AI-Based Multi Scale Convolutional Models for Four Class Brain Tumor Diagnosis
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
The detection and classification of a brain tumor is predominant for effective treatment that gives better patient outcomes. The present study proposes a deep learning system that performs automated classification of brain tumors based on MRI images. In this context, the Plain and Multiscale variants of ResNet-50, EfficientNet-B0, and DenseNet-121 were applied. This system categorizes Brain Tumor MRI Data into the following classes: glioma, meningioma, pituitary tumor and no tumor. For the purpose of enhancing model generalization, each MRI image is preprocessed by resizing, normalizing, and augmenting. Multiscale architectures have the potential to extract tumor features at different resolutions, therefore yielding higher classification accuracy than their plain versions. Comparison of the models shows that Multiscale DenseNet-121 obtained the highest accuracy of 90.11 %, followed by Multiscale EfficientNet-B0 (84.5 %) and Multiscale ResNet-50 (82.30 %), improving its plain variant's performance by 7-18 \%. These observations demonstrate that multiscale architectures, in particular DenseNet-121, can effectively capture discriminative features from brain tumors. The results bring forward the potential of transparent and efficient CNN architecture for clinically active use, computer-assisted brain tumor diagnosis, which outperforms ResNet-50 and EfficientNet-B0 baselines in terms of robustness.
Fields of Study & Subjects
Brain Tumor Detection and Classification
AI in cancer detection
Glioma Diagnosis and Treatment
Publication details
- Journal / source
- International Conference on Intelligent Systems for Communication, IoT and Security (ICISCoIS)
- Pages
- 1–6
- Publication date
- 8 January 2026
- Language
- EN