AI-based reconstruction from low-resolution acquisitions for diffusion MRI: evaluation of FOD similarity

Area of Science:

• Neuroimaging
• Medical Physics
• Biomedical Engineering

Background:

• Magnetic Resonance Imaging (MRI) offers insights into tissue microstructure.
• High Angular Resolution Diffusion Imaging (HARDI) assesses white matter architecture via Fiber Orientation Distribution (FOD).
• HARDI's limitations include low signal-to-noise ratio (SNR) and long scan times, with conventional processing (ZIP) causing artifacts.

Purpose of the Study:

• To evaluate an AI-based reconstruction method, Precise IQ Engine (PIQE), for HARDI data.
• To compare PIQE's performance against conventional zero-fill interpolation processing (ZIP) combined with Advanced Intelligent Clear-IQ Engine (AiCE).
• To assess PIQE's effectiveness in reconstructing low-resolution diffusion data to standard resolution.

Main Methods:

• Ten healthy volunteers underwent 3T HARDI scans.
• Fiber Orientation Distributions (FODs) were estimated using constrained spherical deconvolution.
• Low-resolution data processed with PIQE were quantitatively compared to standard-resolution acquisitions and ZIP+AiCE.

Main Results:

• PIQE demonstrated higher distributional similarity, indicated by lower Jensen-Shannon divergence.
• PIQE showed superior directional agreement, evidenced by a higher angular correlation coefficient compared to ZIP+AiCE.
• Statistically significant improvements in similarity were observed with PIQE.

Conclusions:

• PIQE shows potential as an effective AI-based reconstruction method for HARDI.
• PIQE may enhance the quality and efficiency of advanced diffusion MRI applications.
• The findings suggest PIQE can overcome limitations of conventional processing for detailed white matter microstructure assessment.