Segmentation and optical flow estimation in cardiac CT sequences based on a spatiotemporal PDM with a correction...

Area of Science:

• Cardiovascular imaging and analysis
• Medical image processing
• Computational anatomy

Background:

• The left ventricle and myocardium are critical for cardiac function assessment.
• Accurate isolation and functional characterization of heart structures are essential for clinical diagnosis.
• Existing methods may have limitations in segmentation accuracy and motion analysis.

Purpose of the Study:

• To propose a novel framework for isolating and displaying functional characteristics of the heart using cardiac CT sequences.
• To combine shape extraction with a new segmentation correction scheme and motion estimation.
• To enhance the assessment of heart mechanical properties.

Main Methods:

• Development of a Spatiotemporal Point Distribution Model (STPDM) for heart structure segmentation, guided by intensity and gradient information.
• Introduction of a novel deformable correction scheme using local histograms, gradients, and binary patterns to improve segmentation accuracy.
• Application of a bio-inspired image representation model based on the Hermite transform for motion estimation.

Main Results:

• The framework was evaluated using multiple cardiac CT sequences, focusing on the left ventricle.
• Quantitative metrics validated the proposed framework's performance.
• The method demonstrated superior efficiency compared to existing techniques.

Conclusions:

• The developed tool assists physicians in identifying cardiac mechanical problems more effectively.
• The new segmentation correction scheme significantly enhances segmentation performance.
• This approach represents a promising technique for assessing heart mechanics.