Klasifikasi Tumor Otak Pada Citra Magnetic Resonance Imaging (MRI) Menggunakan Residual Network

Rayyansyah, Arayzi (2026) Klasifikasi Tumor Otak Pada Citra Magnetic Resonance Imaging (MRI) Menggunakan Residual Network. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026221194-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Tingginya angka kasus tumor otak mendorong kebutuhan akan teknologi diagnostik yang efektif, seperti sistem klasifikasi citra. Salah satu metode yang terbukti efektif adalah klasifikasi citra menggunakan citra magnetic resonance imaging (MRI) dengan metode deep learning convolutional neural network (CNN). Namun, metode ini memiliki kelemahan dalam mengatasi masalah vanishing gradient. Oleh karena itu, dalam tugas akhir ini mengusulkan penggunaan model residual network (ResNet), yang dilengkapi dengan skip connection untuk mengatasi masalah vanishing gradient dan meningkatkan performa klasifikasi. Penelitian ini fokus pada klasifikasi jenis tumor otak (glioma, pituitari, dan meningioma) menggunakan arsitektur ResNet50 pada citra MRI dari dataset Kaggle. Evaluasi kinerja model dilakukan dengan menggunakan parameter confusion matrix, akurasi, presisi, recall, f1-score, dan ROC-AUC. Hasil yang diharapkan adalah perangkat lunak yang dapat melakukan klasifikasi tumor otak pada citra MRI secara akurat. Pada dataset usulan, implementasi model dan konfigurasi residual network menghasilkan rata-rata akurasi 95,42%, presisi 95,67%, recall 95,42%, dan F1-score 95,54%. Kinerja model meningkat secara signifikan hingga mencapai rata-rata optimal sebesar 98,23% pada metrik akurasi, presisi, recall, dan f1-score saat menggunakan konfigurasi terbaik, yaitu rasio pembagian data 90:10, ukuran batch 32, dan 20 epoch. Tingginya performa ini selaras dengan nilai ROC-AUC per kelas menunjukkan glioma bernilai 99,93%, meningioma bernilai 99,93%, dan pituitari bernilai 100%. Secara kesuluruhan, tugas akhir ini menunjukan bahwa model residual network usulan dapat melakukan permasalahan klasifikasi tumor otak pada citra MRI.
==============================================================================================================================
The high incidence of brain tumor cases drives the need for effective diagnostic technologies, such as image classification systems. One method proven to be effective is image classification using Magnetic Resonance Imaging (MRI) scans based on deep learning with Convolutional Neural Networks (CNN). However, this method has a drawback in addressing the vanishing gradient problem. Therefore, this final project proposes the use of a Residual Network (ResNet) model, which features skip connections to overcome the vanishing gradient issue and enhance classification performance. This research focuses on classifying types of brain tumors (glioma, pituitary, and meningioma) using the ResNet50 architecture on MRI image datasets from Kaggle. Model performance is evaluated using a confusion matrix, accuracy, precision, recall, f1-score, and ROC-AUC, with the expected outcome being software capable of accurately classifying brain tumors. In the initial testing, the implementation of the baseline residual network model and configuration yielded an average accuracy of 95.42%, precision of 95.67%, recall of 95.42%, and an F1-score of 95.54%. The model's performance significantly improved to an optimal average of 98.23% across accuracy, precision, recall, and f1-score when applying the best configuration, which includes a 90:10 data splitting ratio, a batch size of 32, and 20 epochs. This high performance aligns with the ROC-AUC values per class, which reached 99.93% for glioma, 99.93% for meningioma, and 100% for pituitary. Overall, this final project demonstrates that the proposed residual network model can effectively solve the problem of brain tumor classification on MRI images.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Klasifikasi Citra, Tumor Otak, Magnetic Resonance Imaging, Residual Network.
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Arayzi Rayyansyah
Date Deposited:	28 Jul 2026 04:28
Last Modified:	28 Jul 2026 04:28
URI:	http://repository.its.ac.id/id/eprint/137690

Actions (login required)

View Item