Automated Brain Tumor Analysis from MRI Using Deep Learning

Read full text Summarise with VIRAT
Authors: Aamir Ali, Aamir Raza, Iqra Nazir, Swetank Mohan, Khaleelullah Syed, Nikhat Fatima, Misbah Ali
Cited by: 0
References: 26

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Accurate brain tumor classification from MRI remains essential for computer-assisted diagnosis, yet manual interpretation is time-consuming and variable. This study presents an EfficientNet-B0-based convolutional neural network for multi-class classification of glioma, meningioma, pituitary tumors, and no-tumor cases. The model was trained and evaluated on a public MRI dataset of 7023 images using a strict patient-level split to ensure unbiased assessment. A fixed EfficientNet-B0 backbone with a lightweight classification head reduces overfitting while maintaining stable learning. Performance was assessed via accuracy, precision, recall, F1-score, and specificity. The model achieved class-wise (one-vs-rest) accuracies of 96.5% for glioma, 99.1% for no tumor, 95.6% for meningioma, and 97.9% for pituitary tumors, with high specificity across all classes. A controlled comparison against ResNet50 and MobileNetV2 under identical conditions shows that EfficientNet-B0 offers a balanced trade-off between predictive performance and computational cost, achieving competitive accuracy with significantly fewer parameters and faster inference than deeper architectures. This study provides a reproducible evaluation framework, patient-level validation protocol, and systematic backbone comparison to support efficient and reliable deep learning models for multi-class brain tumor classification.

Fields of Study & Subjects

Brain Tumor Detection and Classification Glioma Diagnosis and Treatment AI in cancer detection

Publication details

Journal / source
Biomedical Informatics and Smart Healthcare
Volume
2
Issue
1
Pages
62–78
Publication date
27 March 2026
Language
EN