Bayesian DOSY: A New Approach to Diffusion Data Processing

by Stanislav Sykora (Extra Byte, Castano Primo, Italy)
and Carlos Cobas (Mestrelab Research, Santiago de Compostela, Spain)

at the XXXVIII National Congress on Magnetic Resonance of GIDRM,
Bressanone - Brixen (Italy), September 10-13, 2008.

Presentation SLIDES (PDF, 2228 KB). NMR Articles and Posters

Presentation Abstract

Certain types of multi-array NMR techniques present problems with the visual representation of the results in terms of the physical parameter associated with each spectral peak. This regards, in particular, the Diffusion Ordered Spectroscopy (DOSY) where the non-spectral parameter is the diffusion coefficient, and also ROSY (Relaxation Ordered Spectroscopy) for which it is the relaxation time. In such cases, the transformation of the original data set into a suitable final 2D representation (such as, for example, the DOSY transform) is conceptually difficult to manage.

Current approaches are often little more than graphic sketches based on the evaluation of the decays at a selected set of spectral points (usually those corresponding to peak tops). Unfortunately, the physical significance of such graphic representations does not exceed that of tabulated results and, more generally, the approaches can be criticized for treating different spectral points in a different way (trying to extend them to spectral regions with little or no signal gives rise to characteristic, noise-induced artifacts). Other, physically more acceptable approaches, such as the MaxEnt algorithm of Marc-Andre Delsuc [1] et al, unfortunately require very long processing times.

We have developed [2] a novel Bayesian approach to this problem which is computationally very efficient and physically eminently meaningful, treats all data points in the same way, and gives very satisfactory and artifact-free results. Applied specifically to DOSY data sets, it leads to what we call the BDT algorithm (standing for Bayesian DOSY Transform). It is also an excellent example of the broad class of Bayesian approaches to the evaluation of NMR data.

The BDT algorithm is the core of the BayDOSY evaluation and presentation method implemented in Mnova, which is now beta-tested in practice on real experimental data. Apart from BDT, BayDOSY includes also a novel algorithm for a substantial improvement of the resolution in the diffusion-constant dimension and a contextual improvement in the alignment of peaks belonging to the same molecular species. Ongoing work aims at Bayesian handling of overlapping spectral peaks belonging to different sample components.

In this talk we will discuss in detail the BDT algorithm and illustrate on practical examples the performance of the BayDOSY method.


[1] M.A.Delsuc, DOSY and diffusion measurements in modern NMR,
    a talk presented at the 23rd NMR Valtice meeting in Valtice (Czech Republic), April 20-23, 2008.
[2] C.Cobas, M.Sordo, N.Larin, S.Sykora, Novel Data Evaluation Algorithms: Bayesian DOSY and ROSY Transforms,
    a poster presented at 49th ENC Conference, Asilomar, CA (USA), March 9-14, 2008

Stan Sykora's Publications, Posters and Courses & Talks Stan's Library | Stan's HUB
Copyright ©2008 Sýkora Stanislav,    DOI: 10.3247/SL2Nmr08.015 Page design by Stan Sýkora