Segmentasi Citra Dari Magnetic Resonance Imaging Meningioma Menggunakan Metode 3-Dimensional Volume To Volume Generative Adversarial Network

Urbach, Fitria (0022) Segmentasi Citra Dari Magnetic Resonance Imaging Meningioma Menggunakan Metode 3-Dimensional Volume To Volume Generative Adversarial Network. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 06111840000020_Undergraduate_Thesis-1.pdf Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Meningioma merupakan salah satu tumor otak primer yang paling sering ditemukan. Salah satu tahap penting dalam penanganan tumor adalah segmentasi citra medis, seperti Magnetic Resonance Imaging (MRI). Segmentasi manual oleh dokter radiologi merupakan proses yang memakan waktu dan bergantung pada pengalaman dokter. Penelitian ini bertujuan untuk mengotomatisasi segmentasi citra MRI meningioma dengan menggunakan metode 3-Dimensional Volume-to-Volume Generative Adversarial Network (3D V2V-GAN). Metode ini dipilih karena kemampuannya dalam mempelajari representasi spasial 3D dari data volumetrik MRI secara langsung. Model GAN terdiri dari dua komponen utama, yaitu generator dan diskriminator. Generator bertugas untuk menghasilkan segmentasi tumor, sedangkan diskriminator bertugas untuk membedakan hasil segmentasi model dengan segmentasi asli (ground truth). Hasil eksperimen menunjukkan bahwa metode 3D V2V-GAN mampu memberikan hasil segmentasi yang akurat dengan nilai Dice Similarity Coefficient (DSC) yang kompetitif dibandingkan dengan metode segmentasi berbasis deep learning lainnya. Implementasi metode ini diharapkan dapat membantu tenaga medis dalam mempercepat proses diagnosis dan perencanaan terapi bagi pasien meningioma.
==============================================================================================================================
Meningioma is one of the most common primary brain tumors. An important stage in tumor treatment is medical image segmentation, such as Magnetic Resonance Imaging (MRI). Manual segmentation by radiologists is a time-consuming process and depends on the doctor's experience. This study aims to automate MRI meningioma image segmentation using the 3-Dimensional Volume-to-Volume Generative Adversarial Network (3D V2V-GAN) method. This method was chosen because of its ability to learn 3D spatial representations from volumetric MRI data directly. The GAN model consists of two main components, the generator and the discriminator. The generator is tasked with producing tumor segmentations, while the discriminator is tasked with distinguishing the model's segmentation results from the original segmentation (ground truth). Experimental results show that the 3D V2V-GAN method is able to provide accurate segmentation results with a competitive Dice Similarity Coefficient (DSC) value compared to other deep learning-based segmentation methods. The implementation of this method is expected to assist medical personnel in accelerating the diagnostic process and treatment planning for meningioma patients.

Item Type:	Thesis (Other)
Additional Information:	RSMa 006.42 Urb s-1 2022
Uncontrolled Keywords:	Tumor Otak, Meningioma, Segmentasi Citra, Generative Adversarial Network. Meningioma, Image Segmentation, Generative Adversarial Network.
Subjects:	Q Science > QA Mathematics
Divisions:	Faculty of Mathematics and Science > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Mr. Marsudiyana -
Date Deposited:	05 Jun 2026 01:35
Last Modified:	05 Jun 2026 01:35
URI:	http://repository.its.ac.id/id/eprint/133587

Actions (login required)

View Item