancellation with a two-channel microphone array is presented. This gain function combines ideas from one- and multichannel algorithms. It is developed using a minimum mean square error estimator for the amplitude of the speech signal from the cross spectrum between two microphone signals. To consider speech pauses and the absence of spectral components of the speech, an extension of this gain function is presented. The performance of the overall gain function is shown in terms of the cancellation of (diffuse) driving noise as well as the cancellation of an interfering speech signal, both recorded in a car.

Algorithms↗

PubMed2006

Support vector machines for dyadic data.

We describe a new technique for the analysis of dyadic data, where two sets of objects (row and column objects) are characterized by a matrix of numerical values that describe their mutual relationships. The new technique, called potential support vector machine (P-SVM), is a large-margin method for the construction of classifiers and regression functions for the column objects. Contrary to standard support vector machine approaches, the P-SVM minimizes a scale-invariant capacity measure and requires a new set of constraints. As a result, the P-SVM method leads to a usually sparse expansion of the classification and regression functions in terms of the row rather than the column objects and can handle data and kernel matrices that are neither positive definite nor square. We then describe two complementary regularization schemes. The first scheme improves generalization performance for classification and regression tasks; the second scheme leads to the selection of a small, informative set of row support objects and can be applied to feature selection. Benchmarks for classification, regression, and feature selection tasks are performed with toy data as well as with several real-world data sets. The results show that the new method is at least competitive with but often performs better than the benchmarked standard methods for standard vectorial as well as true dyadic data sets. In addition, a theoretical justification is provided for the new approach.

Artificial Intelligence↗

PubMed2005

Learning top-down gain control of feature selectivity in a recurrent network model of a visual cortical area.

We propose that the effects of attentional top-down modulations observed in the visual cortex reflect the simple strategy of strengthening currently relevant pathways in a task-dependent manner. To exemplify this idea, we set up a network model of a visual area and simulate the learning of a context-dependent 'go/no-go'-task. The model learns top-down gain-modulations of sensory representations based on reinforcements received from the environment. We also discuss how this idea relates to alternative interpretations like optimal coding hypotheses.

Adaptation, Physiological↗

PubMed2005

Adaptivity of tuning functions in a generic recurrent network model of a cortical hypercolumn.

The representation of orientation information in the adult visual cortex is plastic as exemplified by phenomena such as perceptual learning or attention. Although these phenomena operate on different time scales and give rise to different changes in the response properties of neurons, both lead to an improvement in visual discrimination or detection tasks. If, however, optimal performance is indeed the goal, the question arises as to why the changes in neuronal response properties are so different. Here, we hypothesize that these differences arise naturally if optimal performance is achieved by means of different mechanisms. To evaluate this hypothesis, we set up a recurrent network model of a visual cortical hypercolumn and asked how each of four different parameter sets (strength of afferent and recurrent synapses, neuronal gains, and additive background inputs) must be changed to optimally improve the encoding accuracy of a particular set of visual stimuli. We find that the predicted changes in the population responses and the tuning functions were different for each set of parameters, hence were strongly dependent on the plasticity mechanism that was operative. An optimal change in the strength of the recurrent connections, for example, led to changes in the response properties that are similar to the changes observed in perceptual learning experiments. An optimal change in the neuronal gains led to changes mimicking neural effects of attention. Assuming the validity of the optimal encoding hypothesis, these model predictions can be used to disentangle the mechanisms of perceptual learning, attention, and other adaptation phenomena.

Animals↗

PubMed2005

Adaptation at synaptic connections to layer 2/3 pyramidal cells in rat visual cortex.

Neocortical synapses express differential dynamic properties. When activated at high frequencies, the amplitudes of the subsequent postsynaptic responses may increase or decrease, depending on the stimulation frequency and on the properties of that particular synapse. Changes in the synaptic dynamics can dramatically affect the communication between nerve cells. Motivated by this question, we studied dynamic properties at synapses to layer 2/3 pyramidal cells with intracellular recordings in slices of rat visual cortex. Synaptic responses were evoked by trains of test stimuli, which consisted of 10 pulses at different frequencies (5-40 Hz). Test stimulation was applied either without any adaptation (control) or 2 s after an adaptation stimulus, which consisted of 4 s stimulation of these same synapses at 10, 25, or 40 Hz. The synaptic parameters were then assessed from fitting the data with a model of synaptic dynamics. Our estimates of the synaptic parameters in control, without adaptation are broadly consistent with previous studies. Adaptation led to pronounced changes of synaptic transmission. After adaptation, the amplitude of the response to the first pulse in the test train decreased for several seconds and then recovered back to the control level with a time constant of 2-18 s. Analysis of the data with extended models, which include interaction between different pools of synaptic vesicles, suggests that the decrease of the response amplitude was due to a synergistic action of two factors, decrease of the release probability and depletion of the available transmitter. After a weak (10 Hz) adaptation, the decrease of the response amplitude was accompanied by and correlated with the decrease of the release probability. After a strong adaptation (25 or 40 Hz), the depletion of synaptic resources was the main cause for the reduced response amplitude. Adaptation also led to pronounced changes of the time constants of facilitation and recovery, however, these changes were not uniform in all synapses, and on the population level, the only consistent and significant effect was an acceleration of the recovery after a strong adaptation. Taken together, our results suggest, that apart from decreasing the amplitude of postsynaptic responses, adaptation may produce synapse-specific effects, which could result in a kind of re-distribution of activity within neural networks.

Adaptation, Physiological↗

PubMed2005

Detection of pulses in a colored noise setting.

Cortical neurons are exposed to a considerable amount of synaptic background activity, which increases the neurons' conductance and which leads to a fluctuating membrane potential. Here we investigate how the presence and the properties of this background noise influence the ability of a neuron to detect transient inputs, a task that is important for coincidence detection as well as for the detection of synchronous spiking events in a neural system. Using a leaky integrate-and-fire neuron as well as a biologically more realistic Hodgkin-Huxley type point neuron we find that noise enhances the detection of subthreshold input pulses and that the phenomenon of stochastic resonance occurs. When the noise is colored, pulse detection becomes more robust, because the number of false positive events decreases with increasing temporal correlation while the number of correctly detected events is almost unaffected. Therefore, the optimal variance of the noise also changes with the degree of temporal correlations of the background activity. For the integrate-and-fire model these effects can be described using an ansatz by Brunel and Sergi [J. Theor. Biol. 195, 87 (1998)]. Numerical simulations show that the leaky integrate-and-fire model and the Hodgkin-Huxley type point neuron behave qualitatively similarly.

Action Potentials↗

PubMed2005

Invariant computations in local cortical networks with balanced excitation and inhibition.

Cortical computations critically involve local neuronal circuits. The computations are often invariant across a cortical area yet are carried out by networks that can vary widely within an area according to its functional architecture. Here we demonstrate a mechanism by which orientation selectivity is computed invariantly in cat primary visual cortex across an orientation preference map that provides a wide diversity of local circuits. Visually evoked excitatory and inhibitory synaptic conductances are balanced exquisitely in cortical neurons and thus keep the spike response sharply tuned at all map locations. This functional balance derives from spatially isotropic local connectivity of both excitatory and inhibitory cells. Modeling results demonstrate that such covariation is a signature of recurrent rather than purely feed-forward processing and that the observed isotropic local circuit is sufficient to generate invariant spike tuning.

Action Potentials↗

PubMed2005

Optimal filtering for spike sorting of multi-site electrode recordings.

We derive an optimal linear filter, to reduce the distortions of the peak amplitudes of action potentials in extracellular multitrode recordings, which are due to background activity and overlapping spikes. This filter is being learned very efficiently from the raw recordings in an unsupervised manner and responds to the average waveform with an impulse of minimal width. The average waveform does not have to be known in advance, but is learned together with the optimal filter. The peak amplitude of a filtered waveform is a more reliable estimate for the amplitude of an action potential than the peak of the biphasic waveform and can improve the accuracy of the event detection and clustering procedures. We demonstrate a spike-sorting application, in which events are detected using the Mahalanobis distance in the N-dimensional space of filtered recordings as a distance measure, and the event amplitudes of the filtered recordings are clustered to assign events to individual units. This method is fast and robust, and we show its performance by applying it to real tetrode recordings of spontaneous activity in the visual cortex of an anaesthetized cat and to realistic artificial data derived therefrom.

Action Potentials↗

PubMed2004

A method for the automatic segmentation of autoradiographic image stacks and spatial normalization of functional cortical activity patterns.

This paper introduces two new methods for the automatic anatomical and functional analysis of neurobiological autoradiographic image stacks, such as 2-fluoro-deoxyglucose (2FDG) images. The difficulty in the evaluation of these "2(1/2)D" datasets is that they do not inherently represent a continuous 3D data volume (as generated by MRI or CT), but consist of a stack of images from single tissue slices, suffering from unavoidable preparation artifacts. In the first part of the paper, a semi-automatic segmentation method is presented which generates a 3D surface model of certain brain structures and which is robust against different cutting directions with respect to the brain coordinate system. The method saves man-hours compared to manual segmentation and the results are highly reproducible. In the second part, a fully automatic method for the extraction, analysis and 3D visualization of functional information is described, which allows not only a more accurate localization of activation sites, but also greatly enhances the comparability of different individuals. Results are shown for 2FDG autoradiographs from rat brains under acoustical stimulation.

Animals↗

PubMed2004

New methods for the computer-assisted 3-D reconstruction of neurons from confocal image stacks.

Exact geometrical reconstructions of neuronal architecture are indispensable for the investigation of neuronal function. Neuronal shape is important for the wiring of networks, and dendritic architecture strongly affects neuronal integration and firing properties as demonstrated by modeling approaches. Confocal microscopy allows to scan neurons with submicron resolution. However, it is still a tedious task to reconstruct complex dendritic trees with fine structures just above voxel resolution. We present a framework assisting the reconstruction. User time investment is strongly reduced by automatic methods, which fit a skeleton and a surface to the data, while the user can interact and thus keeps full control to ensure a high quality reconstruction. The reconstruction process composes a successive gain of metric parameters. First, a structural description of the neuron is built, including the topology and the exact dendritic lengths and diameters. We use generalized cylinders with circular cross sections. The user provides a rough initialization by marking the branching points. The axes and radii are fitted to the data by minimizing an energy functional, which is regularized by a smoothness constraint. The investigation of proximity to other structures throughout dendritic trees requires a precise surface reconstruction. In order to achieve accuracy of 0.1 microm and below, we additionally implemented a segmentation algorithm based on geodesic active contours that allow for arbitrary cross sections and uses locally adapted thresholds. In summary, this new reconstruction tool saves time and increases quality as compared to other methods, which have previously been applied to real neurons.

Algorithms↗

PubMed2003

Optimal noise-aided signal transmission through populations of neurons.

Metabolic considerations and neurophysiological measurements indicate that biological neural systems prefer information transmission via many parallel low intensity channels, compared to few high intensity ones [S. B. Laughlin et al., Nature Neurosci. 1, 36 (1998)]. Furthermore, cortical neurons are exposed to a considerable amount of synaptic background activity, which increases the neurons' conductance and leads to a fluctuating membrane potential that, on average, is close to the threshold [A. Destexhe and D. Paré, J. Neurophysiol. 81, 1531 (1999)]. Recent studies have shown that noise can improve the transmission of subthreshold signals in populations of neurons, e.g., if their response is pooled. In general, the optimal noise level depends on the stimulus distribution and on the number of neurons in the population. In this contribution we show that for a large enough number of neurons the latter dependency becomes weak, such that the optimal noise level becomes almost independent of the number of neurons in the population. First we investigate a binary threshold model of neurons. We derive an analytic expression for the optimal noise level at each single neuron, which-for a large enough population size-depends only on quantities that are locally available to a single neuron. Using numerical simulations, we then verify the weak dependence of the optimal noise level on population size in a more realistic framework using leaky integrate-and-fire as well as Hodgkin-Huxley-type model neurons. Next we construct a cost function, where quality of information transmission is traded against its metabolic costs. Again we find that-for subthreshold signals-there is an optimal noise level which maximizes this cost. This noise level, however, is almost independent of the number of neurons, even for small population sizes, as numerical simulations using the Hodgkin-Huxley model show. Since the dependence of the optimal noise level on population size is weak for large enough populations, local neural adaptation is sufficient to adjust the level of noise to its optimal value.

Animals↗

PubMed2003

Modeling the adaptive visual system: a survey of principled approaches.

Modeling the visual system can be done at multiple levels of description ranging from computer simulations of detailed biophysical models to firing rate and so-called 'black-box' models. Re-introducing David Marr's analysis levels for the visual system, we motivate the use of more abstract models in order to answer the question of what the visual system is computing. The approaches we selected to review in this article concentrate on modeling the changes of sensory representations. The considered time-scales, range from the developmental time-scale of receptive field formation to fast transient neuronal dynamics during a single stimulus presentation. Common to all approaches is their focus on providing functional interpretations, instead of 'only' explanations in terms of mechanisms. Although the concrete approaches can be distinguished along different lines, a common theme is emerging which may qualify as a paradigm for providing functional interpretations for changes of receptive field properties, i.e. the dynamic adjustment of sensory representations to varying external or internal conditions.

Adaptation, Physiological↗

PubMed2003

A new point-based warping method for enhanced and simplified analysis of functional brain image data.

Comparison of brain imaging data requires the exact matching of data sets from different individuals. Warping methods, used to optimize matching of data sets, can exploit either local gray value distribution or identifiable reference points within the images to be compared. Gray value-based warping, which is more comfortable, cannot be used if gray values include functional information that should be compared between images. A major drawback in the use of point-based warping methods is the lack of methods for efficient and precise definition of reference points (landmarks) within comparable data sets. Here, we present a novel approach to automatically detect sufficient numbers of landmarks, which is based on 3D differential operators. In addition, we have developed a new distance-weighted warping method, which optimizes individual local weighting factors of displacement vectors. The quality of the methods was evaluated using a set of autoradiographs documenting the metabolic activity of gerbil brains after acoustic stimulation. The new warping method was compared with known methods of landmark-based warping, i.e., warping with radial basis functions and with distance-weighted methods. For the data sets presented in this study our new optimized warping method produced an increase in linear cross correlation of 4.44%, an increase in volume overlap index of 1.55%, and a decrease in the registration error of 36.2%. In addition, the detection of functional differences was improved after warping. Therefore, the new method is a powerful tool, which enhances the comparison of complex biological structures and the quantitative evaluation of functional imaging data.

Algorithms↗

PubMed2003

Soft learning vector quantization.

Learning vector quantization (LVQ) is a popular class of adaptive nearest prototype classifiers for multiclass classification, but learning algorithms from this family have so far been proposed on heuristic grounds. Here, we take a more principled approach and derive two variants of LVQ using a gaussian mixture ansatz. We propose an objective function based on a likelihood ratio and derive a learning rule using gradient descent. The new approach provides a way to extend the algorithms of the LVQ family to different distance measure and allows for the design of "soft" LVQ algorithms. Benchmark results show that the new methods lead to better classification performance than LVQ 2.1. An additional benefit of the new method is that model assumptions are made explicit, so that the method can be adapted more easily to different kinds of problems.

Costs and Cost Analysis↗

PubMed2002

Rapid adaptation to internal states as a coding strategy in visual cortex?

Adaptation is a prominent feature of biological neuronal systems. A common interpretation of adaptation in terms of function is that it provides flexibility for a neuronal system to perform well under varying external conditions, for example by adjusting the input/output relation of a sensory system with reference to the ensemble of stimuli the organism currently perceives. This interpretation, however, only applies if the time-scale of adaptation is slower than the time-scale at which the environment changes. Experimentally it is observed, however, that adaptation can be very rapid. Spike-frequency adaptation of cortical neurons, for example, occurs on a time-scale of approximately 100 ms. Here we show that those rapid adaptation processes can also be understood within the framework of information theory. We start with the hypothesis that neuronal codes are designed to optimize the information a neuronal representation conveys about an input stimulus for any increasing time window beginning with stimulus onset, and we show that this implies a rapid adaptation of the neuronal code on the time-scale of stimulus presentation. Adaptation, however, does not occur because the state of the environment changes. Rather it is a reaction to changes of the organisms own internal state, e.g. the level of noise in the neuronal representation. We apply this approach to a model of an orientation hypercolumn in the primary visual cortex, and predict that inter-columnar interactions should adapt on the time-scale of a typical fixation period ( approximately 300 ms).

Adaptation, Psychological↗

PubMedDate not supplied

Self-organizing maps and clustering methods for matrix data.

In this contribution we present extensions of the Self Organizing Map and clustering methods for the categorization and visualization of data which are described by matrices rather than feature vectors. Rows and Columns of these matrices correspond to objects which may or may not belong to the same set, and the entries in the matrix describe the relationships between them. The clustering task is formulated as an optimization problem: Model complexity is minimized under the constraint, that the error one makes when reconstructing objects from class information is fixed, usually to a small value. The data is then visualized with help of modified Self Organizing Maps methods, i.e. by constructing a neighborhood preserving non-linear projection into a low-dimensional "map-space". Grouping of data objects is done using an improved optimization technique, which combines deterministic annealing with "growing" techniques. Performance of the new methods is evaluated by applying them to two kinds of matrix data: (i) pairwise data, where row and column objects are from the same set and where matrix elements denote dissimilarity values and (ii) co-occurrence data, where row and column objects are from different sets and where the matrix elements describe how often object pairs occur.

Cluster Analysis↗