CQH-MPN: A Classical-Quantum Hybrid Prototype Network With Fuzzy Proximity-Based Classification for Early Glaucoma...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Quantum Computing

Background:

• Glaucoma is a leading cause of irreversible blindness globally.
• Accurate and early diagnosis is critical for vision preservation.
• Deep learning for medical imaging requires extensive labeled data, a common limitation in clinical settings.

Purpose of the Study:

• To develop a novel few-shot learning model for glaucoma diagnosis using limited data.
• To integrate quantum computing principles with classical deep learning for enhanced feature representation.
• To improve diagnostic accuracy in resource-scarce environments.

Main Methods:

• Proposed a Classical-Quantum Hybrid Mean Prototype Network (CQH-MPN).
• Utilized a quantum feature encoder for global representation and a classical convolutional encoder for local features.
• Introduced a fuzzy proximity-based metric for improved classification under uncertainty.
• Evaluated on ACRIMA and ORIGA retinal fundus image datasets in 1-shot, 3-shot, and 5-shot settings.

Main Results:

• CQH-MPN achieved 94.50% accuracy on the ACRIMA dataset in the 1-shot setting.
• The model consistently outperformed existing methods across various few-shot configurations.
• Demonstrated robust generalization capabilities in data-scarce scenarios.

Conclusions:

• CQH-MPN offers a viable solution for few-shot glaucoma diagnosis in low-resource settings.
• The hybrid approach effectively combines quantum and classical computing for medical image analysis.
• This work pioneers quantum-augmented few-shot learning for medical diagnostics.