A semi-automatic clustering-based level set method for segmentation of endocardium from MSCT images

Area of Science:

• Medical Imaging
• Computational Anatomy
• Cardiovascular Diagnostics

Background:

• Multi-slice Computed Tomography (MSCT) provides crucial 3D cardiac data for disease diagnosis.
• Manual segmentation of cardiac MSCT data is time-consuming and inefficient.
• Accurate endocardium segmentation is vital for quantitative cardiac analysis.

Purpose of the Study:

• To develop a semi-automatic method for robust endocardium surface segmentation from cardiac MSCT images.
• To introduce a novel speed function for level set segmentation of cardiac structures.
• To evaluate the performance of the proposed method in terms of speed and accuracy.

Main Methods:

• A level set approach is employed for flexible and accurate segmentation of complex cardiac anatomy.
• A new speed function is proposed, utilizing image intensity clustering for region and background differentiation.
• The method is implemented and tested in 2D, 3D, and 4D cardiac MSCT datasets.

Main Results:

• The proposed semi-automatic method demonstrates robust segmentation of the endocardium surface.
• The novel speed function effectively handles non-homogeneous blood pool intensities.
• Experimental results show favorable computational speed and segmentation accuracy across different dimensions.

Conclusions:

• The developed semi-automatic level set method offers an efficient and accurate solution for cardiac endocardium segmentation.
• This technique simplifies the analysis of cardiac MSCT data, aiding in the diagnosis of cardiovascular diseases.
• The method's ability to segment non-homogeneous regions enhances its clinical applicability.