Left atrial appendage segmentation from 3D CCTA images for occluder placement procedure

Area of Science:

• Medical Imaging
• Cardiovascular Imaging
• Image Segmentation

Background:

• Percutaneous left atrial appendage (LAA) closure is a novel stroke prevention strategy for atrial fibrillation patients.
• Accurate LAA measurements from computed tomography (CT) images are crucial for planning LAA closure procedures.

Purpose of the Study:

• To develop and validate a semi-automatic method for LAA segmentation from 3D coronary CT angiography (CCTA) images.
• To enable accurate LAA measurements and propose optimal occluder placement locations for procedural planning.

Main Methods:

• A semi-automatic segmentation approach using 3D CCTA images.
• Requires minimal user input: a threshold value and a single seed point within the LAA.
• Includes an intuitive correction mechanism for the proposed delineation plane and calculates necessary measurements.

Main Results:

• Validated on 17 CCTA images with high agreement to manual segmentations (average Dice coefficient of 92.52%).
• Achieved high accuracy in LAA orifice localization, with an average distance of 2.51 mm from expert-determined locations.
• Demonstrated strong correspondence between automated segmentation and ground truth data.

Conclusions:

• The proposed semi-automatic LAA segmentation method achieves high accuracy and correspondence with expert segmentations.
• The method's ability to accurately localize the LAA orifice and provide measurements shows significant potential for clinical application in LAA closure procedure planning.