Spiking neural network models of the visual cortex Dynamics, plasticity and feature processing

A central element in the architecture of the visual cortex is the column-like circuitry, known as cortical column. Considered a building block of neural function, it acts as a mini-processor and plays a crucial role in interpreting visual information. This thesis presents models of the cortical column in the mouse primary visual cortex, focusing on modulation by feedforward (FF) and feedback (FB) inputs, oscillatory activity, and orientation selectivity.
In Chapter 2, a biologically detailed V1 cortical column model was developed, incorporating multiple interneuron types (PV, SST, VIP), excitatory pyramidal cells, and AMPA, GABAA, and NMDA receptors. We examined how spontaneous states, FF, and FB inputs affect neuronal populations per layer. FF inputs to layer 4 generally excited the column, whereas FB had an inhibitory effect across layers, particularly in 2/3 and 6, due to recruitment of inhibitory neurons.
Chapter 3 introduced synaptic plasticity to investigate the genesis and maintenance of neural oscillations. We analyzed the role of FF and FB inputs and identified PV cells as key modulators of oscillation frequencies. We also determined the necessary conditions—layer-specific plasticity and connectivity patterns—required to generate oscillations.
Chapter 4 explored orientation selectivity across cortical layers. Connecting orientation-tuned LGN neurons to layer 4 excitatory cells in a fixed-weight model produced unrealistic selectivity indices and firing rates. Implementing a Hebbian plasticity rule reinforcing active like-to-like connections yielded realistic orientation selectivity indices and firing rates, consistent with experimental data.
Overall, this work offers mechanistic insights into how inputs, plasticity, and connectivity shape computations in the cortical column.