Qonita, Sandra Firda (2018) Segmentasi Citra MRI Tumor Otak Menggunakan Gaussian Mixture Model dan Hybrid Gaussian Mixture Model - Spatially Variant Finite Mixture Model dengan Algoritma Expectation-Maximization. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
06211440000119-Undergraduate_Theses.pdf - Accepted Version Download (1MB) | Preview |
Abstract
Tumor otak merupakan salah satu bagian dari tumor pada sistem saraf. Berbagai penelitian telah dilakukan untuk membantu tenaga medis dalam menangani tumor otak, salah satunya dengan melakukan pendeteksian tumor otak melalui segmentasi citra medis berdasarkan MRI. Pada kasus citra MRI, segmentasi dilakukan untuk memisahkan Region of Interest (ROI) atau segmen yang dianggap penting dalam sudut pandang medis, dengan segmen-segmen lainnya (Non-ROI) termasuk noise. Metode segmentasi citra yang umum digunakan adalah model based clustering dengan Gaussian Mixture Model (GMM). Namun, kelemahan GMM adalah antar pixel pada citra dianggap independen sehingga hasil segmentasi tidak memiliki ketahanan terhadap noise dalam segmentasi citra. Untuk mengurangi efek negatif dari noise, dalam penelitian ini akan digunakan model Markov Random Field (MRF) yang secara penuh mempertimbangkan dependensi spasial antara pixel dan proporsi probabilitas label secara eksplisit akan dimodelkan sebagai vektor probabilitas. Sehingga metode yang digunakan adalah Gaussian Mixture Model (GMM) dan GMM yang dibatasi secara spasial oleh Markov Random Fields, atau yang diberi nama Spatially Variant Finite Mixture Model (SVFMM), dimana inisial parameter didapatkan dari GMM, sehingga model yang diajukan adalah hybrid GMM-SVFMM. Dalam proses inferensi, metode estimasi maximum likelihood digunakan untuk mengestimasi parameter model yang diusulkan menggunakan algoritma Expectation-Maximization (EM). Hasil penelitian menunjukkan bahwa segmentasi citra MRI tumor otak dengan hybrid GMM-SVFMM mampu memberikan hasil yang lebih akurat untuk memisahkan ROI dengan noise, dibandingkan jika menggunakan metode GMM.
==================================================================================================
A brain tumor is one part of the tumor in the nervous system. Various studies have been conducted to assist medical personnel in dealing with brain tumors, one of them is by performing brain tumor detection through image-based medical segmentation of MRI. In the case of MRI, segmentation is performed to separate the Region of Interest (ROI) or segments that are considered important in the medical point of view, with other segments (Non-ROI) including noise. The commonly used image segmentation method is the model-based clustering with Gaussian Mixture Model (GMM). However, the weakness of GMM is that between the pixels in the image are considered independent, so that the segmentation results do not have the noise robustness in image segmentation. To minimize the negative effects of the noise, in this research we will use the Markov Random Field (MRF) model which fully takes into account the spatial dependencies between pixels. The proportion of label of pixels probabilities will be explicitly modeled as probability vectors. At the same time, pixel component functions are also relatively related to neighboring pixels. This scenario could be implemented as the GMM that is spatially limited by MRF, called the Spatially Variant Finite Mixture Model (SVFMM), in which the initial parameter generated from the GMM, so the proposed model is hybrid GMM-SVFMM.. In the inference process, the maximum likelihood estimation method is used to estimate the proposed model parameters using the Expectation-Maximization (EM)algorithm. The results from the correct classification ratio (CCR) showed that MRI-based brain image segmentation couple with hybrid GMM-SVFMM was able to provide more accurate results to separate the ROI with noise compared to GMM.
| Item Type: | Thesis (Undergraduate) |
|---|---|
| Uncontrolled Keywords: | EM, GMM, Markov Random Fields, Segmentasi Citra MRI, SVFMM |
| Subjects: | H Social Sciences > HA Statistics H Social Sciences > HA Statistics > HA31.7 Estimation |
| Divisions: | Faculty of Mathematics and Science > Statistics > 49201-(S1) Undergraduate Thesis |
| Depositing User: | Sandra Firda Qonita |
| Date Deposited: | 05 Aug 2021 22:18 |
| Last Modified: | 05 Aug 2021 22:18 |
| URI: | http://repository.its.ac.id/id/eprint/57088 |
Actions (login required)
 |
View Item |