Improving left ventricular segmentation in four-dimensional flow MRI using intramodality image registration for...

Area of Science:

• Cardiovascular Imaging
• Medical Image Analysis
• Biomedical Engineering

Background:

• Accurate left-ventricle segmentation is crucial for four-dimensional flow MRI blood flow analysis.
• Low contrast between blood and myocardium in MRI often hinders precise segmentation.
• Existing methods struggle with inter-slice misalignment due to patient motion.

Purpose of the Study:

• To develop and validate an automated framework for improving left-ventricle segmentation in four-dimensional flow MRI.
• To address challenges in aligning cine-MRI segmentations with four-dimensional flow MRI data.
• To ensure reliable blood flow analysis through enhanced segmentation accuracy.

Main Methods:

• Utilized morphological cine-MRI for initial left-ventricle segmentation.
• Developed a robust image registration framework to automatically align segmentations.
• Mitigated inter-slice misalignment errors caused by patient motion and respiratory drift.
• Evaluated the approach on data from 20 healthy volunteers and patients.

Main Results:

• Achieved high spatial correspondence between manual and automatically aligned segmentations.
• Demonstrated significant improvements in alignment compared to uncorrected segmentations (P < 0.01).
• Found no significant differences in blood flow analysis between manual and automatically corrected segmentations (P > 0.05).

Conclusions:

• The proposed automated registration framework effectively improves left-ventricle segmentation in four-dimensional flow MRI.
• This method shows significant potential for enabling reliable and accurate blood flow analysis.
• The findings support the clinical utility of the approach in cardiovascular research and diagnostics.