Advanced

Method to measure the quality of a neuron model compared to recorded data

Flodin, Oscar (2016)
Department of Automatic Control
Abstract
In this thesis, a method for evaluating the accuracy of mathematical neuron models is developed, with the objective to both improve the quality and the speed of parameter optimization. To evaluate the method, we used a conductance based model of the cuneate nucleus. The model behavior was evaluated against neurons recorded in vivo. A big challenge when modeling neurons is the timing of the action potentials, since the initiation is stochastic. Therefore, the model artificially inserts the spikes at the recorded spike times. The accuracy of the model is evaluated as its capacity to capture the time evolution of the membrane potential both following and preceding the spikes. The evaluation method uses the L2-norm when comparing a few... (More)
In this thesis, a method for evaluating the accuracy of mathematical neuron models is developed, with the objective to both improve the quality and the speed of parameter optimization. To evaluate the method, we used a conductance based model of the cuneate nucleus. The model behavior was evaluated against neurons recorded in vivo. A big challenge when modeling neurons is the timing of the action potentials, since the initiation is stochastic. Therefore, the model artificially inserts the spikes at the recorded spike times. The accuracy of the model is evaluated as its capacity to capture the time evolution of the membrane potential both following and preceding the spikes. The evaluation method uses the L2-norm when comparing a few important time windows of the recording and the model. The method was used to optimize the parameters of the model. Sensitivity analysis is used to learn which model parameters affect the model accuracy the most. The results from the optimization showed that the developed method well describes the quality of the neuron model for different parameters. This tool will make it easier to build models that capture the responsive properties of the membrane potential in neuronal in vivo recordings. (Less)
Please use this url to cite or link to this publication:
author
Flodin, Oscar
supervisor
organization
year
type
H3 - Professional qualifications (4 Years - )
subject
report number
TFRT-6005
other publication id
0280-5316
language
English
id
8887642
date added to LUP
2016-08-17 09:32:30
date last changed
2016-08-17 09:32:30
@misc{8887642,
  abstract     = {In this thesis, a method for evaluating the accuracy of mathematical neuron models is developed, with the objective to both improve the quality and the speed of parameter optimization. To evaluate the method, we used a conductance based model of the cuneate nucleus. The model behavior was evaluated against neurons recorded in vivo. A big challenge when modeling neurons is the timing of the action potentials, since the initiation is stochastic. Therefore, the model artificially inserts the spikes at the recorded spike times. The accuracy of the model is evaluated as its capacity to capture the time evolution of the membrane potential both following and preceding the spikes. The evaluation method uses the L2-norm when comparing a few important time windows of the recording and the model. The method was used to optimize the parameters of the model. Sensitivity analysis is used to learn which model parameters affect the model accuracy the most. The results from the optimization showed that the developed method well describes the quality of the neuron model for different parameters. This tool will make it easier to build models that capture the responsive properties of the membrane potential in neuronal in vivo recordings.},
  author       = {Flodin, Oscar},
  language     = {eng},
  note         = {Student Paper},
  title        = {Method to measure the quality of a neuron model compared to recorded data},
  year         = {2016},
}