Aleatoric-Uncertainty-Aware Maximum Intensity Projection-Based GAN for 7T-Like Generation From 3T TOF-MRA.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Time-of-flight magnetic resonance angiography (TOF-MRA) is crucial for diagnosing cerebrovascular diseases.
• 7T TOF-MRA offers superior signal-to-noise ratio and vessel contrast over 3T, but its clinical use is limited.
• Generating high-quality 7T-like TOF-MRA from 3T data is needed.

Purpose of the Study:

• To develop a method for generating high-quality 7T-like TOF-MRA from routine 3T TOF-MRA.
• To address challenges like spatial sparsity, visibility differences in small vessels, and image misalignment.

Main Methods:

• Proposed a novel aleatoric-uncertainty-aware maximum intensity projection-based generative adversarial network (AU-MIPGAN).
• Utilized a knowledge distillation (KD) framework to integrate 3D vascular morphology from multi-directional MIPs.
• Employed AU modeling to handle spatial misalignment between 3T and 7T paired data.

Main Results:

• The AU-MIPGAN successfully generated 7T-like TOF-MRA from 3T data.
• Qualitative and quantitative evaluations confirmed the method's promising performance.
• The approach effectively enhanced vessel details and compensated for inherent data discrepancies.

Conclusions:

• The AU-MIPGAN provides a viable solution for generating high-fidelity 7T-like TOF-MRA from 3T scans.
• This technique can potentially expand the clinical applicability of advanced MRA.
• The method offers improved visualization of cerebrovascular structures.