Registration of coronary arteries in computed tomography angiography images using Hidden Markov Model

Area of Science:

• Cardiovascular Imaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Computed tomography angiography (CTA) is crucial for diagnosing coronary artery disease (CAD) and monitoring vessel wall remodeling.
• Accurate alignment of coronary arteries across longitudinal CTA scans (3-7 year intervals) is essential for quantitative analysis of geometric and structural changes.
• Existing image registration methods are suboptimal for the complex geometry of coronary arteries.

Purpose of the Study:

• To develop a novel method for directly aligning straightened coronary arteries in CTA images.
• To enable quantitative analysis of CAD progression by visualizing changes in vessel geometry and structure over time.

Main Methods:

• A new image registration method was developed to align straightened coronary arteries within a cylindrical coordinate system, guided by extracted centerlines.
• A Hidden Markov Model (HMM) was employed to integrate CTA image intensity data with coronary artery geometric information for alignment.
• Synchronized visualization of corresponding cross-sectional slices and rotation angles allows for side-by-side comparison of pathological features.

Main Results:

• The novel registration method successfully aligned coronary arteries from longitudinal CTA scans.
• Quantitative evaluation using manually labeled landmarks demonstrated an average distance error of 1.6 mm.
• The method allows for direct side-by-side visualization of pathological changes in coronary arteries.

Conclusions:

• The proposed HMM-guided registration method effectively aligns coronary arteries in CTA, facilitating quantitative analysis of CAD progression.
• This technique offers improved visualization of vessel remodeling and pathological changes over time.
• The method achieves high accuracy, as evidenced by the low average distance error in landmark evaluation.