Registration-based segmentation of murine 4D cardiac micro-CT data using symmetric normalization.

Area of Science:

• Biomedical Imaging
• Cardiovascular Research
• Preclinical Studies

Background:

• 4D cardiac micro-CT offers high resolution for preclinical cardiovascular disease studies.
• Manual segmentation of cardiac chambers in 4D images is time-consuming.
• Efficient quantitative analysis is crucial for throughput in research.

Purpose of the Study:

• To develop and validate a registration-based pipeline for automated segmentation and functional analysis of 4D cardiac micro-CT data in mice.
• To improve the efficiency of quantitative analysis of cardiac function from high-resolution micro-CT imaging.
• To assess the accuracy and speed of the proposed automated segmentation pipeline.

Main Methods:

• A pipeline utilizing deformable registration for label propagation was developed.
• Edge-preserving smoothing using 4D bilateral filtration was applied.
• Manual segmentation of one cardiac phase was used to initiate automated segmentation for all phases.
• Functional metrics (stroke volume, ejection fraction, cardiac output, cardiac index) were derived from segmented chamber volumes.
• Validation involved Dice coefficients and volume accuracy comparisons with manual segmentations.

Main Results:

• The pipeline achieved an 80% increase in speed compared to manual segmentation.
• Dice coefficients >0.90 for the left ventricle and >0.80 for other major chambers (except left atrium) demonstrated good accuracy.
• Functional metrics were underestimated by an average of 6.76% due to errors around atrioventricular valves.
• Segmentation of the left atrium was found to be less reliable.

Conclusions:

• The registration-based pipeline offers a significantly faster approach for analyzing 4D cardiac micro-CT data.
• The method provides accurate segmentation for most cardiac chambers, enabling reliable functional analysis.
• Further refinement is needed to address segmentation inaccuracies around atrioventricular valves for improved functional metric precision.