We present a method for segmenting white matter as well as the gray matter structures from diffusion tensor magnetic resonance images (DT-MRI). The segmentation is done evolving a set of coupled level set functions. The zero level set of each level set function forms a surface in 3D that is driven by the region-based force including all tensors belonging to a certain region. The region-based force is defined by using a very sensitive similarity measure between DT. We apply our method for segmenting the thalamus and its nuclei.
This technical paper proposes several new strategies for level set methods to segment efficiently complex objects as present in DT-MRI. First of all, we present a very sensitive similarity measure that distinguishes very subtle differences between regions within, for example, the thalamus. Secondly, we present a new way of selecting the most representative tensor for group of tensors for these kinds of applications. We argue for the importance to use the tensor minimizing the variation within the group of tensors instead of the mean tensor as suggested in other papers on tensor segmentation. The third important point is the necessity of using several coupled level sets to define the background. Methods differentiating only between foreground and background will fail when applied to complex structures such as the brain. It is crucial for a region-based approach to consider all the surrounding structures for a correct definition of the forces driving the segmentation.