A Precise Classification of Medical Images with aCost-Efficient Quantum-Classical Hybrid Neural Network

초록

This paper presents a comprehensive study on a cost-efficient hybrid quantum-classical neural network for scalable medical image classification. By integrating a compact 4-qubit variational quantum circuit (VQC) into a lightweight convolutional neural network (CNN) backbone, the proposed architecture achieves 99.89% test accuracy, requiring remarkably less numbers of trainable parameters and less training times compared to its well-known classical counterparts (CNN, ResNet-18, and VGG-16) for a large-scale dataset comprising the six clinically relevant classes across three different medical image modalities (X-ray, MRI & CT scan). The seamless fusion of quantum circuit outputs with classical CNN features, which drives rigorous and costefficient learning under limited parameter budgets, not only demonstrate the hybrid quantum-classical neural network can serve as a practical, energy-efficient deep learning frameworks in healthcare applications, but also indicates its strong potential for the broad applicability to other image tasks.

키워드