Motion-aware stroke volume quantification in 4D PC-MRI data of the human aorta

Area of Science:

• Medical Imaging
• Cardiovascular Physiology
• Computational Fluid Dynamics

Background:

• Four-dimensional phase-contrast magnetic resonance imaging (4D PC-MRI) provides noninvasive, time-resolved, 3D blood flow data crucial for assessing cardiac function and cardiovascular diseases.
• Stroke volume calculation, essential for hemodynamic assessment, relies on accurate vessel wall position and orientation, which dynamically change throughout the cardiac cycle.
• Current 4D PC-MRI analysis often uses static vessel approximations, neglecting the dynamic nature of the cardiovascular system.

Purpose of the Study:

• To quantify the differences in stroke volume calculations in the aorta with and without considering vessel dynamics during the cardiac cycle.
• To develop and evaluate a novel method for motion-aware 4D PC-MRI data analysis.

Main Methods:

• A novel method employing 3D segmentation to initialize temporal segmentation algorithms, enabling 4D segmentation using graph cuts.
• Extraction of vessel surfaces and centerlines for each temporal frame to derive motion information.
• Comparison of stroke volume quantification using static vessel approximations, fixed planes in dynamic vessels (2D PC-MRI equivalent), and motion-aware moving planes.

Main Results:

• Motion-aware quantification demonstrated a 1.57% improvement over calculations without motion consideration when compared to manual estimations by radiologists.
• A mean difference of 7.82% was observed between stroke volumes calculated with and without accounting for vessel motion.
• Automatically generated 4D segmentations showed an 85.75% overlap with manual segmentations.

Conclusions:

• Incorporating vessel motion into stroke volume quantification offers slight improvements, though not statistically significant in this study.
• The presented motion-aware method is feasible for clinical routine due to low computation times and automated processing.
• The generated 4D segmentations have broader applicability for other algorithms, and simultaneous visualization/quantification can enhance understanding of cardiac blood flow.