Performance Comparison of ResNet50, VGG16, and MobileNetV2 for Brain Tumor Classification on MRI Images

Brain tumor classification using MRI images is a significant challenge in medical diagnosis, requiring models with high accuracy and efficient training. This study aims to compare the performance of three Convolutional Neural Network (CNN) models—ResNet50, VGG16, and MobileNetV2—for brain tumor clas...

Full description

Saved in:
Bibliographic Details
Main Authors: Muhammad Bayu Kurniawan, Ema Utami
Format: Article
Language:Indonesian
Published: Islamic University of Indragiri 2025-03-01
Series:Sistemasi: Jurnal Sistem Informasi
Subjects:
Online Access:https://sistemasi.ftik.unisi.ac.id/index.php/stmsi/article/view/5054
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850262380736413696
author Muhammad Bayu Kurniawan
Ema Utami
author_facet Muhammad Bayu Kurniawan
Ema Utami
author_sort Muhammad Bayu Kurniawan
collection DOAJ
description Brain tumor classification using MRI images is a significant challenge in medical diagnosis, requiring models with high accuracy and efficient training. This study aims to compare the performance of three Convolutional Neural Network (CNN) models—ResNet50, VGG16, and MobileNetV2—for brain tumor classification based on MRI images. The dataset consists of four brain tumor categories: glioma, meningioma, pituitary, and no tumor, with data split into training, validation, and testing sets. Each model was evaluated using metrics including accuracy, precision, recall, F1-score, specificity, and training time to assess their effectiveness in predicting brain tumors with optimal accuracy and efficiency. Experimental results indicate that VGG16 achieved the best overall performance, with an accuracy of 94.93%, precision of 94.68%, and specificity of 98.33%, while also having the shortest training time of 47.15 minutes. MobileNetV2 demonstrated strong performance with a recall of 94.08% but required a longer training time of 79.53 minutes. ResNet50 recorded the lowest accuracy (91.67%) despite excelling in precision (91.79%), but it underperformed in recall (91.25%) and specificity (97.2%). Overall, this study confirms that VGG16 is the most efficient and effective model for MRI-based brain tumor classification.
format Article
id doaj-art-e143359f008f419c961f54eec3fbb2ed
institution OA Journals
issn 2302-8149
2540-9719
language Indonesian
publishDate 2025-03-01
publisher Islamic University of Indragiri
record_format Article
series Sistemasi: Jurnal Sistem Informasi
spelling doaj-art-e143359f008f419c961f54eec3fbb2ed2025-08-20T01:55:11ZindIslamic University of IndragiriSistemasi: Jurnal Sistem Informasi2302-81492540-97192025-03-0114276777710.32520/stmsi.v14i2.50541039Performance Comparison of ResNet50, VGG16, and MobileNetV2 for Brain Tumor Classification on MRI ImagesMuhammad Bayu Kurniawan0Ema Utami1Universitas Amikom YogyakartaUniversitas Amikom YogyakartaBrain tumor classification using MRI images is a significant challenge in medical diagnosis, requiring models with high accuracy and efficient training. This study aims to compare the performance of three Convolutional Neural Network (CNN) models—ResNet50, VGG16, and MobileNetV2—for brain tumor classification based on MRI images. The dataset consists of four brain tumor categories: glioma, meningioma, pituitary, and no tumor, with data split into training, validation, and testing sets. Each model was evaluated using metrics including accuracy, precision, recall, F1-score, specificity, and training time to assess their effectiveness in predicting brain tumors with optimal accuracy and efficiency. Experimental results indicate that VGG16 achieved the best overall performance, with an accuracy of 94.93%, precision of 94.68%, and specificity of 98.33%, while also having the shortest training time of 47.15 minutes. MobileNetV2 demonstrated strong performance with a recall of 94.08% but required a longer training time of 79.53 minutes. ResNet50 recorded the lowest accuracy (91.67%) despite excelling in precision (91.79%), but it underperformed in recall (91.25%) and specificity (97.2%). Overall, this study confirms that VGG16 is the most efficient and effective model for MRI-based brain tumor classification.https://sistemasi.ftik.unisi.ac.id/index.php/stmsi/article/view/5054resnet50vgg16mobilenetv2brain tumor classificationmri image
spellingShingle Muhammad Bayu Kurniawan
Ema Utami
Performance Comparison of ResNet50, VGG16, and MobileNetV2 for Brain Tumor Classification on MRI Images
Sistemasi: Jurnal Sistem Informasi
resnet50
vgg16
mobilenetv2
brain tumor classification
mri image
title Performance Comparison of ResNet50, VGG16, and MobileNetV2 for Brain Tumor Classification on MRI Images
title_full Performance Comparison of ResNet50, VGG16, and MobileNetV2 for Brain Tumor Classification on MRI Images
title_fullStr Performance Comparison of ResNet50, VGG16, and MobileNetV2 for Brain Tumor Classification on MRI Images
title_full_unstemmed Performance Comparison of ResNet50, VGG16, and MobileNetV2 for Brain Tumor Classification on MRI Images
title_short Performance Comparison of ResNet50, VGG16, and MobileNetV2 for Brain Tumor Classification on MRI Images
title_sort performance comparison of resnet50 vgg16 and mobilenetv2 for brain tumor classification on mri images
topic resnet50
vgg16
mobilenetv2
brain tumor classification
mri image
url https://sistemasi.ftik.unisi.ac.id/index.php/stmsi/article/view/5054
work_keys_str_mv AT muhammadbayukurniawan performancecomparisonofresnet50vgg16andmobilenetv2forbraintumorclassificationonmriimages
AT emautami performancecomparisonofresnet50vgg16andmobilenetv2forbraintumorclassificationonmriimages