A Method for Automatic Delineation of the Left Ventricle Borders in Echographic Images with Use Active Contours and Speckle Tracking Techniques

Abstract

At present, an ultrasound imaging of the heart is widely used method for heart disease diagnostics. The main object of ultrasound cardiac examination is the left ventricle (LV) - the most powerful chamber of the heart. In order to evaluate the necessary functional parameters of the LV, we need to delineate borders of the LV in ultrasound image. Manual borders tracing is very time-consuming procedure and it can produce subjective errors. Therefore, it is a vital problem to develop automatic (or semi-automatic) method for LV border delineation in echographic images. For solving this problem, we used as a base the active contours model (snake) - one of the most frequently used method for objects delineation in images. As the mitral valve should not be included in the analysis, in the first frame of the image sequence the user should annotate two points (so called anchor points) - one on each side of the mitral valve, and the third anchor point on the apex of the LV. Then the modified method of speckle tracking with prediction is used for automatically detection these points in the rest frames. Using these three anchor points, an initialization of inner contour of LV border as ellipse is performed. To reduce the time-consuming and increase method’s convergence, we applied multi-scaling technique. As the ultrasound images often suffer from the effects of noise, the preliminary image filtering is carried out. We tested proposed technique with images obtained during clinical cardiac ultrasound examination. The method showed good results of LV border delineation in images with high and middle quality.