Brain Tumor Detection Using EfficientNet and Variational Quantum Circuits

Abstract

Abstract—Accurate classification of brain tumors from MRI scans is essential for treatment planning and clinical decision support. This work proposes a hybrid quantum–classical deep learning model for multi-class brain tumor classification. The framework combines EfficientNet-B0 for feature extraction with a variational quantum circuit (VQC) layer for nonlinear feature transformation. The model categorizes MRI scans into four classes: glioma, meningioma, pituitary tumor, and normal brain tissue. Experiments conducted on a dataset of 3,264 MRI images show that the hybrid model achieves 85.1% validation accuracy, outperforming the classical EfficientNet baseline by 2.8%. The system also integrates Grad-CAM visualization to highlight image regions influencing predictions, improving interpretability. For practical deployment, a Telegram-based interface enables real-time MRI analysis. The results indicate that lightweight hybrid quantum-classical architectures can be explored as a complementary enhancement to conventional deep learning models in medical image classification tasks.