Robust Tensor Splines for Approximation of Diffusion Tensor MRI Data.

Area of Science:

• Medical Imaging
• Computational Geometry
• Differential Geometry

Background:

• Diffusion Tensor (DT) Magnetic Resonance Imaging (MRI) generates symmetric positive definite (SPD) tensor fields.
• Accurate approximation and interpolation of these tensor fields are crucial for medical image analysis.
• Existing methods struggle with noise and outliers in DT-MRI data.

Purpose of the Study:

• To develop a novel and robust spline approximation algorithm for noisy SPD tensor fields.
• To improve the accuracy of tensor field interpolation in medical imaging, specifically DT-MRI.
• To provide a method that effectively handles noise and outliers in tensor data.

Main Methods:

• A statistically robust algorithm using a tensor product of B-splines is developed.
• The algorithm utilizes the Riemannian metric of the manifold of SPD tensors.
• A two-step iterative procedure alternates Riemannian distance-based tensor spline evaluation and data fitting.

Main Results:

• The proposed algorithm demonstrates significantly improved results compared to four existing tensor interpolation methods.
• The method shows superior performance on noisy DT-MRI data from rabbit heart slices, including handling outliers.
• Validation using synthetically generated noisy tensor field data with outliers confirms the algorithm's robustness.

Conclusions:

• The developed spline approximation algorithm offers a robust and accurate solution for interpolating noisy SPD tensor fields.
• This method has significant potential applications in DT-MRI registration and atlas construction.
• The algorithm provides enhanced accuracy and reliability for medical image analysis involving tensor data.