Quantum-Enhanced Multi-Head Attention for Hybrid Classical-Quantum Neural Networks

Chufan Lyu

Rajalaxmi Hegde, Sandeep Kumar Hegde

Lovish Bhatia, Shreyansh Gupta, Rizil Patel +1 more

Abstract This study demonstrates that a hybrid CNN–QNN architecture outperforms classical CNNs for multi-class retinal OCT classification by leveraging variational quantum circuits to enhance feature separability. Early detection of retinal diseases through optical coherence tomography (OCT) imaging is critical for timely intervention and vision preservation. While classical convolutional neural networks (CNNs) excel at pattern recognition in retinal OCT B-scans, their ability to capture complex, non-linear feature relationships in subtle disease manifestations remains limited. Hybrid quantum-classical architectures offer a promising approach by leveraging variational quantum circuits to enhance feature separability in challenging classification regimes. This study presents a hybrid CNN–QNN architecture for multi-class classification of retinal OCT images dataset (CNV, DME, DRUSEN, NORMAL). A classical CNN backbone extracts hierarchical spatial features from B-scans, which are then processed through parallel classical and quantum classification heads. The quantum head employs a 5-qubit variational quantum circuit with angle embedding and strongly entangling layers to generate quantum-enhanced feature representations. Outputs from both heads are concatenated and passed through a final softmax layer for four-class prediction. The findings establish hybrid quantum-classical neural networks as a viable enhancement to conventional deep learning for retinal OCT analysis, offering improved discriminative performance on established medical imaging benchmarks through quantum feature transformations.

Riza Alaudin Syah, Irwan Alnarus Kautsar, Gunawan Witjaksono +1 more

Breast cancer diagnosis through thermographic image analysis remains a critical challenge in medical AI, with classical deep learning approaches facing limitations in complex thermal pattern classification tasks. This paper presents a novel Hybrid Quantum Neural Network (HQNN) architecture that integrates quantum computing principles with classical convolutional neural networks for enhanced breast cancer classification. Our approach employs parameterized quantum circuits with multi-head attention mechanisms for quantum-aware feature encoding, coupled with classical convolutional layers for comprehensive pattern recognition. The quantum component utilizes a 4qubit variational circuit with strongly entangling layers, while the classical component incorporates advanced attention mechanisms for feature fusion. Experimental validation on breast cancer thermographic data demonstrates substantial performance improvements over state-of-the-art classical architectures, with the quantum-enhanced approach exhibiting superior convergence dynamics and enhanced feature representation capabilities. Our findings provide evidence for quantum advantage in medical image classification through classical simulation, establishing a framework for quantum-classical hybrid systems in healthcare applications. The methodology addresses key challenges in quantum machine learning deployment while maintaining computational feasibility on near-term quantum devices.

R Shyam, G. Chiranjeevi, R. Abhishek Reddy

Dhanya Boyapally, Srikar Alla, Ali Shiri Sichani +2 more

This work presents QSANN-ViQ, an enhanced hybrid quantum-classical architecture for explainable dementia classification using MRI data. The model integrates a lightweight convolutional neural network (CNN) for spatial feature extraction with multiple quantum neural network (QNN) heads of varying qubit sizes (2,4,8), executed on IBM Qiskit simulators. A trustaware routing mechanism dynamically selects the most appropriate quantum path based on inferred confidence, optimizing both accuracy and interpretability. The framework combines SHAP-based quantum explainability with Grad-CAM visualizations from the CNN backbone, offering transparent multi-level reasoning. QSANN-ViQ demonstrates strong classification performance across four dementia stages, with modular design for future quantum hardware integration and practical diagnostic deployment.