Parameter Estimation Of Nuclear Magnetic Resonance Spectroscopy Based On Reversible Jump MCMC

The NMR-based metabonomics approach evolved form the pioneering work of Nicholson and co-workers in 1999 has become a novel analytical technique. NMR-based metabonomics are closely related to NMR analysis technology. How to extract form the NMR signal of useful information is a critical part of NMR-based metabonomics research. The NMR signal is considered as a time-domain complex free induction decay(FID) signal which is a sum of exponentially damped sinusoidal components. The detection and estimation problem for damped sinusoidal model has received considerable attention over the past two decades, and this has increased since the advent of NMR-based metabonomics, where it is often both laborious and mathematically challenging to identify all the components.After a biological matrix such as urine, blood plasma or a tissue extract as detected by 1H NMR spectroscopy. Ideally for this application one would like a method which is fully automated but still capable of resolving as many resonances as possible in a complex biological sample. In this paper, a Reversible Jump Markov chain Monte Carlo(RJMCMC) algorithm is applied to signal processing of NMR-based metabonomics. First, after modeling the NMR signal according to its physical characteristics, we derived the posteriori probability from the parameters'prior probability based on Bayesian theory. Then we designed and implemented RJMCMC algorithm with the posterior probability as objective function, and the maximum posteriori probability is corresponding to the optimal parameter. Finally, we tested and analyzed the algorithm based on simulation and real data. In addition, we proposed a filtering algorithm according to signal feature to filter out the low-frequency interference component in the NMR signal.