Segmentasi Citra MRI Tumor Otak Menggunakan Modified Stable Student-t Burr Mixture Model Dengan Pendekatan Bayesian

SAFA, MIFTAKHUL ARDI IKHWANUS (2018) Segmentasi Citra MRI Tumor Otak Menggunakan Modified Stable Student-t Burr Mixture Model Dengan Pendekatan Bayesian. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 06211645000013-Undergraduate_Theses.pdf]
Preview
Text
06211645000013-Undergraduate_Theses.pdf - Accepted Version

Download (2MB) | Preview

Abstract

Salah satu pendekatan komputasi untuk mendapatkan gambaran lokasi tumor otak pada citra MRI tumor otak yaitu segmentasi citra digital. Teknik segmentasi citra yang sering digunakan yaitu clustering, dimana pixel dalam citra akan dikelompokkan berdasarkan intensitas warna (derajat keabuan/grayscale) yang sama. Model based clustering merupakan metode pengelompokkan yang mengoptimalkan kemiripan antara objek berdasarkan pada distribusi probabilistik data. Model mixture yang paling sering digunakan dalam model based clustering khusunya pada segmentasi citra adalah Gaussian mixture model (GMM). Namun, histogram pada citra MRI tumor otak cenderung menunjukkan pola yang miring dan tidak simetri. Sehingga penggunaan GMM memiliki kelemahan yaitu kurang fleksibel terhadap bentuk data, karena distribusi normal memiliki bentuk simetris dan berekor pendek. Oleh karena itu diperlukan pendekatan distribusi yang mampu mengatasi penyimpangan dari distribusi normal. Modified Stable Student-t Burr Distribution atau distribusi MSTBurr dikembangkan dengan tujuan membuat distribusi yang adaptif terhadap perubahan data inputnya. Solusi analitis untuk estimasi parameter distribusi MSTBurr bukan pekerjaan yang mudah karena fungsi likelihood dari distribusi stable tidak bisa direpresentasikan sebagai bentuk analisis yang sederhana. Sehingga, cara untuk mendapatkan estimasi parameter dari distribusi MSTBurr adalah menggunakan pendekatan Bayesian dengan Marcov Chain Monte Carlo (MCMC). Hasil analisis menunjukkan bahwa MSTBurr Mixture Model lebih mampu menangkap pola citra MRI tumor otak. ================================================================================================================== One of the computational approach to get an overview of the location of brain tumors in brain tumor MRI images are digital image segmentation. Image segmentation technique frequently used is clustering, where the pixels in the image will be grouped based on the same color intensity (grayscale). Model-based clustering is a grouping method that optimizes the similarity between objects based on the probabilistic distribution of data. The most common mixture model used in model-based clustering, especially in image segmentation is a Gaussian Mixture Model (GMM). However, the histogram of the brain tumor MRI image tends to show a skewed and asymmetric. So the use of GMM has a weakness that is less flexible to the form of data because the normal distribution has a symmetrical shape and short-tailed. Therefore, it needs a distribution approach that able to overcome the deviation from the normal distribution. Modified Stable Student-t Burr Distribution or MSTBurr distribution has been developed with the aim of creating an adaptive distribution of changes to its input data. Analytical solution for estimating MSTBurr distribution parameters is not an easy job because the likelihood function of this distribution cannot be represented as a simple form of analysis. Thus, the way to obtain parameter estimates from the MSTBurr distribution is using the Bayesian approach couple with Markov Chain Monte Carlo (MCMC). The results of the analysis showed that the MSTBurr Mixture Model could have a better ability to capture the pattern image of the MRI brain tumor.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Bayesian, Mixture Model, Modified Stable Student-t Burr, Segmentasi Citra, Tumor Otak
Subjects: Q Science > QA Mathematics > QA278.55 Cluster analysis
Q Science > QA Mathematics > QA279.5 Bayesian statistical decision theory.
Divisions: Faculty of Mathematics, Computation, and Data Science > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User: Miftakhul Ardi Ikhwanus Safa
Date Deposited: 08 Jul 2021 08:48
Last Modified: 08 Jul 2021 08:48
URI: http://repository.its.ac.id/id/eprint/57303

Actions (login required)

View Item View Item