• No se han encontrado resultados

Study of sequential information processing in electroreception through modelling and closed-loop stimulation techniques

N/A
N/A
Protected

Academic year: 2023

Share "Study of sequential information processing in electroreception through modelling and closed-loop stimulation techniques"

Copied!
172
0
0

Texto completo

Representative example of distinct current inputs in the simulation at different locations on the graph (Right). Inputs are given to the model as a function (f) of SPIs recorded from the biological system.

Thesis structure

From a neuroscientific perspective, studying the signaling behavior of fsh is closely related to studying the mechanics of the electromotor command system. Chapter 11 talks about the main conclusions of the thesis, obtained from the development of TCDS, its use for active electroreception in weakly electric fsh of the pulse type, and the development of a computer model of the electric motor system.

Aim and main objectives

Neurons

In the 1950s, electron microscopy also revealed the existence of gaps between neurons, which had previously been suggested by the presence of synaptic delays in neuronal communication. An action potential is a rapid voltage transient in the cell membrane where it rapidly changes from a negative voltage value to a positive value (depolarization) and then back to negative values ​​(repolarization), reaching a greater negative voltage than that from the original resting state (hyperpolarization) ( see Fig. 2.2).

Synapses and neural networks

At the end of the axon, the action potential triggers the release of neurotransmitters near the membrane of the target neuron. There is a delay of 1 to 5 ms since the action potential in the presynaptic neuron arrives at the synaptic cleft until an effect is induced in the postsynaptic membrane.

Neurons and synapses as dynamical systems

Temporal sequences in sensory-motor systems

Temporal coding in the nervous system states that information is encoded in the precise timing of spikes (Theunissen & Miller, 1995). Nevertheless, information can be equivalently encoded in the interval between event times (which is mainly the approach used in this thesis).

Information theory and applications

Closed-loop stimulation techniques provide an experimental viewpoint that is much more similar to how the nervous system works in the real world. Some brilliant approaches in the past have exploited the inherent closed-loop nature of the nervous system to mimic closed-loop stimulation techniques and reveal circular processes in the nervous system.

Real-time software for closed-loop experimentation using general-purpose

More details on these and other real-time solutions for conducting closed-loop investigations can be found in the next section. Real Time Application Interface (RTAI)4 is an RTOS based on a patch for the Linux kernel.

Generalization of the dynamic clamp concept: Neurophysiology and

Regarding weak electrical fsh, Madhav et al. 2013) applied closed-loop stimulation to generate a signal with the same frequency as that detected by an Eigenmannia virescens specimen - a sinusoidal electric wave Fsh -. In the case of mormyrids, the electrical signal is directly controlled by a neural network located in the fsh central nervous system, the electromotor system (Caputi et al., 2005; Carlson, 2002a).

Modeling the mormyrids signaling behavior

Electrophysiology of the electromotor command network

The neural system responsible for controlling the timing of the EODs is the electromotor system (see Fig.4.3), located at the central nervous system of the fsh (Caputi et al.,2005). The intrinsic dynamics of the constituent neurons and the network topology of such CPGs are well known.

Mormyrids signaling and SPI generation

The electrical signal from the fsh was picked up by four differential dipoles placed in the water tank. Upon detection of the predefined trigger code, stimulation is delivered to the system using a 13 cm dipole placed in the center of the water tank (stimulator).

Ethical note

Output from the print is connected to the input channel of the DAQ board and obtained by the computer in real time. This stimulator is connected via a transformer that prevents reverse currents to the output channel of the DAQ.

Experimental framework

RT software framework

The hardware characteristics of the computer used in the experiments are described in the previous section 5.2. New considerations during the development of this thesis led us to also consider the Real-Time Experimental Interface (RTXI, Real-Time Framework) as an interesting possibility to increase the diffusion of the protocol.

SPI patterns from Gnathonemus petersii recordings

Like other closed-loop techniques, TCDS allows the stimulation to be delivered and adjusted depending on the system's activity. To overcome this problem, we choose the bin time ∆t that maximizes the entropy of the binary signal (Jaynes, 1957).

TCDS implementation

Functional requirements

This section defines the functional requirements of the implementation, i.e. a description of the services that the software must provide. Shows instant (in a given time window) and accumulated word histograms of the system's ongoing activity.

Stimulation protocols

The open-loop protocol allows us to compare changes in the system due to stimulation triggered by prior fsh activity (closed-loop) with stimulation that ignores fsh activity (open-loop). It is important to note that in the open-loop protocol the stimulation does not trigger the activity of the biological system (Table 6.2).

Quantifying changes in IPI distributions

In the exemplary case shown in Figure 7.7-(I), the cumulative distributional difference is always positive, since at any given point the shorter IPIs are more (or ultimately equally) more likely than the first distribution. As a result, IPIs below 110 ms are more likely in the first distribution (mainly because there are almost no IPIs below 100 ms in the case of the second).

Synapses model

Adjusting the parameters of these synapses is essential to generate the four types of SPI patterns exhibited by the electromotor command network. An ad hoc iterative tuning of the parameters of the model was performed to match previously described dynamics in the electromotor command network: (i) DP/PCN units that break sporadically before CN, (ii) DP/PCN that remain silent for tens to hundreds of milliseconds after an action potential from CN and (iii) VPd edge burst of action potentials during DP/PCN quiescence starting ≈1–8 msec after CN activation.

Target SPI patterns: synthetic and recorded

The topology of the model was set up using a standard configuration of the model of chemical synapses (EDP, EPCN,IDP, IPCN, ECDP in Fig. 8.1). The inherent complexity of manually tuning all the parameters to reflect these dynamics shown by the real network led to the development of an automatic method for tuning synaptic parameters in the model.

Model parameters

As described in Section 8.6.1, during evaluation, the SPIs were normalized to the same duration (1000 arbitrary units) and linearly interpolated (every 20 arbitrary units, for a total of n = 50 points for each model) for characterized the shape of the model.

Simulation parameters

Automatic adjustment of synaptic parameters: Genetic algorithm

Fitness function

Finally, the evaluation procedure calls evalPattern for each of the four patterns (scallop, acceleration, rasp, and stop) and adds them to return the overall individual's ftness value (ftValue, equal to f(I) in Eq. 8.6). First, TCDS using 2-bit codes representing the minimum sequential activity of the fsh was applied to validate TCDS.

Comparative of control sessions

Subsequently, TCDS with all 4-bit codes ending in 11 were applied to study the response to codes representing a longer sequence of activities. Finally, the open loop protocol was applied for assessing differences from the results obtained by 4-bit code-driven stimulation.

TCDS results using 2-bit trigger codes

This code is related to pulse generation under control (due to maximum entropy criterion for choosing ∆t). The second code (11) represented two consecutive time periods emitting impulses, this acceleration in the generation of impulses is related to the active fsh probing the environment or searching for social response.

TCDS 4-bit codes

Robustness to input variability

Representative example of distinct current inputs in the simulation at different locations on the graph (Right). -T moderately improved endurance scores when an increase in the duration of inputs was offset by a decrease in input intensity.

Sensitivity of the robustness analysis

FIGURE 10.6: Mean and variance of simulated SPIs using the R-GA model under different simulation conditions of the robustness analysis (see Fig. 10.4). This result shows that there is considerable sensitivity to the relative value of ftting used in the robustness test, as significant drops in ∆f occur due to corresponding changes in the time structure of the resulting SPI patterns (Fig. 10.7 compared to the target patterns of recorded in Fig. 5.9).

Continuity between S-T and S-GA confgurations

In conclusion, the robustness test is able to detect strong changes in the temporal structure of SPI patterns. The most severe decreases in relative ftting value occur with changes in synaptic parameters, rather than changes in the duration or intensity of inputs.

Modifed network topologies

  • Homogeneous phasic spiking neurons
  • No corollary discharge
  • No PCN pathway
  • Additional considerations about the stimulation in TCDS gen-
  • RTXI-based implementation of TCDS

FIGURE 10.11: Simulation of the four SPI patterns in the best fit model without corollary discharge path. FIGURE 10.12: Simulation of the four SPI patterns in the best fit model without PCN core pathway.

Future work applying TCDS in the study of electroreception

Modeling the electromotor command network

Our computational model of the electromotive command system (Lareo et al., 2018; Lareo et al., 2022b) enables studies that better simulate social contexts. The resting rhythm of the system (IPIs ranging from ∼100-300 ms, following a bimodal distribution) can be easily reproduced by the model using biologically plausible input values.

Integrating the electromotor model in real-time TCDS

Morphology and physiology of brainstem nuclei controlling electrical organ discharge in the mormyrid fsh. Electrical and motor responses of the weak electric vole, Gnathone-mus petersii (Mormyridae), to playback of social cues.

TCDS version using RTBiomanager (RTAI)

TCDS version using RTXI (Xenomai)

Library: binwords

Stores the decimal values ​​of the code words obtained from the processed binary signal. The histogram of the code words stored in the WordsBuffer after processing the binary signal can be obtained using the wbCreateHistogram function.

Preprocessing

When a new word is completed after the arrival of its last bit, the contents of this buffer are used to calculate the word's decimal value, which is stored in another data structure (WordsBuffer). Overlap between consecutive words can be controlled by setting wb->overlap to the desired number of overlapping bits between words.

Electromotor model and ftting GA

Compilation and install

Library: neun

This appendix shows, separated by sample fsh, the results of all the TCDS application experiments that are listed in Table 5.2. In both, (A) Frequency of IPIs in control sessions —no stimulation—, 01 stimulation session and 11 stimulation sessions;.

Results from 4-bit trigger codes TCDS disaggregated by experimental

In both, (A) Frequency of IPIs in control-no-stimulation sessions, 01 stimulation sessions and 11 stimulation sessions; (B) Cumulative IPI frequency difference between control and code stimulation sessions (01-Control, 11-Control); (C) 11–01 accumulated IPIs frequency difference. D) Frequency of IPIs in control-no-stimulation sessions and 1 stimulation session; (E) Cumulative IPI frequency difference between control and code 1 stimulation sessions (1-control); (F) 11-1 and 01-1 accumulated IPIs frequency differences. In both, (A) Frequency of IPIs in control-no-stimulation sessions, 01 stimulation sessions and 11 stimulation sessions; (B) Cumulative IPI frequency difference between control sessions and code stimulation sessions (01-Control, 11-Control);.

Contributions to international congresses

This appendix provides a list of publications indexed in JCR journals and international congresses during this thesis, and relates them to their corresponding chapters.

Referencias

Documento similar

In addition, it was also found that RecN and RecA proteins were able to modulate the activities of PNPase, through the stimulation of polymerase or

While in random data sets, the mean of the frequency dis- tribution is 23 pairs of proximal genes, the observed frequency in our experimental data set is 41.. The probability

Furthermore, we did not find evidence of entrainment of a neural oscillation at the stimulation frequency, as the shape of the transient ERPs obtained from jittered sequences

∆t p seconds where s = [0, ∞), and the index p denotes the time placement of the window measured in seconds from the beginning of the trial. For instance, in Fig. To determine whether

a) The original precalculated regulator, which applies a precalculated duty cycle controller without any frequency correction. b) The frequency loop modifying the duty cycle vector

The estimated phosphene size evoked by single electrode stimulation was usually very small (0.8 ± 0.8; mean ± SD of sub- ject’s size estimates) and resembled “pin points” of light

Apoptosis of osteocytes and osteoblasts precedes bone resorption and bone loss with reduced mechanical stimulation, and receptor activator of NF- ␬ B ligand (RANKL) expression

Figure 4 Normalized mean square error of the training data set for the OLS, k-means clustering, and random selection of centers procedures (frequency of the electric scalar