Pierre Fillard1, Cyril Poupon1, Jean-François Mangin1
1LNAO, CEA/DSV/I2BM/Neurospin, Gif/Yvette, France
We propose a novel tractography algorithm able to handle various fiber configurations like crossings. Fibers are modeled as chains of unit elements called spins. Spins are allowed to move, merge with other spins, and duplicate to form longer pieces of fibers. The entire white matter is tracked at once, which allows pieces of fibers coming from the entire volume to help find the optimal neuronal pathway in crossing regions. Experiments on two realistic MR phantoms show that such approach is successful in reconstructing fiber crossings and kissings, even using a simple diffusion model like the tensor.