An optimum principle predicts the distribution of axon diameters in normal white matter

Many important functional properties affecting nerve conduction are influenced by axon diameter. It is also known that the axon diameter distribution (ADD) in normal nerve fascicles is heterogeneous and skewed. A recent attempt to model and explain the parametric form of these distributions was based on biomechanical principles. Here we explore a neurophysiologically-based hypothesis that the observed ADD can be obtained by optimizing the information flow through a fascicle subject to reasonable anatomical and metabolic constraints. Specifically, we use a variational framework to find an optimal distribution based on the fascicle's channel capacity and informative upper bound (IUB), subject to constraints of fixed available fascicle cross-sectional area and fixed number of axons, to derive two novel probability density functions, which we then compare to other previously used distributions. We show, using experimental histological data, that the distributions based on this optimum principle outperform other distributions. Moreover, the new distribution that optimizes the IUB is extremely robust in fitting ADD data obtained histologically, making it well-suited for use in MRI techniques to measure ADDs in vivo, e.g., AxCaliber MRI.