Biomedical Research

In this paper, an efficient approach for ultrasonic object segmentation with special application for left ventricle segmentation in echocardiography images is proposed. At first, an efficient hybrid trend for ultrasonic image edge detection is suggested. Then, a modified level set approach is introduced based on the extracted edges and the computed probabilistic map as the stopping criteria for the contour evolution. Both synthetic and clinical images are utilized as validation measures with respect to the prior techniques which indicate outperform results quantitatively and qualitatively. Left ventricle segmentation using proposed method illustrates expert-approved performance, providing a reliable tool for clinical practice, as evidenced by less than 3% ejection fraction error. Making use of probabilistic analysis and elaborated edge detection with level set approach helps to avoid the local minima and excessive contour expansion, while the left ventricle valve is open or there is not clear edge.

Author(s): Nima Sahba, Emad Fatemizadeh, Hamid Behnam
Abstract | Full-Text | PDF

Share this