4-dimensional local radial basis function interpolation of large, uniformly spaced datasets

Area of Science:

• Medical Imaging
• Computational Fluid Dynamics
• Biomedical Engineering

Background:

• High-resolution medical image velocimetry generates large, complex, time-varying datasets.
• Accurate reconstruction of 4-dimensional (4D) physiological flow phenomena requires advanced interpolation techniques.
• Existing methods struggle with the scale and complexity of transient flow data.

Purpose of the Study:

• To evaluate a 4D local radial basis function (RBF) algorithm for interpolating complex, time-varying medical image velocimetry data.
• To address the need for effective reconstruction of 4D functional relationships from physiological flow datasets.
• To assess the algorithm's suitability for laminar flow analysis.

Main Methods:

• Proposed a 4D interpolation algorithm using Local Hermitian Interpolation (LHI) with multi-quadric RBF and a partition of unity scheme.
• Decomposed the domain into local sub-systems of neighboring points for computational efficiency.
• Validated the algorithm on analytical datasets, CFD-based phantoms, and magnetic resonance imaging (MRI) measurements of cerebrospinal fluid (CSF) velocities.

Main Results:

• The 4D local RBF algorithm demonstrated superior accuracy compared to quad-linear interpolation.
• The technique proved robust, computationally efficient, and capable of refined spatial and temporal interpolation.
• Application to MRI velocimetry data yielded promising results for 4D reconstruction of transient flow fields and boundary movements.

Conclusions:

• The study confirms the feasibility of an accurate, stable, and efficient 4D local RBF interpolation method for large, transient laminar flow datasets.
• Domain decomposition into local stencils prevents ill-conditioning and high computational costs.
• This method offers a potential tool for improved medical diagnoses and drug delivery via better understanding of physiological flow fields like CSF.