A 2D Cortical Flat Map Space for Computationally Efficient Mammalian Brain Simulation

Alexander Woodward, Rui Gong, Ken Nakae, Patrice Delmas

Abstract: We present a method for computationally efficient cortical brain simulation by constructing a 2D cortical flat map space on a regular grid. Neuroscience data can be mapped into this space to provide experimental information and constraints for the simulation. Neuron locations can be determined probabilistically by treating neuron densities as empirical probability distributions that can be sampled from. Therefore, this approach can be used for specifying parameters for small-scale to large-scale brain simulations (that could simulate the true number of neurons in the brain). The spatial warping of the cortical surface, when going between the flattened 2D space back into 3D, is accounted for by an estimated scale factor. This can be used to scale properties such as diffusion rates of neural activity across the flat map. We demonstrate the approach using neuroimaging data of the common marmoset, a New World primate.