Hybrid-CNNViT: A Deep Learning Framework for Multi-Class Brain Tumor Classification and Computer-Assisted Intervention Support

Md Fahimul Kabir Chowdhury, Jannatul Ferdous, Wael Korani

Keywords: Brain tumor classification, Deep learning, MRI, Vision Transformer

Abstract: Accurate and early detection of brain tumors is critical for effective treatment planning and surgical decision-making. Manual evaluation of Magnetic Resonance Imaging (MRI) scans is time-consuming and subject to inter-observer variability. This study aims to develop a robust deep learning framework that automates tumor classification while supporting computer-assisted intervention (CAI) workflows. We propose a practical Hybrid-CNNViT architecture that integrates the local feature extraction strength of Convolutional Neural Networks (CNNs) with the global contextual understanding of the Vision Transformer (ViT). The model was trained and evaluated on two benchmark brain tumor MRI datasets using ten-fold cross-validation. Performance was assessed through accuracy, precision, recall, F1-score, and ROC metrics. The proposed framework achieved superior performance, obtaining accuracies of 99.29% (95% CI: 99.12-99.46%) and 98.46% (95% CI: 98.19-98.73%) on the two datasets, with consistently high precision, recall, and F1-scores across all tumor classes. Comparative experiments confirmed that the Hybrid-CNNViT outperformed several state-of-the-art pre-trained CNN and transformer-based models, demonstrating strong generalization and stability. The Hybrid-CNNViT framework delivers accurate and interpretable brain tumor classification, positioning it as a potential component of CAI-driven diagnostic systems. By combining efficiency, precision, and scalability, this approach advances automated neuroimaging analysis and offers meaningful support for clinical decision-making in neurosurgical contexts.

Primary Subject Area: Application: Radiology

Secondary Subject Area: Transfer Learning and Domain Adaptation

Registration Requirement: Yes

Reproducibility: https://github.com/fahimulkabir/Hybrid-CNNViT

Visa & Travel: Yes

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Submission Number: 220