Application Of Optimization In NMR Data Processing

This paper focuses on the application of optimization in NMR data processing, including application of Genetic Algorithm in curve-fitting of NMR data and image denoising of MRI images.Many different algorithms can be used in curve-fitting, such as Simplex Method, Variable Dimension Method and Conjugate-Gradient Method. Common restrictions of these algorithms include requirement of correctly-set initialize values, limited number of parameters to search. Genetic algorithm made a break-through in these aspects and thus gets applied widely. The Simple Genetic Algorithm, when applying to curve-fitting, often shows its weakness in local search capacity and convergence accuracy. To improve this, we proposed a Population Redistribution Genetic Algorithm (PRGA). In the progress of curve-fitting, PRGA adjust the distribution of the population around the optimum solution according to the quality of the solution, thus improves the local search ability of the algorithm. Experimental results prove that PRGA gives better results in curve-fitting compared with simple GA and numeric algorithms.Optimization is also applied to the image denoising. By optimizing a wavelet-domain image denoising algorithm, namely NeighShrink, we proposed an Optimized NeighShrink (ONS) algorithm that is adaptive to the image detail level. ONS uses different threshold and shrinkage for different sub bands according to the statistics and the image detail level. Based on ONS, an algorithm for MRI image denoising (ONS-MRI) is also proposed. Both algorithms outperformed the other thresholding algorithms in our experiments.