A 3-D Active Contour Method for Automated Segmentation of the Left Ventricle From Magnetic Resonance Images

Area of Science:

• Medical Imaging
• Computational Anatomy
• Cardiovascular Research

Background:

• Cardiac magnetic resonance imaging (MRI) is crucial for diagnosing heart conditions.
• Accurate 3-D segmentation of cardiac structures is essential for quantitative analysis.
• Existing automated methods often require prior statistical knowledge or user interaction.

Purpose of the Study:

• To develop and validate a fast, automated 3-D segmentation technique for cardiac MRI.
• To reconstruct cardiac MRI DICOM data into a 3-D model directly (volumetric segmentation).
• To eliminate the need for training datasets or user-driven segmentation.

Main Methods:

• A novel 3-D active contour method was developed, incorporating unique internal and external energy terms.
• Histogram matching was utilized to enhance segmentation performance.
• A convex hull interpolation was employed to accurately include papillary muscles within the left ventricle segmentation.
• The method was applied to 33 subjects with heterogeneous heart diseases.

Main Results:

• The developed method demonstrated significant improvements in segmentation performance.
• Comparison with manual segmentation and other automated methods showed superior results.
• The technique successfully segmented the left ventricular cavity and myocardium.

Conclusions:

• A true 3-D reconstruction technique for cardiac MRI has been successfully developed.
• The method is automated, requiring no training data or user input.
• This approach offers enhanced segmentation accuracy and efficiency for clinical applications.