Motion-guided segmentation for cine DENSE MRI.

Area of Science:

• Cardiovascular Imaging
• Medical Image Analysis
• Biomedical Engineering

Background:

• Defining myocardial contours is a critical yet time-consuming step in dynamic cardiac MRI analysis.
• Displacement Encoding with Stimulated Echoes (DENSE) MRI provides quantitative displacement data for non-invasive myocardial kinematics studies.
• Accurate epicardial and endocardial contour definition is essential for quantifying regional displacement and strain over time.

Purpose of the Study:

• To develop and validate a reliable, semi-automated 2D segmentation technique for cine DENSE cardiac MRI.
• To leverage encoded motion within DENSE images to efficiently propagate contours across cardiac phases.
• To reduce the manual effort required for contour definition in dynamic cardiac MRI analysis.

Main Methods:

• A semi-automated 2D segmentation technique was developed, using manually defined contours in initial frames.
• The technique projects these contours through time by utilizing the motion-encoded phase shifts intrinsic to DENSE MRI.
• The method was validated by comparing motion-guided contours to manually drawn contours on conventional and slice-followed cine DENSE datasets.

Main Results:

• The motion-guided segmentation accuracy was comparable to inter-observer variability based on area measures of segmentation error.
• Average radial epicardial segmentation error was 0.36±0.08 pixels (slice-followed) and 0.40±0.10 pixels (conventional).
• Average radial endocardial segmentation error was 0.46±0.12 pixels (slice-followed) and 0.46±0.16 pixels (conventional).

Conclusions:

• Motion-guided segmentation offers an effective and reliable method for semi-automated contour definition in cine DENSE MRI.
• This technique significantly reduces analysis time while maintaining accuracy comparable to manual segmentation.
• The approach is versatile, applicable to various views and regions of interest, and handles challenging early cardiac phases.