Medinria : Dt-mri Processing and Visualization Software

Major advances in medical imaging raised the need for adapted methods and softwares. The recent emergence of diffusion tensor MRI (DT-MRI) was challenging since data produced by this modality are not simple grey-value images but complex diffusion tensor fields. Tensors are symmetric, positive definite matrices and suffer from a lack of adapted theoretical tools to manipulate them. In this paper, we propose two solutions to tensor computing. First, we endow the tensor space with an affine-invariant Riemannian metric and show how some well-known numerical schemes for scalaror vector-valued images can be adapted to tensors. Second, we present the Log-Euclidean (LE) metrics as a more judicious choice as they are less computationally-intensive than the previous one. We show how LE metrics are implemented into a software called MedINRIA, and more especially in a module named DTI Track dedicated to DT-MRI processing. In this tool targeting the clinicians, we offer the full pipeline for DTI analysis, including diffusion tensor estimation, anisotropic tensor smoothing, and fiber tracking. A complementary module called TensorViewer is presented as a nice way to visually inspect the quality of tensor fields produced by DTI Track. This set of softwares, freely available on-line, offers extra features for further DTI analysis, like volumetric image visualization and fiber bundling.