Diffusion- and kurtosis tensor estimators generally assume independent and a Gaussian distributed noise. While with adequate signal-to-noise ratios (SNR) this assumption can be made without significant bias in the estimated models, this is not true when utilizing high diffusion weighting thus low SNR. In such case, the Rician noise should be considered with maximum likelihood (ML) estimators for example. Moreover, the mathematical properties of ML estimators could unveil novel details of the underlying diffusion process within the brain. In this work, we present super-fast, GPU accelerated ML estimator with Matlab interface to provide a practical tool for such improved estimation.
This abstract and the presentation materials are available to members only; a login is required.