MR imaging and computer vision

To parallel the rapid progress in MR data acquisition, the authors have developed advanced computer vision methods specifically adapted to the multidimensional and multispectral (T1- and T2-weighted) nature of MR data to extract, analyze, and visualize the morphologic properties of biologic tissues. A multistage image processing scheme is proposed, which performs the three-dimensional (3D) segmentation of the brain (gray and white matter) and ventricular system from two-echo MR volume data with only minimal user interaction. The quality of the segmentation demonstrates the high potential of MR acquisition along with 3D segmentation and 3D visualization for diagnosis, preoperative planning, and research. With segmentation, a fully quantitative 3D exploration is accessible