COMPLEXITY MEASURES AND NOISE EFFECTS ON DIFFUSION MAGNETIC RESONANCE IMAGING OF THE NEURON AXONS NETWORK IN THE HUMAN BRAIN

摘要

Diffusion Magnetic Resonance Imaging (dMRI) is a novel technique that mirrors the complex architecture of the neuron axons fiber networks in the human brain. Based on the dMRI scans, fractal dimensions (Box Counting and Multifractal Dimensions) are used to quantify the complexity of the neuron axons networks. The values of the fractal dimensions are calculated as a function of the intensity threshold τ in an effort to remove the effects of stochastic noise, always present in the molecular diffusion of water in the brain. It is shown that intermediate values of the noise threshold τ are better for estimating the complexity of the neuron network architecture, because for small τ the presence of stochastic noise often masks the underlying structure, while for τ > 0.6 important parts of the axon network structure are ignored. Calculations of the multifractal dimensions in healthy brains as a function of τ, give consistent scaling results in the medium intensity thresholds, where the neuron axons activity is better discerned. In these intermediate τ scales, deviations are recorded in the multifractal spectra of pathological brains, which indicate damaged network architectures.