ISLAMIYAH, MIFTAKHUL ILMI DINUL (2018) Fernandez-Steel Skew Normal Mixture Model dengan Pendekatan Bayesian untuk Segmentasi Citra MRI Tumor Otak. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
06211645000014-Undergraduate-Theses.pdf - Accepted Version Download (3MB) | Preview |
Abstract
Teknologi citra digital medis yang sering digunakan oleh pakar kesehatan untuk mendeteksi tumor otak pada pasien adalah Magnetic Resonance Imaging (MRI). Kesulitan dalam mengolah citra digital hasil MRI adalah memisahkan Region of Interest (ROI) dengan objek lain, sehingga perlu dilakukan segmentasi citra. Segmentasi citra dapat dilakukan dengan clustering. Metode clustering yang sering digunakan untuk segmentasi citra adalah Gaussian Mixture Model (GMM). Namun terda-pat kelemahan dari distribusi Gaussian, yaitu sifatnya yang berekor pendek dan simetris sehingga jika memerlukan model dengan ekor lebih panjang dapat didekati dengan banyak komponen distribusi Gaussian dalam membentuk mixture model. Hal tersebut mengakibatkan sifat parsimoni model kurang terjaga. Selain itu, pada kenyataannya histogram citra MRI tumor otak mengindikasikan adanya skewness. Oleh karena itu, alternatif dari permasalahan tersebut dengan menggunakan distribusi Neo-Normal. Pada penelitian ini dilakukan segmentasi citra MRI untuk mendeteksi lokasi tumor otak menggunakan Fernandez-Steel Skew Normal (FSSN) Mixture Model dengan Pendekatan Bayesian. Distribusi FSSN merupakan salah satu distribusi Neo-Normal yang membentuk distribusi Gaussian maupun Student's t yang dapat stabil dalam modus distribusinya. Pendekatan Bayesian digunakan karena pendekatan statistika klasik untuk estimasi parameter distribusi FSSN sangatlah rumit dan kompleks untuk diimplementasikan secara numerik. Hasil analisis menunjukkan bahwa distribusi FSSN lebih mampu merepresentasikan citra MRI tumor otak serta model yang didapatkan untuk segmentasi citra MRI tumor otak lebih parsimoni dibandingkan GMM.
==================================================================Medical digital imaging technology that is often used by health professionals to detect brain tumors in patients is Magnetic Resonance Imaging (MRI). Difficulty in processing digital image of the MRI is to identifying the separating Region of Interest (ROI) with other objects, so image segmentation is needed. Image segmentation can be done by clustering. Clustering method which is often used for image segmentation is the Gaussian Mixture Model (GMM). GMM has started to be abandoned because, in reality, the symmetric distribution approach is less able to explain the MRI data pattern. In addition, the use of symmetric distribution cannot compete for the model parsimony of an asymmetric distribution to model the long and heavy tail pattern of data. It needs more components in GMM. Therefore, an alternative to these problems is to employ the Neo-Normal distribution. Neo-Normal distribution is a relaxation of normality that is more adaptive to various characteristics of data than the Gaussian distribution. In this research, MRI image segmentation was performed to detect the location of brain tumors using Fernandez-Steel Skew Normal (FSSN) Mixture Model with Bayesian Approach. The FSSN distribution is one of the Neo-Normal distributions that can be skewed adaptively but still stable in its mode. Bayesian approach is used because the classical statistical approach for estimating FSSN distribution parameters is very complex to be implemented numerically. The results indicate that FSSN mixture model has a better performance to represent the data pattern of brain tumor MRI, more parsimony, and able to detect the brain tumor more precisely than the original GMM approach.
| Item Type: | Thesis (Undergraduate) |
|---|---|
| Uncontrolled Keywords: | Bayesian, Fernandez-Steel Skew Normal, Mixture Model, 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 Ilmi Dinul Islamiyah |
| Date Deposited: | 09 Jul 2021 09:26 |
| Last Modified: | 09 Jul 2021 09:26 |
| URI: | http://repository.its.ac.id/id/eprint/57305 |
Actions (login required)
 |
View Item |