However, unlike in both these experimental Hermann, and model-based Labecki et al. The recordings made for the purposes of this work are at 10, 20, and 40 Hz, representative samples of the upper-alpha, beta and gamma band EEG frequencies, respectively.
The model response validates experimental findings of entrainment with periodic input stimulus. Both cases corresponding to the IN being dominant or suppressed in the LGN circuit are tested with the electronic retina-generated spike trains. The power spectral plots corresponding to the 20 and 40 Hz inputs are in agreement with the model response corresponding to the synthetic input stimulus. In addition to the loss of fidelity with suppression of IN inhibition in the model output corresponding to the 40 Hz spike-train input from the electronic retina, we also note a similar loss in fidelity corresponding to the 10 Hz input from the electronic retina, when the maximum output power is within the second harmonic.
However, this is not the case for the synthetic model input, and is unlike in a recent work with population models of the LGN Sen-Bhattacharya et al.
One reason may be the noise in the electronic retina output spike-train, reflecting a realistic external environment that drives the retinal spiking neuron. In comparison, the model synthetic data is devoid of any noise. Another speculation toward the difference in behavior is the reduced number of neurons in our LGN model; lower frequency band EEG rhythms, for example alpha rhythms, are known to be generated by synchronous behavior of larger neuronal populations compared to localized population activity corresponding to higher frequencies.
Thus, we have simulated a relatively larger neuronal population of the LGN tissue consisting of neurons. Indeed, the response of this multi-nodal network architecture corresponding to suppression of IN inhibition in the circuit shows a loss of output fidelity corresponding to a 10 Hz periodic visual stimulus. A more rigorous testing, and for frequency ranges wider than the current range, is suggested as future work on further scaled up multi-nodal LGN model. To consolidate our observation of model entrainment with periodic input, and to confirm that such behavior is not an artefact of the materials and methods adopted in this work, we have tested the model with aperiodic visual stimulus.
This is simulated by a spike-train following Poisson distribution with parameters in the same range as the periodic input frequencies. Results show no trace of entrainment. Furthermore, the model output power spectra is invariant in the tested range.
However, this would need to be confirmed with larger LGN networks on the SpiNNaker, which is aligned to the short-term plans for future work. In contrast, our study shows that corresponding to periodic stimulus, even the small network is able to simulate SSVEP, which is a higher-level network dynamics.
As above-mentioned, these observations will be further tested with a scaled-up LGN model as a part of ongoing research.
We note three specific areas that may contribute to enhance the framework presented here:. First, the power spectra and amplitude in the model are bound to change with increased stimulus strength, which can be effected in the model by changing connectivity parameter attributes. In the current model, these attributes are set to the minimal threshold values for initiating a spike output response in the model. During preliminary investigations, we have indeed noted increases in both the spike rate and amplitude of membrane voltage for increased connectivity parameter values, along the lines as noted in Notbohm et al.
We leave this to be taken up in future research on the model. Second, we have explored the tonic behavior for all cell types. However, it is known that the bursting nature for all cell types in the LGN underpins brain rhythms not only in the resting state but also in regulating attention.
Simulating both tonic and bursting behavior in the LGN will certainly enhance the biological plausibility of the model. It may be noted that for brevity in this work, the retinal spike-train output is not provided to the model in real time. However, the interface of the electronic retina with the SpiNNaker machine is already available Galluppi et al. The main drawback of the model is the lack of cortical circuitry, as thalamo-cortico-thalamic dynamics form the basis of brain rhythms observed via EEG and LFP.
However, a decorticated disconnected from the cortex model of the LGN in this work is by design rather than any other constraint, and is justified as a necessary step prior to building larger interfaced structures; similarities can be drawn with several decades of research on isolated thalamic slices of mammals and rodents, that has paved the way for the current advanced understanding of the thalamo-cortical dynamics.
In terms of performance evaluation of model simulation on SpiNNaker, two areas take precedence: Real time implementation—we note that the differential equations defining a single neuro-computational unit in the model current-based Izhikevich's neuron need a 0. Thus, all simulation of the LGN model on SpiNNaker ran 10 times slower than real time; for example, if the simulation duration is set to 3 s with a time-step of 0.
However, such scaling up of simulation time is not a concern for this small LGN network, serving as a test-bed for future large-scale simulations, particularly because the model is guaranteed to run in the expected time. However, sPyNNaker is in development mode, and we do expect to see issues upon scaling up the model.
Thus, we expect to be able to provide a more realistic evaluation on the real time performance of SpiNNaker when we run scaled-up versions of this model, as well as others that are under development. Power consumption—Prior work has provided estimated figures for maximum power consumption by a SpiNNaker chip as 1 W.
This is verified by an in-house power measurement set-up that records the active power wattage drawn from the mains by the SpiNNaker board. Computational neurology and psychiatry provide an excellent means for in-depth investigations of vital structures such as the thalamus, which are otherwise hard to study in wet-laboratories Sen-Bhattacharya et al.
Besides dealing with sensory and cortical inputs, the thalamic nuclei play strategic roles in the functioning of the limbic brain, and are known to link decision making and action selection circuitry to the motor circuitry. Thalamic dysfunction is often believed to underpin several neurological and psychiatric disorders. On the other hand, the thalamus forms the primary output target of the Basal Ganglia BG circuit, a brain structure and mechanism that has been the primary focus of the autonomous robotics community toward incorporating learning and decision making in machines Humphries and Gurney, BS planned and designed the research and ran model simulation for generating results.
TS performed data acquisition, data processing, model simulation and generating results related to electron retina. LB and AB made voluntary research contributions to the software code of the model; LB contributed toward object oriented Python script for the multi-node LGN architecture and related data visualization; AB contributed toward Python script for data processing and visualization of results.
IS performed power analysis of the model simulation on SpiNNaker. SF provided guidance and support on the research.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer VV and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review.
BS is supported by the ERC grant. SpiNNaker has been 15 years in conception and 10 years in construction, and many folk in Manchester and in our various collaborating groups around the world have contributed to get the project to its current state.
We gratefully acknowledge all of these contributions. The authors acknowledge the contribution of John V. Woods toward careful proof-reading of the document. Adams, S. Bal, T. Synaptic and membrane mechanisms underlying synchronized oscillations in the ferret lateral geniculate nucleus in vitro.
Benjamin, B. Neurogrid: a mixed-analog-digital multichip system for large-scale neural simulations. IEEE , — Rhythms of the Brain, 1st Edn. Cox, C. Functional synaptic contacts by intranuclear axon collaterals of thalamic relay neurons. PubMed Abstract Google Scholar. Crunelli, V. Thalamic t-type calcium channels and nrem sleep. Cell Calcium 40, — Organisation of thalamic and cortical alpha rhythms: spectra and coherences.
Davison, A. PyNN: a common interface for neuronal network simulators. Diez, P. Asynchronous bci control using high-frequency ssvep. Eliasmith, C. The use and abuse of large-scale brain models. Elibol, R. Google Scholar. Furber, S. Large-scale neuromorphic computing systems. Neural Eng. The spinnaker project. Overview of the spinnaker system architecture. IEEE Trans. Galbraith, B. Neural Modelling with Python Part Galluppi, F. Huang, Z.
Zeng, C. Li, and C. Leung Berlin; Heidelberg: Springer , — Guger, C. Halassa, M. Thalamic inhibition: diverse sources, diverse scales. Trends Neurosci. Hermann, C. Human EEG responses to hz flicker: resonance phenomena in visual cortex and their potential correlation to cognitive phenomena. Brain Res. Hirsch, J. How inhibitory circuits in the thalamus serve vision.
Hopkins, M. Accuracy and efficiency in fixed-point neural ODE solvers. Neural Comput. Horn, S. Relative distribution of synapses in the A-laminae of the lateral geniculate nucleus of the cat.
Humphries, M. The role of intra-thalamic and thalamocortical circuits in action selection. Networks 13, — Huntsman, M. Cells in the six layers that align have receptive fields in the same area of the retina. In addition, the optic nerve fibers from the two eyes project to distinct layers of LGN.
Fibers from the eye on the same side project to layers 2, 3, and 5 of LGN, whereas fibers from the opposite eye project to layers 1, 4, and 6. Reading a neural code. Shannon, C. The Mathematical Theory of Communication Univ. Illinois, Chicago, Sutter, E. Marmarelis — Univ. Southern California, Los Angeles, Reid, R. The use of m-sequences in the analysis of visual neurons: Linear receptive field properties.
Usrey, W. M, Reppas, J. Paired-spike interactions and synaptic efficacy of retinal inputs to thalamus. Nature in press. Cleland B. Simultaneous recording of input and output of lateral geniculate neurones.
CAS Google Scholar. Sustained and transient neurones in the cat's retina and lateral geniculate nucleus. R — Cambridge Univ. Press, London, Hubel, D.
Integrative action in the cat's lateral geniculate body. Kaplan E. Contrast affects the transmission of visual information through the mammalian lateral geniculate nucleus. Synchronous activity in the visual system. Brivanlou I. Mechanisms of concerted firing among retinal ganglion cells. Neuron 20 , — Panzeri, S. Analytical estimates of limited sampling biases in different information measures.
Network: Computation in Neural Systems 7 , 87— Perkel, D. Neuronal spike trains and stochastic point processes. Simultaneous spike trains. Download references. We are grateful to T. Wiesel for his support during the early phase of this work. Save Word. Medical Definition of lateral geniculate nucleus. Learn More About lateral geniculate nucleus. Share lateral geniculate nucleus Post the Definition of lateral geniculate nucleus to Facebook Share the Definition of lateral geniculate nucleus on Twitter.
Dictionary Entries Near lateral geniculate nucleus lateral geniculate body lateral geniculate nucleus lateral horn See More Nearby Entries. Style: MLA.
0コメント