Skip to main content

Representation Of Stimuli As Neural Activity

Deliverables

A polyimide based, flexible, 27 to 54 channel bi-directional, implantable and regenerative type sieve was developed for application to investigate animal behaviour. The materials used for processing the sieve electrode was tested for biocompatibility according ISO 10993 required standards. The supplies of the sieve electrode include several types of adapters and connectors. Primary investigations showed that the sieve electrode was capable to deliver physiological signals from the animal model. The advantage of a sieve electrode for recorded signals is the high selectivity. There are several results associated with the regeneration of neurons showing the modification of nerve anatomy after regeneration in comparison to the original nerve, with this type of electrode. The electrode could be used basic research.
Three versions of tactile stimulator have been developed for use in neurophysiological investigations on animals: First version contactors: 12 arranged in a square matrix contactor size: 0.6mm diameter contactor separation: 1.2mm between centres maximum displacement: 3.0 ± 0.5mm for 80V input Final version A contactors: 15 arranged in a hexagonal matrix contactor size: 0.6mm diameter contactor separation: 1.5mm between centres maximum displacement: 5.0 ± 0.5mm for 68V input (± 5.0mm for ± 68V input range) Final version B contactors: 5 arranged in a hexagonal matrix contactor size: 0.6mm diameter contactor separation: 1.2mm between centres maximum displacement: 4.0 ± 0.5mm for 68V input (± 4.0mm for ± 68V input range) Associated control hardware and software has also been developed. All three versions of the stimulator have been used successfully in investigations on rats. Their performance is a close match to their design specifications. The ROSANA stimulators produce well-specified spatiotemporal patterns of tactile stimulation which are easy to control and modify. These stimulators offer considerable scope for neurophysiological investigations in the future. At present there is significant interest in the development of tactile stimulators for use as an enhancement to virtual environments. Technological developments from the ROSANA project have been incorporated into ongoing work for two other EU-funded projects: HAPTEX (co-ordinated by the University of Geneva); this is concerned with the development of virtual textiles; UNEXE is developing tactile stimulators to produce virtual texture. ENACTIVE (coordinated by the University of Pisa); this is concerned with multimodal computer interfaces; UNEXE is developing tactile stimulators to provide information on contact area, edges, corners and texture on the fingertip.
A new connector concept for easy connection of flexible electrodes was developed. The whole connector based of a thick film ceramic substrate with gold contacts. An elastic clamp creates the connection pressure and the force is distributed by an elastic piece of silicon on the clamp. A flexible cable connection to the ceramic by using different types of connector plugs according to the application requirements are possible. This system is adaptable for different types of flexible electrodes and numbers of single electrodes. By using a fixation system the electrode can be easy mounted on the connector and easy removed as well.
Electrophysiological experimental data from extra-cellular recordings can be considered as time series of discrete events (spike trains) generated by several neurons, without any direct information on the network structure, type of neurons, synapses, etc. Thus the first step therefore, is the spike sorting on the basis of the extra-cellular recordings. The electrode configuration (four probes placed in a rhomboidal geometry) allows recording the spike waveform from different perspectives. This information is used to characterize with more detail the action potential and improve the use of Principal Component Analysis, PCA. To solve some of the limitations of PCA we have proposed a new method that makes use of Independent Component Analysis, over spike scores found by PCA. We also used another technique, based on wavelet transform developed by our group. The combination of ICA together with PCA allows solving those cases where clustering is difficult even with the simple combination of features from different electrodes and different clusters (identified later with the spike density plots) are mixed up. For example in figure 1 we can see that that before ICA transformation clusters are vaguely separated (there are no clear peaks in the distribution), however, after ICA we perfectly see two peaks supporting the power of the proposed method. Another important but frequently underestimated problem we considered here is the appropriate waveform filtering. The wideness of this problem can be understood just mentioning that despite of the methods used for features extraction and clustering any electrophysiological recording is initially filtered by a hardware or/and software. The frequency band of the filter is usually taken subjectively without attention to the peculiarities of the method of spike features extraction subsequently used. Here we analysed this problem providing a recipe for each of the considered techniques.
The fourth objective and one of the main tasks of UCM was the development of reliable models of the internal representation of real-world stimuli in early stations of the somatosensory system, mainly based on the signal processing performed by oscillating neurons, and the rhythmic interactions existing among neurons belonging to local networks in those stations. The brain is a non-linear dynamic system and just as physical systems can be characterized as evolving across different regimes (periodic, quasi-periodic, chaotic, etc.) under the control of one or more system parameters, experimental data indicate that neurons and neural ensembles also pass through a repertoire of regimes under the control of driving inputs. Within the brain, neural ensembles have intrinsic oscillatory activity and there is also evidence that real neural systems operate near instability which confers them the ability to respond rapidly and with a large flexible repertoire of sensory and motor patterns. Neural ensembles devoted to pattern recognition are synchronized by precisely timed synaptic inputs and in this context we proposed to develop mathematical models of the interactions between incoming electrical stimuli and the oscillating neural activity in the populations of neurons involved in the perception of the incoming stimuli. The study of the nervous system and more concretely its availability for computation depends on both the intrinsic properties of single neurons and on their interactions since individual properties of the neurons play less significant role than the complexity of the neuron ensembles they form. With this idea in mind we mechanically stimulate the glabrous part of the hind limb of the rat and record the extra-cellular activity of the Central Nervous System. This way we obtained spike trains representing the simultaneous activity of a group of neurons. Then all spike trains have been statistically analysed and different stimulus response patterns have been identified. Next we analysed the connectivity among the neurons and patterns changes when different stimuli are applied. We performed identification by the method developed in our lab (INCAM). It provides the topology of the network in different stimulation conditions and allows modelling the neural dynamics. The connectivity pattern and the functional characteristics of the sub-networks that we have been able to identify provide a good basis for the study of information processing conducted in the first relay station of the somatosensory pathway. In the present work we analysed and modelled a large number of experimental recordings of neural activity in the somatosensory pathway and identifying networks of more than 4 neurons. Our new results show more specialised functional networks for the processing of tactile information and that the composition of such networks correspond better to the division of neurons into the two experimentally observed types of "fast" putative inhibitory inter-neurons and "slow" putative excitatory projecting neurons. In a preliminary work on mathematical models we developed functional networks using 3-recorded neurons and we identified different connectivity patterns. In our final work we inferred on the neural circuits in spontaneous conditions and for each stimulation point, using our INCAM method with 4 or more recorded neurons. In figures 5 to 9 we present the functional circuits built by 4 cells that correspond to the neural responses recorded after stimulation to the 4 loci in the hind-paw of the rat and the at rest recorded activity for such cells.
We could determine the magnitude and partial selectivity of fibre types, of axonal regeneration through the sieve electrode in rats sustaining implants in transected sciatic nerves. This study correlated functional results of the regeneration achieved with quantitative assessment of those neurons responsible of the regeneration selectively labelled by retrograde tracing.
A handheld stimulator was designed for easy manageability and fast investigation performance. This stimulator was designed to have several features like bipolar stimulation current, voltage controlled stimulation and different stimulation amplitudes. The stimulator is able to stimulate in single or continuous mode with rectangular pulse shapes and different pulse widths with two different stimulation frequencies (20Hz, 100Hz). The slim line design of the handheld stimulator and its battery operation makes it very flexible to use in nearly every place to stimulate appropriate electrodes.
In order to test the mathematical models developed by UCM we had foreseen to inject artificial electrical stimuli direct to the peripheral nerve that would generate activation patterns in the CNS similar to those of the natural stimuli. In this context the bioelectronic interfaces developed in the framework of the First Objective were used for the generation of stimulation patterns similar of the natural ones and the stimulation of the CNS in a natural way. As an outcome we presented the models of the neural circuits responding to specific electrical stimulation patterns delivered to the peripheral nerve through the sieve electrode and the comparison with the neural circuits obtained by analysing the neural responses to mechanical stimulation of the skin. Consequently we generated electrical patterns in the peripheral nerve, equivalent to the natural ones, capable of stimulating the neural populations of the somatosensory system (DCN and Cx) in a way similar to that of natural stimuli.
Electrophysiological data on interactions between incoming external signals (natural or artificially generated) and the electrical activity of populations of processing neurons in DCN and somatosensory cortex have been obtained by means of single-electrode and multi-tip needle electrodes (Michigan Probes with 16 recording channels: two shanks (3mm long), separated by 150µm, with four tetrodes with recorded points separated by 25µm). Neural dynamics involved in tactile stimuli perception have been studied using the advanced stimulator designed and produced by UNEX with very interesting results for both, DCN and somatosensory cortex.
We focussed a good deal of effort on evaluating in quantitative terms the actual regeneration of axons taking place through the electrode. We first collected data to establish a large database from intact -control- nerves, determining the normal range of variability for the various subcomponents of the nerve and fibre types. We found a sizable variability in the regenerating that took place, both in the rat and cat models, and identified the main factors -including technical limitations of the implant- that could be involved in such variability. We then established the numbers and main morphometric features of myelinated and unmyelinated axons that were present at the implant site. Finally, we obtained valuable data on the molecular phenotypes (and its changes) of the regenerating axons.
We assessed the changes in a variety of molecular features of the target territories in the central nervous system (spinal cord, brain stem and sensory cortex) consequent to the implant and eventual nerve regeneration through it. We found that permanent transection of nerves, keeping them from regenerating, produces detectable changes in the molecular markers used in some or all of the target regions, specifically in those subregions representing the territory affected by the denervation. Most changes consisted in down-regulation or lower expression of the marker, but exceptions occurred, as is the case of over-expression of parvalbumin in the deafferented DCN, which built up with survival time. When transected nerves regenerated succesfully through the sieve electrode most, but not all, of the changes were reverted -or did not occur altogether-. There seems to exist a direct relationship between the actual number of regenerating sensory fibres through the electrode and the expression of some markers in the target regions. This relationship is probably complex, because it likely depends on the relative regenerative success of each kind of fibre subpopulation An additional relevant conclusion was that the number of motoneurons in the ventral horn of the rat spinal cord never regained a normal structure, as revealed by Nissl (fewer motoneuron bodies) and AChE and ChAT (lower staining of cholinoceptive/cholinergic and cholinergic markers, respectively), indicating less than optimal motor recovery. Finally, very long postimplantation survival times in cats showed fairly normal patterns of molecular expression at "early" stations of sensory processing, but various degrees of disorganization or down-regulation of the same markers in the somatosensory cortex, suggesting that regenerating failed to bring about nearly-complete recovery of the reorganized cortex: PV, CB or SS immunostaining was apparently normal in the neurons and neuropil of the dorsal horn of the spinal cord and the DCN, in rats and cats, after regeneration. In contrast, these three markers displayed various degrees of disorganization or down-regulation in the somatosensory cortex.