Ensemble combination of CNN for MRI-based brain tumor classification

Sidqi, Akbar, Santoso, Irwan Budi ORCID: https://orcid.org/0000-0001-5586-9035 and Harini, Sri ORCID: https://orcid.org/0000-0001-9664-027X (2025) Ensemble combination of CNN for MRI-based brain tumor classification. International Journal of Engineering Continuity, 4 (1). pp. 71-84. ISSN 2963-2390

Preview

Text 23639.pdf - Published Version Available under License Creative Commons Attribution Share Alike. Download (1MB) | Preview

Abstract

Classifying 17 types of brain tumors remains a major challenge in the medical field, especially in improving diagnostic accuracy and accelerating patient care. This study proposes a CNN-based model with an ensemble combination approach to improve accuracy by integrating multiple architectures through Majority Voting and Weighted Average for more reliable predictions. The models are evaluated using accuracy, precision, recall, and F1-score metrics. The results show that CNN3 with Nadam achieves the best performance (accuracy: 0.90–0.91), outperforming CNN1 (0.87–0.89) and CNN2 (0.82–0.87). The ensemble combination improves accuracy across all models, with CNN3 achieving the highest accuracy (0.96), followed by CNN1 (0.94–0.95) and CNN2 (0.91–0.92). This study demonstrates that the ensemble combination approach can improve the performance of brain tumor classification using deep learning, contributing to faster and more accurate medical diagnosis. Furthermore, these findings open up opportunities for further research in advancing brain tumor detection systems.

Item Type:	Journal Article
Keywords:	Ensemble Combination; Convolutional Neural Network; Brain Tumor Classification
Subjects:	08 INFORMATION AND COMPUTING SCIENCES > 0801 Artificial Intelligence and Image Processing > 080199 Artificial Intelligence and Image Processing not elsewhere classified
Divisions:	Faculty of Technology > Department of Informatics Engineering
Depositing User:	irwan budi santoso
Date Deposited:	06 May 2025 14:03

Downloads

Origin of downloads

Actions (login required)

View Item