Two different reconstruction methods of bidimensional distributions from nuclear magnetic resonance (NMR) were developed; one is a standard in the industry and the other is a new method developed in our institute. The method allows the production of T1-T2, T2-T2 and D-T2 bidimensional distributions sampled in logarithmic scale at arbitrary spacing. The available methods are very efficient and can handle good resolution sampling a thousand times faster than other methods based on the Kronecker product. The key is that the problem is not transformed into a nonnegative least squares problem with a matrix-vector formulation. Instead, it is solved in the original dimensions of the kernel matrix, and the resulting problem is solved by a new methodology using a specially developed matrix factorization method. The library was produced at the request of Spinlock SRL, a company devoted to industrial NMR solutions, by contract “Research and development of algorithms for estimation of distributions of relaxation times of nuclear magnetic resonance”, 2013.
Contact: Diego Milone, Leandro Di Persia