Quantum-Enhanced Approach for Brain Tumor Classification

Abstract

This paper investigates a quantum-enhanced ap- proach for brain tumor classification using MRI data. We focus on establishing robust classical deep learning baselines and defining the core components for integrating Quantum Convolutional Neural Networks (QCNNs). Classical models, specifically VGG-16, MobileNetV1, and Xception, are employed via transfer learning on a prepared brain tumor dataset to benchmark performance. Standardized preprocessing from .mat files, data augmentation, and training procedures are detailed. Concurrently, we define the architecture for a potential hybrid QCNN system using PennyLane, featuring a 4-qubit variational quantum circuit designed as a Keras Layer for integration with classical feature extractors. The objective is to lay the groundwork for exploring how quantum- inspired processing, combined with powerful classical feature extraction, might enhance pattern recognition and classification accuracy for this critical medical imaging task, particularly in complex cases. This work provides the necessary data handling protocols, baseline results, and quantum component specifications for future implementation and evaluation of the complete hybrid system.