Label noise is prevalent in medical image datasets, arising from observer disagreement and ambiguous cases and deep networks memorize the mislabeled examples at the cost of generalization. Current solutions such as loss correction and sample selection add training complexity and generally assume the noise rate is known or estimable. This paper investigates whether robustness can be obtained architecturally, by routing a frozen backbone's features through a low-dimensional bounded bottleneck and classifying them with a head whose function class is intrinsically smooth. Variational quantum circuits with angle encoding and a linear cost function realise exactly such a class. Their use in medical imaging, however, has been limited by the small number of qubits available on noisy intermediate-scale devices, which typically restricts inputs to low-resolution grayscale images. We therefore couple a frozen ImageNet-pretrained ViT-B/16 backbone to a trainable linear bottleneck and an eight-qubit variational quantum circuit, allowing 224x224 RGB inputs to be processed. Under clean labels the model leads five of six MedMNISTv2 benchmarks against published hybrid quantum-classical models, with margins of 17.4 and 15.7 percentages on BloodMNIST and PathMNIST. The advantage emerges under corrupted labels. At symmetric noise rates of 30% and 50% the model outperforms many published quantum and noise-robust classical baselines on all six datasets without any explicit noise-handling component. Against a control that shares the backbone, bottleneck, parameter count and optimizer and replaces only the circuit with a ReLU layer, it retains 89.75% accuracy on PathMNIST at 50% noise where the control falls to 78.96%. Direct measurement of training dynamics shows that this advantage is not an inability to fit corrupted labels, which the circuit retains, but greater stability under prolonged training on them.
Comments
Accepted at the Second AAAI Symposium on Quantum Information & Machine Learning (QIML), AAAI 2026 Fall Symposium Series. 8 pages, 1 figure
Label noise is prevalent in medical image datasets, arising from observer disagreement and ambiguous cases and deep networks memorize the mislabeled examples at the cost of generalization. Current solutions such as loss correction and sample selection add training complexity and generally assume the noise rate is known or estimable. This paper investigates whether robustness can be obtained architecturally, by routing a frozen backbone's features through a low-dimensional bounded bottleneck and classifying them with a head whose function class is intrinsically smooth. Variational quantum circuits with angle encoding and a linear cost function realise exactly such a class. Their use in medical imaging, however, has been limited by the small number of qubits available on noisy intermediate-scale devices, which typically restricts inputs to low-resolution grayscale images. We therefore couple a frozen ImageNet-pretrained ViT-B/16 backbone to a trainable linear bottleneck and an eight-qubit variational quantum circuit, allowing 224x224 RGB inputs to be processed. Under clean labels the model leads five of six MedMNISTv2 benchmarks against published hybrid quantum-classical models, with margins of 17.4 and 15.7 percentages on BloodMNIST and PathMNIST. The advantage emerges under corrupted labels. At symmetric noise rates of 30% and 50% the model outperforms many published quantum and noise-robust classical baselines on all six datasets without any explicit noise-handling component. Against a control that shares the backbone, bottleneck, parameter count and optimizer and replaces only the circuit with a ReLU layer, it retains 89.75% accuracy on PathMNIST at 50% noise where the control falls to 78.96%. Direct measurement of training dynamics shows that this advantage is not an inability to fit corrupted labels, which the circuit retains, but greater stability under prolonged training on them.
Classification of medical images plays a vital role in medical image analysis; however, it remains challenging due to the limited availability of labeled data, class imbalances, and the complexity of medical patterns. To overcome these challenges, we propose a novel Hybrid Quantum-Classical Neural Network (HQCNN) for both binary and multi-class classification. The architecture of HQCNN integrates a five-layer classical convolutional backbone with a 4-qubit variational quantum circuit that incorporates quantum state encoding, superpositional entanglement, and a Fourier-inspired quantum attention mechanism. We evaluate the model on six MedMNIST v2 benchmark datasets. The HQCNN consistently outperforms classical and quantum baselines, achieving up to 99.91% accuracy and 100.00% AUC on PathMNIST (binary) and 99.95% accuracy on OrganAMNIST (multi-class) with strong robustness on noisy datasets like BreastMNIST (87.18% accuracy). The model demonstrates superior generalization capability and computational efficiency, accomplished with significantly fewer trainable parameters, making it suitable for data-scarce scenarios. Our findings provide strong empirical evidence that hybrid quantum-classical models can advance medical imaging tasks.
Medical image classification is a critical component of modern healthcare; however, accurate diagnosis remains challenging due to limited annotated datasets, class imbalance, and the high dimensionality of medical imaging data. To address these challenges, a hybrid quantum-classical neural network (HQCNN) is proposed, integrating classical deep learning with variational quantum learning for medical image classification. The proposed architecture combines a five-layer convolutional neural network (CNN) for hierarchical feature extraction with a lightweight 4-qubit variational quantum circuit (VQC) incorporating quantum state encoding, superposition and entanglement mechanisms, and a quantum attention-Fourier (QAF) module. This hybrid design aims to improve nonlinear feature representation and quantum parameter efficiency while maintaining a shallow quantum circuit suitable for noisy intermediate-scale quantum (NISQ)-era constraints. Experimental evaluation on six MedMNIST benchmark datasets demonstrated competitive performance across both binary and multi-class classification tasks. HQCNN achieved 98.88% accuracy on the binary subset of PathMNIST (classes 0 vs. 1), 97.61% accuracy on the multi-class OrganAMNIST dataset, and 86.29% accuracy on BreastMNIST. Comparative experiments and statistical analyses demonstrated consistent improvements over the controlled BHQNN baseline, while component-wise ablation studies showed that the QAF module, superposition and entanglement mechanisms, and expressive parameterized rotations contributed to classification performance in a complementary and dataset-dependent manner. Moreover, HQCNN reduced the number of trainable quantum parameters by approximately 55.6% compared with the baseline hybrid quantum neural network (BHQNN). Noise-aware simulations further showed that the model retained relatively stable predictive performance under moderate depolarizing noise, supporting further evaluation under near-term quantum computing conditions. Overall, the results demonstrate that HQCNN provides a parameter-efficient hybrid quantum-classical framework for medical image classification and offers a promising foundation for further investigation of quantum-enhanced medical image analysis.
Medical image classification has been a fundamental task in computer-aided diagnosis. In this work, we propose a hybrid quantum-classical deep learning framework that integrates a parameterized quantum layer into the ResNet-18 architecture for medical image classification. Our experiments are conducted on the OrganMNIST-Axial, OrganMNIST-Coronal, and OrganMNIST-Sagittal datasets from the MedMNIST v2 collection, each consisting of 28×28 grayscale images representing different organ sections. The quantum layer is implemented as a 6-qubit variational quantum circuit (VQC) that partially replaces the classical fully connected module, enabling the network to exploit quantum feature mapping and entanglement for enhanced nonlinear representation learning. This architectural modification reduces the number of classical trainable parameters by approximately 44% compared with the original ResNet-18 while preserving its representational capacity. We evaluate the proposed QResNet18 against classical baselines to investigate its potential advantages in terms of accuracy, robustness, and parameter efficiency. Experimental results demonstrate that the hybrid quantum-classical model achieves competitive or superior performance while maintaining a significantly reduced parameter count, highlighting the feasibility and promise of quantumenhanced medical image classification.
Accurate classification of microscopic blood cells is still a critical task in medical image analysis, where subtle variations and limited data can challenge conventional deep learning models. As such, we investigate in this work the potential of Hybrid Quantum-Classical Neural Networks (HQNNs) to enhance feature representation and improve classification performance in this domain. We propose a modular architecture combining a pre-trained ResNet-50 backbone with a low-dimensional latent bottleneck and a variational quantum circuit, enabling a direct comparison between quantum-enhanced and purely classical transformation mechanisms. To isolate the contribution of the quantum component, we evaluate three architectures: a HQNN model, a Classical Matched Model with an additional nonlinear transformation layer of comparable capacity, and a baseline model without an intermediate transformation stage. Experiments conducted on two publicly available blood cell datasets, namely the Blood Cell Images dataset and the PBC dataset, demonstrate that HQNNs consistently achieve superior or more balanced performance across evaluation metrics. In the Blood Cell Images Dataset, the proposed approach improves macro F1-score by up to 3.7% compared to classical baselines, while improving the F1-score from 98.54% to 98.69% in the more challenging 8-class scenario with near-saturated performance. Additional evaluation on IBM quantum hardware shows that the model remains robust under noise, with only a modest performance degradation relative to simulated results. These results indicate that quantum feature transformations can enhance discriminative representations, particularly in challenging classification scenarios, and highlight the practical potential of HQNN models for medical imaging tasks.
Hybrid quantum-classical neural networks provide a promising approach for incorporating quantum circuits into machine learning in the noisy intermediate-scale quantum regime. However, existing hybrid models often rely on fixed or globally shared fusion strategies, which may limit their ability to exploit complementary information carried by quantum branches, especially under distribution shifts. In this work, we propose a Feature-Adaptive Fusion Hybrid Quantum-Classical Neural Network (FAF-HQNN) for biomedical image classification. The model combines a classical deep feature encoder with a variational quantum circuit (VQC) and introduces a feature-adaptive fusion mechanism to dynamically weight classical and quantum predictions. We evaluate FAF-HQNN on two MedMNIST benchmarks, PathMNIST and BloodMNIST, under clean and corrupted test conditions. FAF-HQNN achieves the strongest overall performance on clean data among the compared methods and shows improved robustness under Gaussian, salt-and-pepper, and Poisson corruptions. Further analysis of circuit layout, measurement basis, and depth shows that even shallow variational quantum circuits can provide useful complementary information. These results demonstrate that feature-adaptive fusion is an effective strategy for improving accuracy and robustness in hybrid quantum-classical models for biomedical image classification.
Quantum transfer learning combines pretrained classical deep learning models with quantum circuits to reuse expressive feature representations while limiting the number of trainable parameters. In this work, we introduce a family of compact quantum transfer learning architectures that attach variational quantum classifiers to frozen convolutional backbones for image classification. We instantiate and evaluate several classical-quantum hybrid models implemented in PennyLane and Qiskit, and systematically compare them with a classical transfer-learning baseline across heterogeneous image datasets. To ensure a realistic assessment, we evaluate all approaches under both ideal simulation and noisy emulation using noise models calibrated from IBM quantum hardware specifications, as well as on real IBM quantum hardware. Experimental results show that the proposed quantum transfer learning architectures achieve competitive and, in several cases, superior accuracy while consistently reducing training time and energy consumption relative to the classical baseline. Among the evaluated approaches, PennyLane-based implementations provide the most favorable trade-off between accuracy and computational efficiency, suggesting that hybrid quantum transfer learning can offer practical benefits in realistic NISQ era settings when feature extraction remains classical.