Uncertainty Quantification in Hemodynamic Metrics from 4D Flow MRI with Super-resolution in a Carotid Bifurcation...

Area of Science:

• Medical Imaging
• Fluid Dynamics
• Biomedical Engineering

Background:

• Super-resolution (SR) techniques enhance 4D flow magnetic resonance imaging (4DMR) data post-acquisition.
• The impact of SR on carotid artery hemodynamic measurements requires further investigation.

Purpose of the Study:

• To quantify uncertainties in SR-processed 4DMR carotid artery data.
• To compare SR-enhanced 4DMR with native 4DMR and computational fluid dynamics (CFD) for hemodynamic accuracy.

Main Methods:

• A phantom carotid artery model was scanned using 3 Tesla 4DMR.
• SR processing was applied using a pretrained 4DFlowNet.
• CFD simulations served as a reference for comparison.
• Geometrical uncertainties were assessed via surface and Hausdorff distances.
• Hemodynamic metrics included velocity, vorticity, flow rate, and wall shear stress (WSS).

Main Results:

• SR processing improved intraluminal hemodynamic statistics (p=0.033) and stabilized flow rate estimates (SD reduced from 14.7% to 2.8%).
• Mean flow rate values showed larger deviations post-SR.
• Qualitative improvements in WSS were observed, but quantitative differences were not significant (p=0.20).
• Segmentation accuracy yielded a mean surface distance of 0.31 ± 0.06 mm.

Conclusions:

• SR processing offers benefits for carotid hemodynamic analysis by improving data quality and stabilizing measurements.
• Caution is advised when applying SR to low-quality 4DMR data due to potential error amplification.
• Further research is needed to enhance the generalizability of SR techniques in cardiovascular imaging.