bridging the micro-macro gap with diffusion MRI

Image Processing

The transition from diffusion-weighted MR images to fullscale microstructural modeling is setting a much higher bar on estimating signal parameters in an accurate, precise and robust way.  However parameter estimation is a challenging problem because diffusion MRI suffers from low signal-to-noise ratio and various imaging artifacts, such as eddy current distortions, EPI distortions, and Gibbs ringing.

Although we strive to optimize our image quality during acquisition, rigorous post processing of the images is required to achieve the desired image quality. The MRI Biophysics group has gained expertise in:

  • Denoising: We developed a selective denoising technique by exploiting data redundancy in the PCA domain using universal properties of the eigenspectrum of random (cf. thermal noise) covariance matrices. Identifying the Marchenko-Pastur distribution in the eigenspectrum allows a model-independent extraction of the signal-contributing components.
  • Gibbs ringing suppression:  The Gibbs ringing effect on the diffusion parameters is unexpectedly strong compared to the 9% alterations in the DW images. To cure the artifact without resolution loss, we extrapolate the k-space beyond truncation frequency by adopting physically reasonable representation of the image that is imposed by the second order TGV function.
  • Parameter estimation: Linear, non-linear, weighted, or constrained … each estimator has its own strengths, limitations, and pitfalls in terms of accuracy, precision, and robustness.


Scholarly Lite is a free theme, contributed to the Drupal Community by More than Themes.