A multiple active contour model for cardiac boundary detection on echocardiographic sequences

Area of Science:

• Cardiology
• Medical Imaging
• Biomedical Engineering

Background:

• Accurate left-ventricular border tracing is vital for cardiac function quantification.
• Echocardiographic sequences require precise epicardial and endocardial boundary detection.

Purpose of the Study:

• To develop and validate an advanced active contour model for detecting left-ventricular borders in cardiac ultrasound.
• To assess the algorithm's accuracy against expert manual delineations.

Main Methods:

• An extension of active contour models was employed to delineate epicardial and endocardial borders.
• The algorithm was tested on short-axis cardiac sequences across the entire cardiac cycle.
• Computer-generated boundaries were compared to manual outlines from four expert observers on 44 datasets.

Main Results:

• Mean boundary distances were 2.80 mm (epicardium) and 3.61 mm (endocardium), comparable to interobserver variability.
• Correlation coefficients for enclosed areas were 0.95 (epicardium) and 0.91 (endocardium).
• The algorithm demonstrated robustness and insensitivity to initial curve selection.

Conclusions:

• An effective and robust algorithm for extracting left-ventricular boundaries from echocardiographic sequences has been developed.
• The method provides accurate and reliable border detection, aiding in cardiac function assessment.