A Novel Richardson-Lucy Model with Dictionary Basis and Spatial Regularization for Isolating Isotropic Signals

Area of Science:

• Neuroscience
• Medical Imaging
• Biophysics

Background:

• Diffusion-weighted magnetic resonance imaging (dMRI) is crucial for neuroscience but affected by partial volume effects (PVEs).
• PVEs lead to inaccurate fiber orientation estimation in dMRI.
• Accurate white and gray matter quantification is essential for understanding brain structure.

Purpose of the Study:

• To develop a novel method to overcome PVEs in dMRI.
• To improve the accuracy of fiber orientation distribution function (fODF) reconstruction.
• To introduce a new index for quantifying white and gray matter.

Main Methods:

• A novel response function (RF) and fiber orientation distribution function (fODF) were used to construct signal models.
• The fODF was represented using a dictionary basis function.
• An innovative iterative Richardson-Lucy (iRL) model integrating spatial regularization was proposed to address spherical deconvolution challenges.
• A new index, Piso, was developed to quantify white and gray matter based on fODF.

Main Results:

• The proposed iRL method robustly reconstructs more accurate fODFs compared to classic methods.
• The novel index Piso demonstrates superior performance in quantifying white and gray matter compared to fractional anisotropy (FA) and general fractional anisotropy (GFA).
• Experimental results on simulated and real data validate the effectiveness of the iRL model and Piso index.

Conclusions:

• The iRL method effectively mitigates PVEs in dMRI, leading to improved fiber orientation estimation.
• The Piso index offers a more accurate and reliable measure for white and gray matter quantification.
• This work advances dMRI analysis for more precise neuroscience imaging and brain structure characterization.