Automatic cardiac LV segmentation in MRI using modified graph cuts with smoothness and interslice constraints

Area of Science:

• Cardiovascular Imaging
• Medical Image Analysis
• Biomedical Engineering

Background:

• Late-enhanced MRI is standard for cardiac viability assessment.
• Myocardial segmentation is essential for MRI-based cardiac viability analysis.
• Automated segmentation is needed for efficient processing of large MRI datasets.

Purpose of the Study:

• To develop a generic, rule-based framework for automatic segmentation of the left ventricle myocardium.
• To enhance robustness against subject- and study-specific variations in cardiac MRI.
• To support the processing of massive quantities of cardiac MRI data for viability assessment.

Main Methods:

• A rule-based framework utilizing intensity, shape, and interslice smoothness constraints.
• Automatic initialization considering geometrical and appearance properties of the left ventricle.
• A decoupled, modified graph cut approach with control points for segmentation.

Main Results:

• The method achieved a Dice coefficient of 0.81±0.05 on late-enhanced MRI and 0.92±0.04 on cine-MRI.
• Evaluated on 20-patient (late-enhanced MRI) and 15-patient (cine-MRI) databases.
• Demonstrated favorable comparison against a 3D Active Shape Model approach.

Conclusions:

• Experimental validation confirms increased accuracy and versatility of the proposed segmentation method.
• The approach is effective across different magnetic resonance sequences (late-enhanced and cine-MRI).
• The automated segmentation framework shows promise for clinical applications in cardiac viability assessment.