I have added the abstracts from the SPARC 2014 diffusion MRI challenge, hosted as a part of the MICCAI workshop on Computational Diffusion MRI (CDMRI 2014). We presented three abstracts, combining denoising with spherical deconvolution techniques, using the SHORE basis to reconstruct the signal and find peaks and also the Sparse Spherical Finite Rate of Innovation (SFRI), which is made to directly find peaks on the sphere. You can grab the abstracts here.
The aim was to both recover original crossings in the phantom as well as predict the signal on missing q-space points. While no best method was declared as the winner, a clear message always come out : There is no one size fits all method, and one must carefully choose what to use depending on the application at hand.
Depending if one wants to recover crossings, fit the signal or aim for imaging microstructure (according to the amount of data available, number of directions and multiple b-values or not), many choices are available. Each user should therefore ask himself "What do I want to do with my data?" and use an appropriate model for the task at hand. As this is a half-website, half-blog, maybe another post regarding diffusion MRI acquisition and usage of various models will come by later, who knows.