Quantum-Enhanced Brain Tumor Classification Using Deep Feature Extraction and ANN-Based Compression

Résumé fourni par la source

Brain tumor classification plays a crucial role in early diagnosis and treatment planning. Magnetic Resonance Imaging (MRI) provides detailed structural information, but the high-dimensional nature of MRI data makes accurate classification challenging. In this study, we present a hybrid framework that integrates classical deep learning, artificial neural networks (ANN), and quantum variational circuits for effective brain tumor classification. Features are first extracted from MRI images using multiple deep learning models, including DenseNet101, ResNet101, VGG16, MobileNetV2, and EfficientNetV3. These high-dimensional features are then reduced to an 8-dimensional vector using an ANN, capturing the most relevant information while reducing computational complexity. The reduced features are encoded into a quantum variational circuit, which leverages parameterized rotations and entanglement operations to learn complex patterns among the features. The proposed approach was evaluated using a 5-fold cross-validation strategy on three tumor types: meningioma, glioma, and pituitary tumor. The DenseNet101-based configuration achieved the highest accuracy of 99.23 %, demonstrating that combining classical and quantum computing can significantly improve diagnostic performance and reliability in brain tumor classification.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.