Abstract:
In this thesis, a new concept for development and simulation of anatomically and functionally constrained models of signal flow in neural networks is described. This approach consists of the following tools:
1. A standardized anatomical reference frame of the brain region studied and registration methods to integrate anatomical data from different experiments with the highest precision possible.
2. A method for determining morphological neuron types to allow correlation between measurements of the morphology and functional responses of individual neurons.
3. A tool to build an average three-dimensional (3D) statistical model of the neural networks in a brain region based on a representative sparse sample of all neuron types present in the brain region. This model contains 3D morphological models for every neuron in the brain region, as well as the total number and 3D distribution of synaptic contacts between them.
4. A method to activate the network based on measured responses of different neuron types, and to simulate the response of individual neurons representative of different cell types within this network model.
The feasibility and validity of this process is demonstrated on the example of rat vibrissal cortex. The 3D model of this primary sensory area in cortex contains ∼ 530,000 neurons of 16 different types and ∼ 6 × 10^9 thalamocortical and intracortical synapses. Activation of this model with functional responses measured after whisker touch and simulation of
the responses of different neuron types shows that the simulated model responses match experimental measurements. This allowed investigating how robust sensory-evoked responses after different sensory stimuli are formed in different neuron types using computer simulations, and to make predictions to experimentally test these hypotheses.