Unsupervised Learning for MRI Brain Tumor Segmentation with Spatially Variant Finite Mixture Model in Reversible Jump MCMC Algorithm

Abstract

MRI brain tumor segmentation is an important topic in medical image processing. Manual segmentation is risky and time-consuming when the MRI is in low quality. The automatic segmentation can be a solution to manage this problem. This paper proposed an improved modeling approach for unsupervised learning trough Spatially Variant Finite Mixture Model (SVFMM). The main contribution is the automation of the SVFMM algorithm to find the optimum number of clusters. This is achieved by employing the birth-death random process in Bayesian Reversible Jump MCMC. Validation of the proposed model is done by calculating the Correct Classification Ration (CCR) in comparison to the original SVFMM and Gaussian Mixture Model (GMM). The proposed model provides similar performance in image segmentation compared to the original SVFMM but is better than GMM. However, SVFMM-RJMCMC is faster and more efficient in finding the optimum number of clusters.

Machine learning and deep learning for brain tumor MRI image segmentation

“…A hybrid GMM with a spatially variant finite mixture model was proposed to mitigate the sensitivity to noise of GMM by taking account of the spatial dependencies between pixels using a Markov random field model. 31 Testing on brain tumor MRI image segmentation showed that this hybrid model is more accurate in separating region of interest, such as tumor from noise than GMMs. Incorporating the spatial association, Chen et al .…”

“…Therefore, several methods were proposed to integrate spatial information into conventional GMM, with the prior probability of each pixel being determined through the utilization of information from neighboring pixels. 29–32 For example, Ji et al . 29 introduced a fuzzy local GMM method that includes spatial constraints within local GMMs for automated brain tumor MRI image segmentation.…”

Remote Sensing Image Segmentation Based on a Novel Gaussian Mixture Model and SURF Algorithm

“…To automatically determine the number of class members, many scholars have proposed a variable class image segmentation method based on Gaussian mixture model. Pravitasari et al (2021) proposed an image segmentation algorithm based on RJMCMC (reversible jump Markov chain Monte Carlo) method with multivariate Gaussian mixture model. The sampling simulation operation of the algorithm included label field sampling, Gaussian distribution parameter sampling, mixed weight coefficient sampling, MRF parameter sampling and class number sampling.…”