Segmentasi Area Hippocampus Otak pada Citra Magnetic Resonance Imaging Menggunakan 3D U-Net Berbasis Transfer Learning

Widodo, Ramadhan Sanyoto Sugiharso (2023) Segmentasi Area Hippocampus Otak pada Citra Magnetic Resonance Imaging Menggunakan 3D U-Net Berbasis Transfer Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 07211940000058-Undergraduate_Thesis.pdf] Text
07211940000058-Undergraduate_Thesis.pdf
Restricted to Repository staff only until 1 September 2025.

Download (13MB) | Request a copy

Abstract

Hippocampus merupakan bagian otak yang memiliki peranan penting dalam pembelajaran, memori, dan navigasi spasial untuk manusia. Bagian otak ini rawan terkena penyakit neuropsikiatri seperti epilepsi, alzheimer, dan depresi. Untuk melakukan diagnosa pada penyakit neuropsikiatri yang terjadi pada hippocampus umumnya diketahui dengan bentuk volume. Diperlukan Magnetic Resonance Imaging (MRI) yang memiliki navigasi spasial yang baik untuk mengetahui kondisi kesehatan hippocampus. Saat ini, penandaian Hippocampus Pada Citra MRI masih banyak dilakukan secara manual dan membutuhkan waktu yang lama, serta rentan akan variabilitas antar-pengamat. Di sisi lain, pemrosesan Citra MRI memerlukan komputasi yang tinggi dan membutuhkan waktu yang panjang, terlebih lagi terdapat keterbatasan ketersediaan dataset hippocampus dalam tiga dimensi untuk melatih model deep learning. Dalam penelitian ini, model 3D U-Net yang memiliki kemampuan baik dengan citra medis dirancang dalam mensegmentasi hippocampus. Untuk mengurangi kompleksitas komputasi dan meningkatkan efisiensi waktu, penerapan Transfer Learning pada 3D U-Net diterapkan untuk meningkatkan generalisasi model meskipun dalam dataset terbatas. Dilakukan penelitian dengan berbagai macam skenario menggunakan metode yang diusulkan dengan penerapan 3D U-Net Berbasis Transfer Learning. Berdasarkan pengujian yang telah dilakukan menggunakan berbagai skenario ditemukan hasil model yang diusulkan memiliki rata-rata Dice Score di atas 85% dan IoU Score di atas 75%.
====================================================================================================================================
The hippocampus is a part of the brain that plays an essential role in human learning, memory, and spatial navigation. This part of the brain is prone to neuropsychiatric diseases such as epilepsy, alzheimer’s, and depression. To diagnose neuropsychiatric diseases that occur in the hippocampus, it is commonly known by its volume. Magnetic Resonance Imaging (MRI) with good spatial navigation is required to determine the health condition of the hippocampus. Currently, marking the hippocampus on MRI images is still mostly done manually, takes a long time, and is prone to inter-observer variability. On the other hand, MRI image processing requires high computation and takes a long time. Moreover, there is a limited availability of hippocampus datasets in three dimensions to train deep learning models. In this study, a 3D U-Net model with good capability with medical images is designed to segment the hippocampus. To reduce computational complexity and improve time efficiency, Transfer Learning on 3D U-Net is applied to improve the model’s generalization even in limited datasets. The research was conducted with various scenarios using the proposed method with the application of 3D U-Net Based on Transfer Learning. Based on the tests that have been carried out using various scenarios, it is found that the proposed model has an average Dice Score above 85% and IoU Score above 75%.

Item Type: Thesis (Other)
Uncontrolled Keywords: 3D U-Net, Hippocampus, MRI, Transfer Learning
Subjects: Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Q Science > QA Mathematics > QA336 Artificial Intelligence
Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
R Medicine > R Medicine (General) > R858 Deep Learning
R Medicine > RC Internal medicine > RC78 Diagnosis, Radioscopic--Examinations, questions, etc.
R Medicine > RC Internal medicine > RC78.7.N83 Magnetic resonance imaging.
T Technology > T Technology (General) > T385 Visualization--Technique
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User: Ramadhan Sanyoto Sugiharso Widodo
Date Deposited: 24 Aug 2023 07:34
Last Modified: 24 Aug 2023 07:34
URI: http://repository.its.ac.id/id/eprint/100918

Actions (login required)

View Item View Item