MRI BRAIN IMAGE SEGMENTATION BASED ON SPATIALLY CONSTRAINED GAUSSIAN MIXTURE MODEL WITH REVERSIBLE JUMP MARKOV CHAIN MONTE CARLO ALGORITHM

• Main Author: Zaenurdin, Nurdianto
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/39586
• Institution: Institut Teknologi Bandung

One of the Gaussian Mixture Model (GMM) applications that is quite successful is in image segmentation. The GMM application assumes pixel independence so that it can make noise in the Region of Interest (ROI). To overcome this, a lot of research integrates spatial information into Markov Random Field (MRF) approaches. In this study, MRI image segmentation of the brain was performed using Spatially Variant Finite Mixture Model (SVFMM) with each mixture component Gaussian distribution so that in this study it will be called the Spatially Constrained Gaussian Mixture Model (SCGMM). The optimization method and select the model simultaneously so that it leads to the automation of image segmentation. For this purpose the Reversible Jump Markov Chain Monte Carlo (RJMCMC) algorithm is used on the grounds that this method is able to determine the number of mixture components that are not known with certainty. Image segmentation is carried out on 30 brain MRI images with the aim of detecting the existence of brain tumors as an early stage in the diagnosis of brain tumors with a Computer Aided Disgnosis (CAD) system. The results of SCGMM-based image segmentation with optimization using the RJMCMC algorithm are compared with the SVFMM segmentation method with optimization using the Expectation Maximization (EM) algorithm. Segmentation results validation is measured by the value of Missclassification Ratio (MCR) in the range 0 to 1. The experimental results show statistically the SCGMM method with RJMCMC optimization is superior in segmentation accuracy, this is shown by a two-mean test with the value of MCR as variables, the value of T-value = 1.85 is greater than the value of = 1,699 and the P-value is greater than 0.05.