Segmentation and Volumetric Analysis of Heart from Cardiac CT Images

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Cardiovascular Imaging

Background:

• Cardiac CT is crucial for cardiovascular disease evaluation.
• Accurate heart segmentation is vital for diagnosis and management but remains challenging.
• Manual segmentation is time-consuming, labor-intensive, and prone to errors.

Purpose of the Study:

• To develop a fast, fully automatic segmentation method for the heart from CT images.
• To overcome the limitations of manual segmentation in cardiac imaging.
• To enhance the efficiency and accuracy of cardiovascular image analysis.

Main Methods:

• An image-driven processing pipeline was developed.
• Techniques include adaptive median filter, multi-level thresholding, active contours, and mathematical morphology.
• Incorporation of human anatomy knowledge for region delineation.

Main Results:

• The automated method achieved an average Dice score of 0.866 and Jaccard score of 0.776.
• Average Hausdorff distance was 33.29 mm for segmenting heart chambers, aorta, and blood vessels.
• Results were validated against ground truths using Bland-Altman plots.

Conclusions:

• The proposed method successfully segmented the heart from CT images automatically.
• This technique can be a foundation for developing AI-based segmentation solutions.
• Automated segmentation offers improved efficiency and accuracy in cardiac image analysis.