Year: 2020
Type: dissertation
Abstract: A current challenge within neuroscience is to model the dynamics of biological networks.This thesis explores a model of the mouse primary visual cortex (V1) and lateral geniculate nucleus (LGN), developed by the Allen Institute for Brain Science.Spatial and temporal tuning properties of model cells are analysed.Of main focus is two phenomena which have been observed in mice LGN and V1.The first, surround suppression, refers to the neural response to a visual stimulus being inhibited by a surrounding stimulus.The second is band-pass tuning in response to stimuli of various spatial frequencies.The original Allen model fails to exhibit both surround suppression and bandpass tuning.It is therefore explored how the model could be improved, specifically by modifying the LGN.To represent the center-surround organisation of receptive fields found in mice, the Gaussian spatial filters of the LGN are replaced with Differenceof-Gaussians (DoG).As a result, the modified model reproduces band-pass spatial frequency tuning for grating stimuli and surround suppression for uniform stimuli.However, no clear suppression is shown in response to patches of grating stimuli, inconsistent with both previous findings in mice V1 and the LGN and V1 of other mammals.Surround suppression is thought to incorporate both linear and nonlinear mechanisms, where only the linear response can be accounted for by the DoG function.How nonlinear suppression arises in the cortex is not yet fully understood, though one potential source is input from LGN.Previous studies have suggested that responses in the visual pathway are effectively normalised through neural inhibition.To model this, the Allen LGN is modified so that firing rates are normalised with respect to overall activity.Suppression is thus reproduced in both LGN and V1 for patch grating stimuli.However, the method has limitations when it comes to accounting for all aspects of V1 suppression, such as time delay and dependence on stimuli orientation.Another potential source of V1 suppression is network effects from within cortex.Recurrent connections in the Allen V1, however, seem to have a limited effect on size tuning.Future studies might therefore want to explore whether changes to these connections could generate suppression.For temporal frequency tuning of the Allen model, the results are closer to experiments than for spatial tuning.However, the preferred frequencies in the Allen V1 are somewhat higher than what was expected based on previous experiments on mice.There is also overall less variation in tuning properties between model cells in V1 than what has been observed experimentally, both regarding temporal and spatial tuning.A wider range of tuning properties is displayed by the LGN, but the variation seems to average out in V1.In the future, a greater variation in V1 frequency tuning may be obtained by adjusting the cortical and thalamocortical connections according to their preferred frequencies.i I would first like to thank my supervisor, Gaute Einevoll, for introducing me to computational neuroscience, and for your help and motivation during my master studies.Thank you for always being available, and for taking the time to answer my many questions during our meetings.I would like to express my thanks to Torbjørn Vefferstad Ness, Alexander Johannes Stasik and Espen Hagen for your time spent helping with the BMTK and setting up the Allen model.I also want to thank Atle Eskeland Rimehaug, for sharing your knowledge about the model.To Camilla Marie Baastad, for working together in exploring the field of neuroscience, and for our breaks during long days in the maths building and at CINPLA.To Julie Uv, for being an inspiration and making my days brighter.Thank you for taking a genuine interest in my
Author: Lena Myklebust
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W3093710138