A powerful probabilistic model for noise analysis in medical images

摘要

The statistical properties in various medical images demonstrate uncorrelatednoise fluctuations. The signal noise fluctuations are generally due to physicalimaging processes and have nothing to do with the tissue textures. Adding thenoise types (e.g., quantization, electronics, photon) usually degrade medicalimages. The noise variation is usually assumed to be additive with zero-mean,constant variance Gaussian distribution. However, close consideration of differentmedical images indicates the need for better model representation tominimise the noise that can be vital in decision-making. This research proposeda probabilistic method to represent all real-type noise in general medicalimages. The method aims to cover most classical statistical models such asGaussian, lognormal, Rayleigh, Weibull, and Nakagami without a prior examinationto test for fitness. The proposed model was applied to actual clinicalimages to test the performance of the noise originating from the physical processes.The noise is assumed to be additive white Gaussian type with a zeromean and constant variance. The theoretical literature indicates that a nonlinearfunction can better represent noise. This research helps to form a relationshipbetween the image intensity and the noise variance that yields thefitting parameters in the introduced nonlinear function. The validity of theproposed method was proved mathematically and tested using the well knowKolmogorov–Smirnov (K-S) and Akaike Information Criteria (AIC) tests. Themethod was successfully applied to various clinical images such as magneticresonance, x-ray, and panoramic images. The model's performance is comparedwith the classical models using root mean squared error (RMSE), relativeerror (RE), and R~2 as the evaluation matrices. The presented model hasoutperformed all classic models.