MRI local metrics

Local diffusion MRI metrics were computed for every subject on DWI images using 4 different local reconstruction models:

Diffusion Tensor Imaging (DTI)

DTI metrics computations are implemented directly in TractoFlow using dipy.reconst.dti (tutorial) and are based on the reconstruction of the diffusion signal with a tensor model quality control metrics: (Tournier, 2011). In our case, using a b-value shell 1000 mm2/s the model estimates diffusion anisotropy metrics using a weighted least squares single-tensor fit.

FreeWater

Freewater and freewater corrected DTI metrics are computed in the freewater_flow pipeline and are based on the separation of the diffusion signal contributions from freewater and the rest with a bi-tensor model (Pasternak, 2009). In our case, using b-value shells 300 and 1000 mm2/s the model estimates freewater water fraction using Accelerated Microstructure Imaging via Convex Optimization (AMICO).

Neurite Orientation Distribution Density Imaging (NODDI)

NODDI metrics are computed with the noddi_flow pipeline and are based on the reconstruction of the WM microstructure with a 3 compartment model: intra-cellular, extra-cellular, and CSF compartments (Zhang, 2012). In our case, using b-value shells 300, 1000 and 2000 mm2/s the model estimates an orientation dispersion index also using Accelerated Microstructure Imaging via Convex Optimization (AMICO).

Fiber Orientation Distribution Function (fODF)

fODF metrics computations are implemented directly in TractoFlow using dipy.reconst.csdeconv (tutorial) and are based on the reconstruction of the fODF with constrained spherical deconvolution (Raffelt, 2012). In our case, using b-value shells 1000 and 2000 mm2/s the model estimates the fODFs by applying a constrained spherical deconvolution on ODFs using a single by default (manually entered) fiber response function.