An Auto Segmentation Method of Gagged Left Ventricle Magnetic Resonance Images Based on Narrow Band Level Set

Qian Chen
2005-01-01
Abstract:An auto segmentation method is presented for tagged left ventricle MR images. It mainly consists of two parts: First, this method localizes the left ventricl based on SVM and gives the initial contour. Second, it evolves the initial curve based on the improved narrow band level set, and achieves the final segmentation. This method improves the generating method of narrow band and decreases the generating time. Focusing on the imaging characters of tagged left ventricle MR images, we introduce the block--pixel variation (BPV) and the intensity comparability to improve the speed term of level set. The results demonstrate that the method advanced in this paper can automatically, quickly and accurately achieve the segmentation of the left ventricle.
What problem does this paper attempt to address?