Two-parametric analysis of magnetic-resonance images by the maximum likelihood technique in comparison with the one-parametric approximation

The paper considers a method of the two-parametric analysis of magnetic-resonance image's data which allows getting a joint estimation of both the useful signal and the noise within the image being analyzed on the basis of the maximum likelihood principle. This technique presents an essentially new approach to data processing in the conditions of the Rice distribution and can be efficiently used in information technologies for solving a wide range of tasks connected with the Rician signals' filtering. Solving the two-parametric task is based on measured samples' data only and is not connected with a priori suppositions concerning the noise value which inevitably limit the precision of the one-parametric method. By means of computer simulation, a comparative analysis of the traditional one-parametric method and the elaborated two-parametric method is made for estimation of the useful component of the signal forming the magnetic-resonance image. The statistical data for the shift and the spread of the sought-for signal and noise parameters are calculated. The advantages of the two-parametric technique of data analysis are shown.