KLASIFIKASI TUMOR OTAK BERDASARKAN CITRA MRI BERBASIS MODEL ENSEMBLE LEARNING EFFICIENTNETV2 DAN XCEPTION

Alhamda, Denisa Septalian (2025) KLASIFIKASI TUMOR OTAK BERDASARKAN CITRA MRI BERBASIS MODEL ENSEMBLE LEARNING EFFICIENTNETV2 DAN XCEPTION. Undergraduate thesis, Universitas Pembangunan Nasional Veteran Jawa TImur.

	Text (Cover) 21083010113_Cover.pdf Download (1MB)
	Text (BAB I) 21083010113_BAB I.pdf Download (100kB)
	Text (BAB II) 21083010113_BAB II.pdf Restricted to Registered users only until 31 January 2028. Download (1MB)
	Text (BAB III) 21083010113_BAB III.pdf Restricted to Registered users only until 31 January 2028. Download (1MB)
	Text (BAB IV) 21083010113_BAB IV.pdf Restricted to Registered users only until 31 January 2028. Download (746kB)
	Text (BAB V) 21083010113_BAB V.pdf Download (12kB)
	Text (DAFTAR PUSTAKA) 21083010113_DAFTAR PUSTAKA.pdf Download (106kB)
	Text (DAFTAR LAMPIRAN) 21083010113_LAMPIRAN.pdf Restricted to Registered users only until 31 January 2028. Download (291kB)

Abstract

Brain tumors are a condition in which cells in or around the brain grow abnormally and uncontrollably. Magnetic resonance imaging (MRI) plays a critical role in the diagnosis of brain tumors, where rapid and accurate diagnosis is essential for optimal treatment. Automated systems based on Convolutional Neural Networks (CNNs) provide an effective solution for image classification with different adaptable architectures. The EfficientNetV2S architecture refines the proportional design of EfficientNet by efficiently optimizing depth, width, and resolution while employing techniques such as Neural Architecture Search (NAS) and progressive learning to accelerate training convergence without compromising accuracy. The Xception architecture is a CNN variant that uses depth-separable convolutions, reducing the number of parameters while maintaining the ability to capture complex features. ResNet50, with its residual learning framework, is designed to address the vanishing gradient problem, enabling the network better to detect errors through its very deep network structure. This study combines the EfficientNetV2S, Xception, and ResNet50 models through ensemble learning and implements Grad-CAM (gradient-weighted class activation mapping) to provide model interpretability in brain tumor classification. Grad-CAM identifies areas in MRI images most relevant to the model's decisions, facilitating expert validation of classification results. The integration of these techniques overcomes the limitations of individual architectures while improving the accuracy and interpretability of results. The results show an accuracy of 99.88%, with a precision of 99.85%, a recall of 99.87%, and an F1 score of 99.88%, confirming the effectiveness of the ensemble method in mitigating overfitting in complex architectures. This research proves that the Grad-CAM approach combined with ensemble learning is highly effective and robust for brain tumor classification based on MRI images across four classes: glioma, meningioma, non-tumor, and pituitary. Keywords: Brain tumor classification, EfficientNetV2S, Ensemble Learning, MRI multi-section, ResNet50, Xception.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Diyasa, I Gede Susrama Mas	NIDN197006192021211009	igsusrama.if@upnjatim.ac.id
Thesis advisor	Saputra, Wahyu Syaifullah Jauharis	NIDN198608252021211003	wahyu.s.j.saputra.if@upnjatim.ac.id

Subjects:

Q Science > Q Science (General)
Q Science > QA Mathematics > QA76.6 Computer Programming

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

Denisa Septalian Alhamda

Date Deposited:

31 Jan 2025 08:28

Last Modified:

31 Jan 2025 08:28

URI:

https://repository.upnjatim.ac.id/id/eprint/34460