Rodrigo de Luis Garcia1,2, Carlos Alberola Lopez2, Gordon Kindlmann1, Carl-Fredrik Westin1
1Laboratory of Mathematics in Imaging, Harvard Medical School, Boston, MA, USA; 2Laboratorio de Procesado de Imagen, Universidad de Valladolid, Valladolid, Spain
This abstract presents a fully automatic DTI segmentation method for anatomical structures in the white matter. Our approach is based on: (a) the use of tensor invariants and the orientation information of the tensor as features, (b) a statistical modeling of the data with a level set implementation, and (c) an automatic initialization with a DTI white matter atlas. This formulation allows to control the relative importance of the different properties of the diffusion tensor, which overcomes limitations of previous approaches in the literature. The method has been validated on two DTI volumes, showing accurate and robust results.