3D Segmentation Of Cardiac Ultrasonic Image

Ultrasound (US), including 2D and real-time 3D US, is one of the most important means for diagnosis of the cardiac diseases. Real-time 3D can have the advantage over 2D US of diagnosing space information, but can't be popular for its high price. 3D reconstruction of cardiac US from 2D sectional images can obtain the same effect as the real time 3D US imaging and the potential quantitative analysis can be explored on it. 3D segmentation is the key basis of 3D reconstruction. At present, few 3D US segmentation methods are proposed based on complicated cardiac models, which need abundant artificial inductive work. Some segmentation software used in clinic can do some initial segmentation only and can't be quantitatively analyzed. A simple and effective method for 3D segmentation of cardiac US based on 2D sectional images is proposed in this study.The segmentation algorithm is mainly based on the combination of ellipse region growing and 3D deformable model including 4 steps. Firstly, initial region on specified slice is segmented by the seed ellipse through simple region growing and ellipse fitting algorithm. Secondly, rest slices are segmented by projected growing and ellipse fitting procedure and the initial contour composed by these ellipses is obtained. Thirdly, the ellipses are dispersed and then the triangular model is built. In the end, the triangular model is deformed to attain the final 3D contour. Every vertex's inner and outer energies are needed to be calculated at deforming, and then its moving velocities are calculated and it changes new position. Some local simplified algorithm has been used to reduce the complexity of resample.The 3D volume data in experiments are acquired by rotational transesophageal echocardiography (TEE) and interpolation. The initial 2D contour by ellipse model can coarsely describe most regions of left ventricle, which can basically determine...