Automatic model-based segmentation of the heart in CT images

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Efficient analysis of cardiac computed tomography (CT) data requires automated image processing.
• Accurate segmentation of cardiac structures is crucial for quantitative analysis.

Purpose of the Study:

• To develop a fully automatic, model-based method for segmenting the whole heart (chambers, myocardium, great vessels) from 3D CT images.
• To improve convergence and segmentation accuracy through progressive model adaptation.

Main Methods:

• Heart localization using 3D generalized Hough transform.
• Pose correction via global similarity transformation.
• Multicompartment mesh initialization using piecewise affine transformations per region.
• Deformable adaptation for precise boundary matching.

Main Results:

• Achieved a mean surface-to-surface error of 0.82 mm in leave-one-out validation on 28 images.
• Piecewise affine transformation demonstrated superior characterization of interphase and interpatient shape variability compared to principal component analysis.

Conclusions:

• The proposed model-based approach enables accurate and automatic segmentation of the entire heart from 3D CT data.
• The novel initialization and adaptation strategy offers improved shape variability analysis for cardiac imaging research.