Research On Real-time Optimization Algorithm Of NMR Measurement Parameters

Nuclear Magnetic Resonance?NMR?is able to explore the structure and properties of materials from a microscopic level and has been widely used in the field of medicine,biology,chemistry and so on.A large number of NMR applications in these fields are based on the analysis of the nuclear magnetic resonance relaxation time T₁,T₂ and the self-diffusion coefficient D₀,and fast and accurate measurement of T₁,T₂,D₀ is desired.The measurement results are largely dependent on spectrometer measurement parameters,and the optimal parameters corresponding to different samples vary widely.Without prior knowledge of the sample,the commonly used T₁,T₂,and D₀ measurement methods have problems such as measurement redundancy,dependence on experiences of the operator.We propose a Monte-Carlo-based measurement optimization algorithm,which evaluates the uncertainty of the sample on-the-fly with a random sampling strategy.It arranges a new sampling in position where reduces the sample uncertainty most,that is,the position with maximum sample information.It achieves the adaptive adjustment of the key parameters which control the measurement position inT₁,T₂,and D₀ experiments.The experiments on pure water and glycerol show that Monte-Carlo-based algorithm has good adaptability to samples with different T₁ and D₀ values,and can achieve about 6times acceleration with a difference from manual-measured results less than 4%.However,the Monte Carlo method is inefficient in evaluating the uncertainty of multi-component samples,and the acceleration effect is not obvious in practice.Therefore,we continue on to a Bayesian-based measurement optimization algorithm,which models the posterior distribution of T₁?T₂?D₀ and performs sampling under the guidance of the posterior distribution.It overcomes the inefficiency of Monte Carlo random sampling in evaluating the uncertainty of multi-component samples.The simulation of the mixture of three single-component materials shows that the algorithm can still achieve 3⁴ times acceleration in the T₁ and D₀ experiment of multi-component samples.The systematic error between the result and the accurate value is less than 5%.Considering the inhomogeneity of the single-sided NMR sensor,here we don't consider parameters adaptive adjustment in T₂ measurement experiment.But in the case of magnetic resonance imaging with high homogeneity,the adaptive adjustment of T₂measurement parameters is not fundamentally different from the adaptive adjustment of T₁ and D₀ measurement parameters in single-sided nuclear magnetics.The proposed algorithm is still universal.Since other measurement parameters inT₁ and D₀ experiments can be determined by simple If-Else programming,we can realize one-click measurement by integrating the algorithm into the embedded system chip,which has the possibility of commercialization.