Bilateral filtering of diffusion tensor magnetic resonance images

Area of Science:

• Medical Imaging
• Image Processing
• Diffusion Tensor Imaging

Background:

• Scalar bilateral filtering is a common technique for edge-preserving smoothing.
• Diffusion Tensor Magnetic Resonance Imaging (DTMRI) generates complex tensor field data.
• Smoothing DTMRI data is crucial for accurate analysis and segmentation.

Purpose of the Study:

• To extend scalar bilateral filtering to DTMRI data.
• To develop an edge-preserving smoothing method for diffusion tensor fields.
• To evaluate the effectiveness of the proposed method on synthetic and real DTMRI data.

Main Methods:

• Extended scalar bilateral filtering to the Log-Euclidean framework for DTMRI.
• Employed weighted averaging of neighboring tensors based on spatial and tensor dissimilarity.
• Derived a noniterative closed-form smoothing equation.
• Evaluated various tensor dissimilarity metrics.

Main Results:

• Demonstrated effective edge-preserving smoothing of DTMRI data.
• Showcased interpolation of DT data as a special case of bilateral filtering.
• Presented qualitative and quantitative results on synthetic and real cardiac/brain DTMRI data.
• Illustrated improved segmentation accuracy after smoothing.

Conclusions:

• The proposed bilateral DT filtering effectively smooths DTMRI data while preserving edges.
• The Log-Euclidean framework ensures valid tensor output.
• The method shows promise for improving DTMRI analysis and segmentation accuracy.