Fernandez-Steel Skew Normal Mixture Model dengan Pendekatan Bayesian dan Markov Random Field untuk Segmentasi Dinding Ventrikel Kiri Jantung

Fahira, Miftah (2026) Fernandez-Steel Skew Normal Mixture Model dengan Pendekatan Bayesian dan Markov Random Field untuk Segmentasi Dinding Ventrikel Kiri Jantung. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003241049-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Myocardial infarction (MI) merupakan penyakit kardiovaskular yang ditandai oleh kerusakan jaringan miokard akibat gangguan aliran darah koroner, yang sering kali berdampak pada perubahan struktur dan fungsi dinding ventrikel kiri. Segmentasi dinding ventrikel kiri pada citra ultrasound jantung menjadi langkah penting dalam mendukung evaluasi klinis dan analisis MI, namun masih menghadapi tan-tangan berupa noise speckle, variasi anatomi, dan ketidakpastian spasial. Penelitian ini mengusulkan metode Spatially Constrained Fernandez–Steel Skew Normal Mixture Model dengan spatial prior, di mana estimasi parameter dilakukan menggunakan pendekatan Bayesian. Model Fernandez–Steel Skew Normal dipilih karena kemampuannya dalam merepresentasikan distribusi intensitas piksel yang asimetris dan berekor panjang, yang tidak dapat dimodelkan secara optimal oleh distribusi Gaussian konvensional. Spatial prior digunakan untuk menangkap keterkaitan spasial antar piksel sehingga meningkatkan konsistensi segmentasi. Eksperimen pada dataset citra ultrasound jantung pasien myocardial infarction menunjukkan bahwa metode yang diusulkan memberikan performa yang lebih unggul dibandingkan dua metode pembanding, dengan nilai Dice Similarity Coefficient (DSC) mencapai sekitar 86.78%, sehingga berpotensi mendukung analisis citra medis yang lebih akurat dalam konteks evaluasi klinis myocardial infarction.
=================================================================================================================================
Myocardial infarction (MI) is a cardiovascular disease characterized by myocardial damage caused by impaired coronary blood flow, which often leads to structural and functional alterations of the left ventricular wall. Segmentation of the left ventricular wall in cardiac ultrasound images is therefore an important step in supporting clinical evaluation and MI analysis; however, this task remains challenging due to speckle noise, anatomical variability, and spatial uncertainty. This study proposes a Spatially Constrained Fernandez–Steel Skew Normal Mixture Model with a spatial prior, in which parameter estimation is performed using a Bayesian approach. The Fernandez–Steel Skew Normal model is selected for its ability to represent asymmetric and heavy-tailed pixel intensity distributions that cannot be adequately modeled by conventional Gaussian distributions. The spatial prior is incorporated to capture spatial dependencies among neighboring pixels, thereby enhancing segmentation consistency. Experimental results on a dataset of cardiac ultrasound images from myocardial infarction patients demonstrate that the proposed method outperforms two comparative approaches, achieving a Dice Similarity Coefficient (DSC) of approximately 86.78%, and thus shows strong potential for supporting more accurate medical image analysis in the clinical evaluation of myocardial infarction.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Fernandez-Steel Skew Normal, Bayesian, Markov Random Field, Spatial Prior, Fernandez-Steel Skew Normal, Bayesian, Markov Random Field, Spatial Prior
Subjects:	Q Science Q Science > QA Mathematics > QA279.5 Bayesian statistical decision theory.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Miftah Fahira
Date Deposited:	29 Jan 2026 09:34
Last Modified:	29 Jan 2026 09:34
URI:	http://repository.its.ac.id/id/eprint/131071

Actions (login required)

View Item