Development of Simplified Auto-Segmentable Functional Cardiac Atlas.

Area of Science:

• Radiotherapy and Medical Imaging
• Cardiovascular Anatomy and Physiology
• Computational Anatomy

Background:

• Radiation doses to cardiac substructures are linked to adverse cardiac events.
• Manual delineation of cardiac substructures is labor-intensive.
• Auto-segmentation of cardiac substructures, particularly the conduction system, remains challenging.

Purpose of the Study:

• To propose and evaluate a simplified, auto-segmentable functional cardiac atlas.
• To assess the accuracy of auto-segmentation for cardiac substructures, including the conduction system.
• To provide a tool for improving radiation therapy planning and cardiac event assessment.

Main Methods:

• A cardiac substructure atlas was created from 20 CT scans, including 4 chambers, a high-risk zone (LAD surrogate), and conduction nodes.
• Atlas-based auto-segmentation was evaluated on 20 additional CT scans.
• Dice similarity coefficients and dosimetric comparisons assessed contour accuracy and dose distribution.

Main Results:

• Average Dice similarity coefficients were 0.78 (chambers), 0.65 (high-risk zone), 0.56 (sinoatrial node), and 0.15 (atrioventricular node).
• Auto-segmented substructures were slightly smaller than manual contours.
• Dosimetric parameters showed similarity between manual and auto-segmented contours.

Conclusions:

• A simplified functional cardiac atlas, including the cardiac conduction system, was successfully developed.
• The atlas demonstrated acceptable auto-segmentation properties for most cardiac substructures.
• This atlas holds potential for use in epidemiologic studies and clinical practice for radiation therapy.