Related Experiment Videos
QG-WRN: A Quantum-Enhanced Graph Convolutional Wide Residual Network for ASD Diagnosis via Neuroimaging Sensing
Nanting Huang1, Xiaoyu Li2, Xin Yang3
1The School of Physics, University of Electronic Science and Technology of China, Chengdu 611731, China.
Sensors (Basel, Switzerland)
|July 15, 2026
Summary
This study introduces a novel Quantum-Enhanced Graph Convolutional Wide Residual Network (QG-WRN) for autism spectrum disorder (ASD) diagnosis. The QG-WRN accurately identifies brain abnormalities, outperforming existing methods for clinical diagnosis.
Area of Science:
- Neuroscience
- Computational Intelligence
- Quantum Computing
Background:
- Autism spectrum disorder (ASD) pathology involves dual heterogeneity: abnormal local energy metabolism and brain-wide topological failure.
- Advanced neuroimaging captures complex signals reflecting these abnormalities.
- Classical methods face limitations in analyzing intricate network connections and mitigating overfitting in medical data.
Purpose of the Study:
- To develop a synergistic computational tool for characterizing dual heterogeneity in ASD using neuroimaging data.
- To propose a novel Quantum-Enhanced Graph Convolutional Wide Residual Network (QG-WRN) for improved ASD diagnosis.
- To leverage quantum computing's capabilities for analyzing complex brain network topology.
Main Methods:
- A decoupled parallel dual-stream architecture (QG-WRN) was designed, integrating classical and quantum branches.
- The classical branch uses a Wide Residual Network (WRN) for Amplitude of Low-Frequency Fluctuation (ALFF) feature extraction.
- The quantum branch employs a variational quantum graph convolutional (QGCN) module with a 4-qubit configuration to process functional connectivity (FC) matrices, utilizing quantum interference.
Main Results:
- The QG-WRN achieved an accuracy of 68.49% in ASD diagnosis, surpassing 10 classic and recent baselines.
- The method demonstrated effective mitigation of overfitting on small-sample medical data through quantum computing principles.
- Interpretability analysis successfully mapped core disease hubs to standard AAL116 atlas coordinates.
Conclusions:
- The QG-WRN offers a powerful computational intelligence tool for sensor-based clinical diagnosis of ASD.
- The study highlights the potential of integrating quantum computing with deep learning for analyzing complex neurological disorders.
- The findings provide a foundation for computationally aided ASD diagnosis and understanding disease mechanisms.