There are two overarching challenges of clinical diffusion-weighted magnetic resonance imaging: the estimation of long-range structural connectivity, and second, the quantification of tissue properties at the a cellular scale of micrometers. Here we propose a global optimization framework that tries to resolve both challenges for human white matter fibers, by unifying global fiber tractography and intra-voxel mesoscopic modeling. The algorithm is based on a Simulated Annealing approach which simultaneously optimizes diffusion parameters and fiber locations. Additionally, fibers can carry their individual set of diffusion microparameters which allows to link them by their structural relationships.
Target audience:
People who are interested in Diffusion Weighted MRI in the context of tractography and microstructure modelling.
Outcome/Objectives:
Learning about the principles of global tractography [1,2] and ways to join tractography and microstructure imaging [3,4,5] in a holistic manner.
Purpose:
Diffusion MRI analysis may devided into two basic branches, on the one hand tractography and structural connectivity, on the other microstructure imaging. Both fields rely on each other and try to solve highly ill-posed problems. The approach to join both fields may help for a better understanding of brain diffusion in white and gray matter.
Methods:
The analysis of dMRI is stated as one global optmization problem and is solved via a Gibbs sampler. Similar to simulated annealing, a polymerization and freezing process is simulated which concurrently optimizes fiber geometry and microstructure parameters [3].
Results:
Examples show that the algorithm is able to solve the formulated problem in a reasonable time. Results are shown for Human Connectome data and some simpler, more clinical, protocols.
Discussion
Estimated parameters and fiber geometries are reasonable. The fiber geometry actually helps to regularize the microstructure estimatation. However, problems remain:the absence of global geometric constraints on fiber geometry let fiber terminate within white matter.Or, what is the choice of the right microstructure model?
Conclusion:
The identification of quantitative and non-invasive biomarkers is essential to characterize the reversibility of a neurological disorder and decide for an effective treatment. This work is one attempt to use diffusion MRI to provide such novel biomarkers.
References:
[1] Global fiber reconstruction becomes practical
Reisert, Marco and Mader, Irina and Anastasopoulos, Constantin and Weigel, Matthias and Schnell, Susanne and Kiselev, Valerij
Neuroimage 54, 955--962,2011
[2] Toward global tractography,
Mangin, J-F and Fillard, Pierre and Cointepas, Yann and Le Bihan, Denis and Frouin, Vincent and Poupon, Cyril,
NeuroImage 80, 290--296, 2013
[3] MesoFT: unifying diffusion modelling and fiber tracking,
Reisert, Marco and Kiselev, Valerij G and Dihtal, Bibek and Kellner, Elias and Novikov, DS,
International Conference on Medical Image Computing and Computer-Assisted Intervention, 201-208, 2014
[4] Global tractography of multi-shell diffusion-weighted imaging data using a multi-tissue model
Christiaens, Daan and Reisert, Marco and Dhollander, Thijs and Sunaert, Stefan and Suetens, Paul and Maes, Frederik,
NeuroImage 123, 89-101, 2015