Australian
Neuroscience Society Meeting (ANS 2006),
(Abstract1),
(Abstract2),
Sydney, Australia, 31 January - 3 February 2006.
International Conference on Intelligent Sensors, Sensor Networks and Information
Processing (ISSNIP 2004),
(Abstract),
Melbourne, Australia, 14-17 December 2004.
Tenth Australian International Conference on Speech Science and Technology (SST 2004),
(Abstract),
Sydney, Australia, 8-10 December 2004.
Thirteenth Annual Computational Neuroscience Meeting (CNS*2004),
(Abstract),
Baltimore, Maryland, USA, 18-22 July 2004.
Australian
Neuroscience Society Meeting (ANS 2004),
(Abstract1),
(Abstract2),
(Abstract3),
Melbourne, Australia, 27-30 January 2004.
Seventh International Conference on Cognitive and Neural Systems (ICCNS 2003),
(Abstract),
Boston, USA, 28-31 May 2003.
Mathematical Biosciences Institute (MBI) Workshop on the Auditory System (2003),
(Abstract),
Ohio State University, USA, 5-9 May 2003.
Eighth Western Pacific Acoustics Conference (WESPAC 2003),
(Abstract),
Melbourne, Australia, 7-9 April 2003.
Australian
Neuroscience Society Meeting (ANS 2003),
(Abstract1),
(Abstract2),
Adelaide, Australia, 28-31 January 2003.
Ninth Australian International Conference on Speech Science and Technology (SST 2002),
(Abstract),
Melbourne, Australia, 3-5 December 2002.
Eleventh Annual Computational Neuroscience Meeting (CNS*2002),
(Abstract),
Chicago, Illinois, USA, 21-25 July 2002.
Australian
Neuroscience Society Meeting (ANS 2002),
(Abstract),
Sydney, Australia, 3-6 February 2002.
International
Joint Conference on Neural Networks 2001 (IJCNN 2001) ,
(Abstract1),
(Abstract2),
Washington DC, USA, 14-19 July 2001.
Tenth Annual Computational Neuroscience Meeting (CNS*2001),
(Abstract),
Pacific Grove, California, USA, 30 June-5 July 2001.
Biomedical Research in 2001 (IEEE Engineering in Medicine and Biology Society),
(Abstract1), (Abstract2),
Monash University, Melbourne, Australia, 19-20 February 2001.
Australian
Neuroscience Society Meeting (ANS 2001),
(Abstract),
Brisbane, Australia, 28-31 January 2001.
Australian
NIH-NINDS Neural Prosthesis Workshop,
(Abstract),
Washington, USA, 25-27 October 2000.
Australian
Neuroscience Society Meeting (ANS 2000),
(Abstract1),
(Abstract2),
(Abstract3),
(Abstract4),
(Abstract5),
Melbourne, Australia, 30 January - 2 February 2000.
Eighth
Annual Computational Neuroscience Meeting (CNS*99),
(Abstract1),
(Abstract2),
Pittsburgh, Pennsylvania, USA, 18-22 July 1999.
International
Joint Conference on Neural Networks 1999 (IJCNN'99) ,
(Abstract1),
(Abstract2),
Washington DC, USA, 10-16 July 1999.
IEEE
Biomedical Research Conference 1999,
(Abstract),
Monash University, Melbourne, Australia, 22-23 February 1999.
Australian
Neuroscience Society Meeting (ANS 1999),
(Abstract),
Hobart, Australia, 31 January - 3 February 1999.
Understanding
the Brain and Engineering Models (UBEM'99) Workshop ,
(Abstract),
Sydney University, Australia, 18-19 January 1999.
Seventh
Annual Computational Neuroscience Meeting (CNS*98),
(Abstract),
Santa Barbara, USA, 26-30 July 1998.
Ninth
Australian Conference on Neural Networks (ACNN'98),
(Abstract),
Brisbane, Australia, 11-13 February 1998.
Purpose: Experimental evidence indicates that synaptic modification depends upon the timing relationship between the presynaptic inputs (excitatory postsynaptic potentials - EPSPs) and the output spikes (action potentials - APs) that they generate. Synaptic structure formation was examined in a class of spike timing-dependent plasticity (STDP) models that is both competitive and stable. Methods: The analysis of synaptic structure formation was based upon the Fokker-Planck equation using a self-consistent formalism, and it was carried out using a homogeneous Poisson spiking-rate neuronal model. The inputs consisted of correlated synaptic subgroups and the process of synaptic structure formation was examined for one class of STDP models. Results: The results of analytical calculations and numerical simulations (each subgroup consisting of 40 synapses, simulations over 500,000 output spikes) showed the following pattern of behaviour: (i) small learning rates (0.0003) produce multiple alternative synaptic structures that depend upon the initial configuration, (ii) large learning rates (0.03) produce synaptic structures that do not stabilize, resulting in neurons without consistent response properties, and (ii) intermediate learning rates (0.003) produce a unique and stable synaptic structure that exhibits input selectivity. The robustness of the selectivity is largely determined by the ratio of the subgroup correlation strength to the number of subgroups. Conclusions: The analysis provides an account of the conditions that permit the formation of synaptic structure through stable and competitive STDP when inputs consist of correlated synaptic subgroups.
Abstract published in the Proceedings of the Australian Neuroscience Society 2006, Vol. 17, p.127 (ISSN No. 1034-3237).
Purpose: T-stellate cells in the cochlear nucleus (CN) provide a robust representation of the spectrum of sound input. While neural models of T-stellate cells have previously focused on specialised membrane properties, experimental evidence indicates that synaptic interactions are affecting the spectral and temporal processing of T-stellate cells. This study develops a network model using Hodgkin-Huxley neurons to investigate the synaptic organisation of interneurons within the CN that are suitable candidates for lateral inhibition of T-stellate cells. Methods: Using the neural simulator NEURON, individual cell models of T-stellate cells, GABAergic golgi cells and glycinergic D-stellate and tuberculoventral cells were created. Network variables (eg. synaptic strength, delays, bandwidth) were manipulated automatically using a genetic algorithm. Each parameter set was given an objective value based on comparisons with experimental physiological results in the literature. Stimuli included CF tones, notch filtered noise, tone+noise maskers and clicks. Results: The genetic algorithm was an effective tool in optimising 20 real-valued parameters after simulating over 6000 combinations. GABAergic inhibition of D-stellate cells by golgi cells plays a significant role in modulating D-stellate post-onset responses. D-stellate-to-tuberculoventral cell projection was fast enough to inhibit response to clicks and was offset by +0.2-0.3 octaves; a corroboration of recent results. Parameters for D-stellate and tuberculoventral cell inhibition of T-stellate cells found both inputs were similar, except for greater bandwidth of D-stellate cells. This suggests that D-stellate and tuberculoventral cells provide complementary mechanisms for lateral inhibition at the onset of broad sounds and throughout narrowband sounds, respectively. Conclusion: The derived CN model demonstrates cell behaviour that reproduces characteristic responses to various stimuli. This allows us to examine the effects of the neural circuit on T-stellate cells..
Abstract published in the Proceedings of the Australian Neuroscience Society 2006, Vol. 17, p.40 (ISSN No. 1034-3237).
The use of auditory models in cochlear implant sound processing strategies aims to improve cochlear implant users’ perception of speech, particularly in noisy environments. To date, cochlear implant sound processing strategies have been designed using simple envelope extraction techniques. Our new strategy simulates the behaviour of the cochlea and the auditory nerve to give a stimulation sequence that produces responses in the auditory nerve that are closer to those in normal hearing. The development and evaluation of the strategy will also help to answer the fundamental question of how much benefit is provided by fine-grained timing information to the perception of different sounds that make up speech.
To be Published in the Proceedings of the International Conference on Intelligent Sensors, Sensor Networks and Information Processing 2004
[Full text] (PDF Document, 353 Kb, 5 pages)
We present a detailed network model of neurons of the cochlear nucleus to explore leveldependent processing by T stellate cells. These cells selectively process input from two types of auditory nerve fibres to enhance their dynamic range. We use a biologically plausible network between T stellate, D stellate and tuberculoventral cells, to show how level-dependent selective processing can be used to recreate rate-level curves seen in vivo and encode the spectrum of the vowel /e/. The compartmental models include recently formulated current models of ventral cochlear nucleus cells to give greater accuracy of cell spiking behaviour.
To be Published in the Proceedings of the Tenth Australian International Conference on Speech Science and Technology 2004
[Full text] (PDF Document, 753 Kb, 6 pages)
An experimentally supported model of cortical background activity is used to investigate the role of such activity in neural gain control. The model demonstrates the feasibility of a scheme for contrast enhancement where by the overall intensity of an input pattern adjusts the dynamic range of a neuron such that it remains sensitive to contrast over a wide range of overall intensities.
To appear in Neurocomputing.
Purpose: Experimental evidence indicates that synaptic modification depends upon the timing relationship between the presynaptic inputs (excitatory postsynaptic potentials - EPSPs) and the output spikes (action potentials - APs) that they generate. Spike timing-dependent plasticity (STDP) with additive potentiation and multiplicative depression was examined and four classes of STDP were identified on the basis of the time-extent of their input-output (EPSP-AP) interactions. Methods: The analysis was based upon the Langevin equation and the Fokker-Planck formalism, and it was carried out using a leaky integrate-and-fire neuronal model with synaptic conductances. Results: The effect upon the potentiation of synapses with different rates of input was analysed to elucidate the relationship of STDP with classical studies of LTP/LTD and rate-based Hebbian learning. The selective potentiation of higher input-rate synapses was found only when both: (i) the time-extent of the input-output interactions were "input restricted" (i.e., restricted to time domains delimited by adjacent synaptic inputs), and (ii) the time-asymmetric learning window had a longer time constant for depression than for potentiation (as observed experimentally). The effect of suppressive interspike interactions upon STDP were also analysed and shown to modify the synaptic dynamics - the amount of potentiation and depression was quantitatively changed, but the qualitative behaviour remained the same. Conclusions: The analysis provides an account of synaptic dynamics determined by a stable fixed-point. Only input-restricted models of STDP were found to have an input-selective stable fixed-point behaviour.
Abstract published in the Proceedings of the Australian Neuroscience Society 2004, Vol. 15, p.77 (ISSN No. 1034-3237).
These experiments were undertaken to determine the role of the different cell types within the ventral cochlear nucleus (VCN) in representing a key temporal feature of speech. The waveform of a vowel is characterised by regular, high amplitude peaks that mark the periodic opening and closing of the vocal folds (glottis); the repetition rate of these glottal pulses determines a vowel's fundamental frequency (F0 or pitch). We examined F0 coding within the VCN by presenting naturally spoken syllables (adult male; F0 ~ 100 Hz) over a range of sound pressure levels (55-85 dB). The problem of analysing neural responses to a natural stimulus with a variable fundamental period was overcome by developing a pitch-synchronous vector strength (VS) program that could be applied to responses to the entire vowel or segments of it. Extracellular recordings were made from 171 single neurons in pigmented rats anaesthetized with 20% urethane; neurons with a wide range of best frequencies (BFs) were sampled (0.66 to 10 kHz). The capacity of cells representing the major VCN response types (primary-like, PL, 67% of sample; chopper, CH, 21%; onset, ON, 12%) to lock to a vowel's F0 was determined by calculating the VS for the entire vowel in the syllable /got/. Significant differences (t-test, p<0.05) were found between cell types. ON cells were superior F0 coders and showed higher average VS values than CH cells, which, in turn, had higher average VSs than PL cells. VS also varied systematically as a function of cell BF and syllable SPL. Analyses of neural responses within different (moving) windows during a relatively long vowel (in the syllable /ka/) revealed that a cell's VS is a dynamic characteristic, and can change quite dramatically during the course of the vowel. (Supported by NHMRC project grant 217303).
Abstract published in the Proceedings of the Australian Neuroscience Society 2004, Vol. 15, p.85 (ISSN No. 1034-3237).
Based on studies in echolocating bats, it has been proposed that the ventral nucleus of the lateral lemniscus (VNLL) plays a prominent role in temporal pattern detection. This function has also been suggested for non-echolocating species, however, this was based on anatomical evidence rather than a direct physiological investigation. As a precursor to such a study, this area's functional organisation needed to be clarified because previous studies have yielded conflicting results. This study recorded either intracellularly or extracellularly from single VNLL neurons (n=53) in urethane-anaesthetised rats (2.6g/kg i.p.), while dichotically presenting noise or tone bursts. A standardised set of coronal sections was constructed through the VNLL and divided into sectors, allowing an assessment of the distribution of response characteristics (e.g. best frequency (BF), binaural interaction (if any), response type). A decision tree was developed to categorise objectively a neuron's response pattern to supra-threshold BF tones. Currently, most recorded units are located in the anterior two-thirds of the VNLL; nevertheless, it appears that some response types are preferentially located in certain sectors. For instance, onset-ideal cells, the most common VNLL response type (52%), tend to be located dorsorostrally within VNLL. Segregation of response types has been described in most species studied. Most cells showed excitation to stimulation of the contralateral ear (77%) and were distributed throughout the VNLL. Cells that showed an excitatory response to stimulation of the contralateral ear and inhibition to ipsilateral ear stimulation (EI: 23%) were only found dorsally, while IE cells (8%) were only found ventrally. This finding is consistent with a previous description in the cat, in which binaural cells were located at the VNLL edge. Presently, we have no physiological evidence for tonotopicity in rat VNLL in agreement with the weight of evidence in other species.
Abstract published in the Proceedings of the Australian Neuroscience Society 2004, Vol. 15, p.90 (ISSN No. 1034-3237).
[Full Text of Abstract] (Word Document, 24 Kb, 1 page)
[Full Text of Abstract] (Word Document, 24 Kb, 1 page)
Additional material is available online from the MBI Workshop website
One of the fundamental problems in speech processing is removal of noise from the signal and extraction of important parameters such as the formant frequencies. We present a method that uses an auditory model and modulation Poisson rate estimation for formant estimation and removal of noise from a speech signal. A noisy speech signal is first put through an auditory model to generate a spike-sequence similar to that observed in the auditory nerve. The firing of the spike train is modelled as a doubly stochastic process where the firing rate is a function of the underlying process. The underlying process is estimated from an observation counting process. The instantaneous frequency obtained from the inter-spike intervals is controlled by the dominant frequency components of the signal. Thus, the estimation procedure results in rate estimates that follow the speech formants closely. The speech is then reconstructed using these formant estimates resulting in an enhanced signal by reducing the contribution of noise outside the formant bandwidths. The procedure and speech enhancement results are presented.
To be published in the Proceedings of the Eighth Western Pacific Acoustics Conference (WESPAC 2003).
[Full text] (PDF Document, 183 Kb, 4 pages)
A network model utilising sparsely connected spiking neurons and conductance-based synapses with reversal potentials was studied analytically to see if cortical background activity could be explained by the random self-sustained firing of excitatory and inhibitory neurons. In agreement with a previous study, which ignored the effect of reversal potentials, the model shows that moderate background activity (1-20 Hz) is sustained provided that (1) external excitatory input exceeds a critical but comparatively small value (~1 Hz for each input neuron), and (2) recurrent excitation is not stronger than recurrent inhibition. Unlike the previous model however, this model also reproduces much of the experimental data that characterises cortical background activity, including the mean firing rate (1-20 Hz), the high coefficient of variation of the inter-spike intervals (0.9), the mean depolarisation of the membrane potential from rest (10-15 mV), the standard deviation of the membrane potential (3-4 mV) and the effective membrane time constant (0.6-6 ms). These results indicate that, over a wide range of external inputs, random self-sustained activity of neurons can account for much of the experimental data on cortical background activity once synaptic input is treated as conductance-based rather than a simple voltage-independent current injection.
Abstract published in the Proceedings of the Australian Neuroscience Society 2003, Vol. 14, p.330 (ISSN No. 1034-3237).
The functional role of the ventral nucleus of the lateral lemniscus (VNLL) is poorly understood within the ascending auditory pathway of mammals. Very little work has been conducted in vivo , and there is have been no intracellular studies to investigate responses to acoustic stimuli. Microelectrodes (1M KAc) were used to record intracellular and extracellular responses in vivo to bilaterally presented acoustic stimuli in the VNLL of urethane anaesthetised rats (2.6g/kg i.p.). Animals were transcardially prefused with formalin and standard histological techniques employed to confirm recorded neuron locations. Intracellular responses of 24 VNLL neurons were classified according to white noise responses: 75% were monaural, of which 78% were contralaterally driven. The binaural response types of the 6 remaining cells were: Excitatory-inhibitory (EI), n=2; EE, n=3; II, n=1. In 16 extracellularly and 4 intracellularly recorded neurons, responses to best frequency tones were used to classify the neurons into the following groups: Onset-Ideal (O-I) = 30%; Onset-Sustained = 30%; Sustained = 20%; Chopper = 5%; Onset-Chopper = 5%; Complex = 10%. All O-I cells were monaural and contralaterally driven. The response pattern, as well as the fact that 65% of the cells featured a predominately onset response, could reflect the major projection from the octopus cell area of the contralateral cochlear nucleus. The results suggest the VNLL may be a binaural structure with a diverse population of neural response types, including inhibitory interactions.
Abstract published in the Proceedings of the Australian Neuroscience Society 2003, Vol. 14, p.265 (ISSN No. 1034-3237).
This paper describes a method of rate estimation for spike trains produced by auditory models. The firing of the spike train is modelled as a doubly stochastic process where the firing rate is a function of an underlying process. The underlying process is estimated from an observation counting process. The variance of the estimator is calculated for tracking a random walk stochastic FM signal to give an empirical performance measure. Formant tracking is also demonstrated as an application for the use of this type of estimation.
Published in the Proceedings of the Ninth Australian International Conference on Speech Science and Technology 2002 (SST-2002, Melbourne, 3-5 December 2002), Australian Speech Science and Technology Association (Inc.) Melbourne 2002, p.16-21 (ISBN 0 9581946 0 2).
[Full text] (Word Document, 257 Kb, 6 pages)
Gain modulation of neural responses by the balanced component of the synaptic input is analyzed in the gaussian approximation using a single compartment conductance-based neural model. The model is analyzed in the ``normal operating regime'', in which the output spiking-rate of the neuron is equal to the spontaneous spiking-rate in the absence of any stimulus. The gain in response to both additional excitatory synaptic input and injected current is found to be modulated in a non-linear way by the level of balanced synaptic input.
(Paper to appear in CNS*2002 Proceedings)
Neurons in vivo are subject to a constant barrage of both excitatory and inhibitory synaptic input, which significantly alters the properties of the neurons relative to their properties in vitro. In order to examine these changes, a conductance-based single compartment model is analysed, in which the total synaptic current is decomposed into a sum of the excitatory and inhibitory synaptic conductances. The model incorporates excitatory and inhibitory reversal potentials and the time distribution of the synaptic inputs is modelled as a Poisson process. The analysis is carried out in the Gaussian approximation, which comparison with numerical simulations confirms is most accurate in the limit of a large number of inputs. The hypothesis that the observed variability in the spike times of cortical neurons is caused by a balance of excitatory and inhibitory synaptic inputs is supported by the results for the coefficient of variation of the interspike intervals. Its value decreases with both increasing numbers and amplitude of inputs, and is consistently lower than 1.0 over a wide range of realistic parameter values. The gain of the neural responses, defined as the sensitivity of the firing rate to injected current, is found to depend upon the rates of excitatory and inhibitory synaptic inputs. A systematic investigation of the dependence of the output spike rate upon the rate, number and amplitude of the synaptic inputs, as well as upon the value of the inhibitory reversal potential, is given. A low spiking-rate of spontaneous activity in both the excitatory and inhibitory neurons is found to be stable over a range of values of the neural parameters.
Abstract published in the Proceedings of the Australian Neuroscience Society 2002, Vol. 13, p.83.
We study the influence of noise modulation on the detection of subthreshold periodic signals in threshold detectors under the paradigm of stochastic resonance. More particularly, we show that having an input noise amplitude modulated by the input signal can enhance the phenomenon of stochastic resonance. We then apply this result to a neuron model for which the noise modulation mechanism is an intrinsic property coming from the shot-noise nature of the neuron membrane potential.
Published in the Proceedings International Joint Conference on Neural Networks (IJCNN'01 , Washington DC , 14-19 July 2001), Vol.1, pp.644-647, Piscataway, New Jersey : IEEE.
[Full text] (PDF 160 Kb, 4 pages)
Neurons in which the level of excitation and inhibition are roughly balanced are shown to be very sensitive to the coherence of their synaptic input. The behavior of such balanced neurons with reversal potentials is analyzed both analytically and numerically using the leaky integrate-and-fire neural model. The investigation uses the Gaussian approximation with synaptic inputs modeled as inhomogeneous Poisson processes. The results indicate that for balanced neurons with $N$ synaptic inputs, it is only necessary for $O(\sqrt{N})$ of the synaptic inputs to have a periodicity in order that their spike outputs become phase-locked to this periodic signal.
Published in the Proceedings International Joint Conference on Neural Networks (IJCNN'01 , Washington DC , 14-19 July 2001), Vol.1, pp.22-27, Piscataway, New Jersey : IEEE. [Full text] (PDF 159 Kb, 6 pages)
The mutual information between the phase of a periodic stimulus and the timing of the output spikes is examined for the leaky integrate-and-fire neuron, both with and without reversal potentials. The analysis is carried out in the Gaussian approximation and in the low output rate regime. The results indicate that, although the output demonstrates a higher synchronization than the input over a large range of parameters, this does not necessarily lead to an increase in the mutual information. The results in the subthreshold regime also shed light upon the role of stochastic resonance in such models.
(Paper to appear in CNS*2001 Proceedings)
We study the effect of noise upon the transmission of information about an input signal containing two frequencies in a leaky integrate-and-fire neural model. This work extends the results of earlier studies on stochastic resonance in neural models, and particularly in models of auditory processing. It is found that the temporal coding of two sub-threshold formants can be enhanced by the presence of noise. This study provides an approximation to the response to vowel speech stimuli and the results have a bearing upon the possible effectiveness of incorporating noise in cochlear implant speech coding strategies.
Published in the Proceedings of the IEEE Engineering in Medicine and Biology Society (EMBS) Conference, pp.21-24, Monash University, Australia, 19-20 February 2001. [Full text] (PDF 286 Kb, 4 pages)
The cause of peak-splitting in the output phase distribution of the leaky integrate-and-fire neuron model in response to low-frequency periodic nerve fibre inputs is analyzed. It is found that peak-splitting largely arises from an increase of the spiking-rate of individual nerve fibre inputs, or from an increase in amplitude of individual input excitatory postsynaptic potentials, or both. These findings add another dimension to the understanding of how peak-splitting arises in the phase histograms of the responses of neurons in the auditory pathway, given that peak-splitting is typically thought to arise as a result of the non-linear dynamics of the basilar membrane and the hair cells. This research has implications for the understanding of the temporal code in the auditory pathway.
Published in the Proceedings of the IEEE Engineering in Medicine and Biology Society (EMBS) Conference, pp.13-16, Monash University, Australia, 19-20 February 2001. [Full text] (PDF 171 Kb, 4 pages)
A model has been developed to determine the relationship between the output synchrony of cochlear nucleus neurons and the site of stimulation in the cochlea. This is an Integrate and Fire Neuron Model in which noisy periodic synaptic inputs to the neuron are summed and a spike is generated when the membrane potential reaches threshold. The model describes the stochastic input that auditory nerve fibres provide to a cochlear nucleus neuron and the corresponding stochastic output. To investigate the relationship between the output synchrony of cochlear nucleus neurons (namely globular bushy cells) and the site of stimulation in the cochlea, phase differences between the periodic inputs of the model were incorporated, in order to mimic how the travelling wave consecutively activates auditory nerve fibres originating over a spatial spread of the basilar membrane. Analysis of the model found that output synchrony decreased with an increase in frequency and spatial spread. Furthermore, enhancement of the output synchrony relative to the input synchrony occurred for small spatial spreads of the basilar membrane over which input primary afferent fibres originate. Adding noise helped to make the model more realistic. As a result enhancement of synchrony occurred with a spatial spread of less than 1.25 mm and 0.75 mm for 0.5 kHz and 1 kHz respectively, while for the higher frequencies analysed (2 kHz and 5 kHz) enhancement of synchrony did not occur. This research has implications for the design of electrode arrays in cochlear implants. The number and geometry of the electrodes and the stimulus patterns to be used will depend on the degree of convergence of fibres and how phase information is processed by neurons in the brainstem.
Abstract published in the Proceedings of the Australian Neuroscience Society 2001, Vol. 12, p.216.
This study examines neural mechanisms in the rat cochlear nucleus, in particular the degree of convergence of the input from auditory nerve fibres to globular and spherical bushy cells with specific reference to developing an improved cochlear implant electrode array and speech processing system. In male rats anesthetised with urethane (1.3g/kg i.p), microelectrodes containing 1M KAc, were inserted into the ventral cochlear nucleus (VCN). Intracellular recordings to sound have fast excitatory postsynaptic potentials (EPSPs) corresponding to the period of the sound wave up to 2.5kHz. Response to intracochlear electrical stimulation of auditory nerve fibres has demonstrated uniform conduction velocities for the inputs to these cells. The amplitudes of the EPSPs are stepped, indicating discrete synaptic inputs. A mathematical model has been developed to determine the relationship between output synchrony and the site and place of stimulation in the cochlea. This is an integrate and fire model which takes into account the stochastic nature of auditory nerve fibre firing statistics. The in vivo intracellular recordings obtained in the VCN have enabled the amplitude, number, time, course, synchrony and duration of post-synaptic events to be quantified, providing realistic parameters to the mathematical model. The model has established an upper bound on the distance along the cochlea over which input converges. This information is important in determining the geometry of an intracochlear electrode for processing the fine temporo-spatial patterns of responses for temporal coding with a cochlear implant. In summary this study demonstrates that it is feasible to use a new generation of electrode arrays with closely placed stimulating pads to transmit phase information associated with the maxima to give the best opportunity for speech processing strategies.
The hypothesis that the variability in the discharge of cortical neurons results from balanced excitation and inhibition is examined using neural models. In order to analyse neurons with significant inhibitory inputs it is necessary to incorporate an inhibitory reversal potential below which the membrane potential can not fall. A new method is presented for such a mathematical analysis of the leaky integrate and fire neural model with reversal potentials. The inputs are modelled as a Poisson process and the interspike interval distribution is calculated using a self-consistency relationship in the Gaussian approximation, which is most accurate when a large number of incoming postsynaptic potentials (PSPs) are summed. The mathematical results for the interspike interval distribution, output spike rate and coefficient of variation show close agreement with numerical simulations for large numbers of small amplitude PSPs. The dependence of the output spike rate upon the rate, number and amplitude of the synaptic inputs, as well as upon the value of the inhibitory reversal potential, are determined. One of the principal determinants of the output interspike interval distribution is the closeness of the firing threshold to the equilibrium value of the membrane potential in the absence of the spiking mechanism. The coefficient of variation of the output spike rate decreases with both increasing numbers and amplitude of inputs, and is consistently lower than 0.8 over a wide range of realistic values for the reversal potentials and the number and amplitude of PSPs, as observed in cortical neurons.
Abstract published in the Proceedings of the Australian Neuroscience Society 2000, Vol. 11, p.218.
The influence of spontaneous neural activity on the time coding of weak vowel stimuli in cochlear implants is studied using neural modelling techniques. The leaky integrate and fire neural model is used, in which the incoming postsynaptic potentials are summed and an outgoing action potential (AP) is generated when threshold is reached. A subthreshold stimulus, too weak to generate threshold crossings by itself, can combine with the spontaneous activity to enable the threshold to be crossed and APs to be generated. This nonlinear phenomenon is called stochastic resonance. Previous work on stochastic resonance has focussed upon sensory neurons driven by a continuous stimulus. Our model extends this by allowing the study of spike train transmission at any stage of the auditory pathway. The statistics of the membrane potential and the output spike train are derived using the theory of stochastic point processes and standard techniques based on the stationary phase distribution of the output APs. The analytical results and computer simulations show that ongoing spontaneous activity can enhance the transmission of a subthreshold polyperiodic stimulus (simulating a vowel stimulus), i.e., the output spike train contains the same frequencies as the input stimulus. Such a phenomenon could possibly be used in cochlear implants speech coding strategies to improve the time coding of weak vowel formants by artificially adding noise, which mimics spontaneous neural activity, to the electrical stimulus.
Abstract published in the Proceedings of the Australian Neuroscience Society 2000, Vol. 11, p.215.
The effects of action potential (AP) propagation delay times and the absolute refractory period upon the synchronization index are analysed for the integrate and fire neuron model, and the results are compared with recordings from auditory ganglion neurons and cochlear nucleus neurons. In the model the noisy periodic synaptic input to the neuron is summed and an AP is generated when the membrane potential reaches threshold. The output phase distribution (phase histogram) is calculated at the site at which the APs are generated. The AP propagation delay times along an axon are modelled using a periodically wrapped Gaussian distribution, with the width fitted from experimental data. This distribution is convolved with the calculated phase distribution to obtain the phase distribution at the axon terminal. The model is implemented using the parameter values for the membrane time constant and the refractory period of both auditory ganglion neurons and cochlear nucleus neurons. It is found that the synchronisation index of the output APs decreases rapidly at high frequencies of the input (greater than 1 kHz). Inclusion of an absolute refractory period to the original model increases the interspike intervals, and the resultant reduction of the synchronization index is most pronounced at higher frequencies of the input. The computed phase distributions of the model show close agreement with experimentally recorded phase histograms.
Abstract published in the Proceedings of the Australian Neuroscience Society 2000, Vol. 11, p.145.
Delay analysis is a method for analysing phase-locked responses to periodic stimuli which is widely used in the study of auditory cells, as it provides an estimate of the delay present in a system from steady-state data. While the usual formulation utilises the assumption that the delay is constant across frequencies, in the auditory system delay varies with frequency. In this paper two new formulations of delay analysis are introduced, and are applied to the analysis of auditory temporal coding. The computed delays correlate with another estimator of delay, click response latency (r=0.71, p<1e-26), for data from the auditory nerve, cochlear nucleus neurons, and units in the trapezoid body. Mean delays (with standard error) computed for these locations are 2.42±0.06ms; 3.20±0.04ms; and 3.59±0.06ms. Using the introduced techniques, an analysis of the temporal code at these locations is carried out, involving the vector strength, delay and phase of response. It is demonstrated that there are deviations from the constant delay approximation, and the implications discussed in terms of basilar membrane properties and neural processing. Data were obtained from 511 units in rats anaesthetised with urethane (1.3g/kg i.p.). Recordings were made extracellularly in vivo using glass microelectrodes filled with 1M potassium acetate (50-70MW).
Abstract published in the Proceedings of the Australian Neuroscience Society 2000, Vol. 11, p.216.
We hypothesised that the loss of the peripheral processes and the partial demyelination of auditory nerve fibres (ANFs) following a sensorineural hearing loss would increase their refractory properties. Normal control, and long-term (2.5 months) systemically deafened rats were anaesthetised (urethane, 1.3 g/kg i.p.), a bipolar stimulating electrode was implanted into the scala tympani and glass microelectrodes (30-80 MW) used to record single ANF activity. Stimuli (pairs of 100 ms/phase charge balanced biphasic pulses with interpulse intervals (IPIs) of 0.34-10 ms) were presented at 6 dB above threshold using a repetition interval of 250 ms. Absolute refractory period (ARP) was defined as the IPI at which the probability of eliciting a spike to the second stimulus was 0.1. In the present results, based on recordings from 62 fibres, ANFs were distinguished from cochlear nucleus (CN) neurones by their significantly shorter median latencies (AN: 0.575ms vs CN: 1.137ms; Whitney-Mann Rank Sum, p<0.0001). There were no significant differences between minimum ANF latencies from normal and deafened animals. Although the median ARP was greater in deafened versus normal animals, this difference was not statistically significant (normals: median 0.658ms, interquartile range 0.554-0.913ms; deafened: 0.772ms and 0.616-1.073ms; p=0.16). Finally, the spike latency associated with the second pulse of a pair systematically increased with decreasing IPI, contrasting with the stable latency of the response to the leading pulse. Although pathological changes to ANFs may increase their refractory properties, at this duration of deafness these changes were not significant.
Abstract published in the Proceedings of the Australian Neuroscience Society 2000, Vol. 11, p.144.
The relationship between the timing of noisy periodic synaptic inputs and the output spikes in leaky integrate and fire neurons is investigated using the new integrated-input technique. The output spike density as a function of input phase is calculated, enabling the output interspike interval distribution and phase distribution to be evaluated. The interspike interval distribution shows the typical multimodal response, and the synchronization index of the output spikes is found to be larger than the input synchronization index over a range of frequencies and neural parameters, this enhancement of synchronization being more pronounced for larger numbers of inputs and lower frequencies.
(Paper to appear in CNS*99 Proceedings 1999)
The hypothesis that the variability in the discharge of cortical neurons results from a balance of excitation and inhibition is analyzed using the new integrated-input technique. The method is extended to include reversal potentials and enables the interspike interval distribution to be calculated in the large N limit (number of inputs). The output spike rate increases monotonically over two orders of magnitude, solving the dynamic range (or gain control) problem. The coefficient of variation is approximately 1.0 for low input rates and increases to around 1.6 at high input rates, well within the range observed in the response of cortical neurons.
(Paper to appear in CNS*99 Proceedings 1999)
The relationship between the timing of the synaptic inputs and the output spikes of leaky integrate and fire neurons with noisy periodic synaptic input is addressed using the recently developed integrated-input technique. The conditional output spike density in response to noisy periodic input is evaluated as a function of the initial phase of the inputs. This enables the phase transition matrix to be calculated, which relates the phase at which the output spike is generated to the initial phase of the inputs. The interspike interval histogram and the period histogram for the neural response to ongoing periodic input are then evaluated by using the leading eigenvector of this phase transition matrix. The dependence of the synchronization index of the neural response upon the number and amplitude of synaptic inputs, the membrane time constant, the average rate of inputs and their frequency of modulation is examined.
Published in the Proceedings International Joint Conference on Neural Networks (IJCNN'99 , Washington DC , 10-16 July 1999), Vol.1, pp.268-273, Piscataway, New Jersey : IEEE.
The observed variability in the spike rate of cortical neurons has been hypothesized to result from a balance in the excitatory and inhibitory synaptic inputs that the neurons receive. The coefficient of variation of the spike rate is calculated in the limit of a large number of inputs using the integrated-input technique, which is extended here to include the effect of reversal potentials. The output spike rate is found to increase monotonically over two orders of magnitude, thereby solving the dynamic range (or gain control) problem. The coefficient of variation is approximately 1.0 for low input rates and increases to around 1.6 at high input rates, well within the range observed in the response of cortical neurons.
Published in the Proceedings International Joint Conference on Neural Networks (IJCNN'99 , Washington DC , 10-16 July 1999), Vol.1, pp.305-308, Piscataway, New Jersey : IEEE.
There are significant differences in the responses of auditory nerves when they are stimulated acoustically (normal hearing situation) or electrically (with a cochlear implant). We address the underlying causes of these differences by studying the interspike interval histogram, the synchronization index, and the entrainment (ability to respond at successive cycles of the stimulus). The new integrated-input technique is used to analyze the response to periodic synaptic input of integrate-and-fire neurons, in which the randomly arriving synaptic inputs are summed and an action potential is generated when the postsynaptic potential reaches threshold. The synaptic inputs in the model are a sinusoidally modulated inhomogeneous Poisson process, and each input generates a characteristic postsynaptic response that subsequently decays according to the membrane decay constant. The results provide a quantitative understanding of both the decrease of the synchronization index with increasing frequency of acoustical stimulation in the auditory pathway and the previously observed enhancement of synchronization in globular bushy cells of the cochlear nucleus. The differences in the responses of neurons in higher stages of the auditory pathway for acoustical and electrical stimulation may be accounted for by the differences in the degree of entrainment that they induce.
Published in the Proceedings of the IEEE Engineering in Medicine and Biology Society (EMBS) Conference, pp.46-49, Monash University, Australia, 22-23 February 1999
The response of a neural system to periodic synaptic input is analyzed using the new integrated-input technique, in which the randomly arriving synaptic inputs are summed and an action potential is generated when the postsynaptic potential reaches threshold (i.e., integrate-and-fire neurons). The results provide a quantitative understanding of the decrease of the synchronization index with increasing frequency of acoustical stimulation in the auditory pathway. The conditions under which the observed enhancement of synchronization in globular bushy cells of the cochlear nucleus may occur are elucidated. The dependence of the synchronization index upon the frequency of stimulation is analyzed for a range of neurophysiological parameters, including the number of afferent fibres, the membrane time constant, the amplitude of the individual postsynaptic potentials, timing jitter in the axonal propagation of spikes, and the time course of the postsynaptic response. The inclusion of an absolute refractory period to the model is found to have little effect upon the period histogram. The interspike interval histogram and the period histogram for the neural response to ongoing periodic inputs are evaluated using the stationary solution to the phase transition matrix, which relates the phase at which the output spike is generated to the initial phase of the inputs. Spontaneous activity is included in the model, and the possible role of stochastic resonance in threshold detection is addressed.
Abstract published in the Proceedings of the Australian Neuroscience Society 1999, Vol. 10, p.173.
The degree of phase locking in response to noisy periodic input plays an important role in auditory processing, where studies indicate that spikes in the auditory pathway are phase locked up to frequencies around 3-5 kHz in mammals and up to frequencies of 8kHz in the Barn Owl. Phase locking has also been postulated to play a central role in temporal coding in the brain, where it may be used in feature linking and pattern segmentation. The results will be presented from an investigation of the relationship between the timing of noisy periodic synaptic inputs and the output spikes that are generated by leaky integrate and fire neurons. A new method of analyzing integrate and fire neural models is used, which allows both the interspike interval distribution and the synchronization index in response to the noisy synaptic input to be calculated.
A new technique for calculating the probability distribution of output spikes for integrate-and-fire models is presented. It proceeds by integrating over the distribution of arrival times of the afferent postsynaptic potentials (PSPs), rather than using conventional stochastic differential equation techniques. For synchronized inputs, the relationship between the temporal jitter of input and output spikes is evaluated. Poisson distributed inputs are analyzed and a solution to the Stein model, incorporating membrane leakage, is given. PSPs with varying amplitudes, including inhibitory PSPs, are also analyzed. The results, which are compared with numerical simulations, are exact for large numbers of small amplitude PSPs.
Published in Neurocomputing 26-27 (1999) 93-99.
We present a new technique for analyzing the probability distribution of output spikes for the integrate and fire model. Using this method we investigate models with arbitrary synaptic response functions and the results, which are compared with numerical simulations, are exact in the limit of a large number of small amplitude inputs. We apply this method to the synchronization problem, in which the relationship between the spread in arrival times of the inputs (the temporal jitter of the synaptic input) and the resultant spread in the times at which the output spikes are generated (output jitter) is analyzed. The results indicate that the ratio of the output jitter to the input jitter is consistently less than one and that it decreases for increasing numbers of inputs, in agreement with earlier studies. We identify the variation in the spike generating thresholds of the neurons and the variation in the number of active inputs as being important factors that determine the timing jitter in layered networks, in addition to those identified previously.
Published in Australian Journal of Intelligent Information Processing Systems 5 (1998) 50-54.
This page, its contents and style, are the responsibility of the author and do
Author Anthony N. Burkitt
Last modified Friday June 02 2006
not represent the views, policies or opinions of The University of Melbourne