Correlation-based discrimination between cardiac tissue and blood for segmentation of the left ventricle in 3-D...

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Cardiovascular Ultrasound

Background:

• Automated segmentation of 3-D echocardiographic images is crucial for cardiac analysis.
• Incorporating temporal information can enhance segmentation accuracy.
• Current methods may benefit from improved contrast between blood and cardiac tissue.

Purpose of the Study:

• To determine optimal settings for calculating temporal cross-correlations (maximum cross-correlation - MCC) for 3-D echocardiography.
• To enhance contrast between blood and cardiac tissue for improved segmentation.
• To integrate MCC values into a deformable model for automated left ventricular segmentation.

Main Methods:

• Optimized calculation of temporal cross-correlations between cardiac image frames.
• Assessed contrast and boundary gradient quality measures to optimize MCC values.
• Evaluated signal choice (radiofrequency vs. envelope data) and axial window size.
• Incorporated optimal MCC values into a deformable model for segmentation.
• Tested MCC values against filtered, demodulated radiofrequency data.

Main Results:

• Optimal MCC values were determined using envelope data with a small axial window (0.7-1.25 mm).
• This approach yielded the best contrast and boundary gradient between blood and cardiac tissue throughout the cardiac cycle.
• Preliminary results show that MCC values enhance automated left ventricle segmentation.

Conclusions:

• Envelope data combined with small axial windows optimizes contrast for cardiac image analysis.
• Maximum cross-correlation (MCC) values significantly improve automated segmentation of the left ventricle.
• Temporal information, via MCC, offers added value for 3-D echocardiographic segmentation.