Continuous, Tract-Constrained Directional Vector Fields from rDCM Effective Connectivity Using Mixture Density Networks

Effective connectivity estimates from dynamic causal modeling (rDCM) are typ-
ically represented as dense directed matrices. However, these matrices do not
directly specify a continuous, geometric description of how directed influence
is routed through 3D white matter anatomy. Because these estimates are ex-
pressed as abstract parcel to parcel weights, they may require fusion with addi-
tional data, such as diffusion MRI tractography in order to define anatomically
plausible pathways. I present a solution for fusing whole-cortex estimates from
regression dynamic causal modeling (Schaefer-400) with a dense tractography
atlas (HCP-1065)