A method to detect a boundary between a tumor region and surrounding normal tissue using tissue motion vector mapping is described. The boundary between the tumor and the surrounding normal tissue can not be detected in the conventional B-mode images because the difference between their echogenicity is too small in several type of tumor. The extent and the aspect of adhesion at the boundary depend on the tumor type. Some possibilities are that metastatic tumors tend to have a loose boundary and move along it relatively smoothly by respiratory motion. Such a boundary may not be detected in a single B-mode frame, but can be recognized in a B-mode movie. Defining of the boundary is essential for non- and minimally invasive therapies, and the estimation of tumor volume based on it is also essential for optimization of medication. We proposed a new imaging method based on the detection of spatial discontinuity of the tissue motion vector, estimated by the image correlation between sequential frames. A VX2 tumor implanted to the liver of a rabbit was chosen to test the proposed method. The boundary between the tumor and the surrounding normal tissue was not clearly detected in B-mode images, but it was clearly shown by the proposed method.