Feature-Adaptive Fusion in Hybrid Quantum-Classical Biomedical Image Classification

A feature-adaptive hybrid quantum–classical neural network has demonstrated competitive biomedical image classification accuracy and enhanced resilience against image noise, according to a research article published in the journal Quantum Machine Intelligence. On the PathMNIST benchmark, the FAF-HQNN model achieved higher aggregate mean accuracy than all evaluated baselines under both Gaussian noise and Poisson noise environments.

The study addresses key architectural challenges encountered when integrating quantum circuits into practical machine learning frameworks during the current noisy intermediate-scale quantum era. As the authors highlighted, fixed or globally shared fusion strategies used in many existing hybrid models may limit the effective integration of complementary classical and quantum information.

To address the limitations of static fusion models, the research team designed the FAF-HQNN framework specifically for biomedical image classification tasks. The resulting architecture integrates a classical deep feature encoder with a variational quantum circuit, also known as a VQC. At the core of this hybrid structure is a feature-adaptive fusion mechanism developed to dynamically assign weights to classical predictions and quantum predictions. Instead of imposing a single global weighting scheme, this adaptive mechanism computes sample-specific and class-specific weights dynamically across different inputs and target classes.

Under the evaluated settings, FAF-HQNN achieves competitive accuracy on PathMNIST and BloodMNIST compared to internal variants and external hybrid quantum–classical baselines. On PathMNIST, FAF-HQNN achieves higher aggregate mean accuracy than all evaluated baselines under both Gaussian and Poisson noise. These results suggest that feature-adaptive fusion may help improve classification performance within the evaluated framework.

Subject of Research: Feature-adaptive fusion in hybrid quantum-classical neural networks for biomedical image classification

Article Title: Feature–adaptive fusion in hybrid quantum–classical neural networks for robust biomedical image classification

Article References: Hou, Y.-Y., Li, J., Ye, C., Li, H., Wang, Z., & Liu, Q. (2026). Feature–adaptive fusion in hybrid quantum–classical neural networks for robust biomedical image classification. Quantum Machine Intelligence, 8(2), Article 122. https://doi.org/10.1007/s42484-026-00467-5

Image Credits: AI Generated

DOI: 10.1007/s42484-026-00467-5

Keywords: quantum machine intelligence, hybrid quantum-classical neural networks, biomedical image classification, feature-adaptive fusion, variational quantum circuit, PathMNIST, BloodMNIST, noisy intermediate-scale quantum era, Feature, adaptive, fusion, hybrid