Reconstruction of scattered data in fetal diffusion MRI

Area of Science:

• Medical Imaging
• Neuroscience
• Biophysics

Background:

• Diffusion-weighted MRI (DW-MRI) data reconstruction from scattered points is crucial for accurate analysis.
• Existing methods often rely on the diffusion tensor model, which is inadequate for intravoxel orientational heterogeneity (IVOH).
• Fetal brain imaging presents unique challenges due to scattered data from registration algorithms.

Purpose of the Study:

• To develop a model-agnostic method for reconstructing DW-MRI data on regular grids from scattered measurements.
• To address limitations of the tensor model in scenarios with IVOH.
• To improve the quality of DW-MRI data, particularly for in utero fetal brain studies.

Main Methods:

• A groupwise registration method was employed for data preprocessing.
• Dual spatio-angular interpolation using radial basis functions (RBF) was utilized for data reconstruction.
• Leave-one-out cross-validation experiments were conducted on adult and fetal datasets.

Main Results:

• The proposed method demonstrated high accuracy in reconstructing adult DW-MRI data.
• Application to fetal data resulted in improved sequence quality, as evidenced by fractional anisotropy (FA) maps.
• Tractography results showed discernible differences, indicating enhanced data representation.

Conclusions:

• The developed reconstruction method offers a robust, model-agnostic approach for DW-MRI data.
• This technique significantly improves data quality for challenging applications like fetal neuroimaging.
• The findings suggest potential for more accurate in vivo brain studies using advanced MRI reconstruction.