| name | quantum-enhanced-coronary-classification |
| description | Lightweight quantum-enhanced ResNet for coronary angiography (CAG) classification. Combines classical CNN backbones with variational quantum circuits for medical image classification. Use when: coronary angiography analysis, cardiac image classification, lightweight QML models, quantum-enhanced CNNs, operator-dependency reduction in CAG interpretation, or hybrid quantum-classical medical imaging. |
| metadata | {"arxiv_ids":"1809","published":"2026-01-22","tags":["quantum","medical-imaging","coronary","resnet","cnn","lightweight","vqc"]} |
Quantum-Enhanced Coronary Angiography Classification
Description
Lightweight quantum-enhanced ResNet framework combining classical CNN feature extractors with variational quantum circuits for coronary angiography (CAG) classification. Addresses operator-dependency in CAG interpretation by providing consistent, quantum-enhanced analysis of coronary vessel images.
Key insight: Classical CNN handles most feature extraction; quantum circuit acts as a lightweight classifier head, leveraging Hilbert space expressivity for complex decision boundaries with minimal qubits.
Architecture
Hybrid Pipeline
Input Image → CNN Backbone (ResNet) → Feature Vector → Quantum Circuit → Classification
- CNN Backbone: Standard ResNet (pretrained on ImageNet or medical data)
- Feature Compression: Reduce to N dimensions (N = number of qubits)
- Quantum Classifier Head: VQC with N qubits, parameterized layers
- Measurement: Readout yields class probabilities
Quantum Circuit Design
- Encoding: Angle encoding of compressed features
- Ansatz: Hardware-efficient (RY/RZ + CZ entanglement layers)
- Layers: 2-4 layers (shallow to reduce noise impact)
- Readout: Pauli-Z expectation values per qubit
Why Lightweight?
- Only classifier head is quantum (1-4 qubits vs full image encoding)
- Classical CNN handles the computationally heavy feature extraction
- Quantum part focuses on high-dimensional decision boundaries
- Compatible with current NISQ hardware
When to Use
- Coronary angiography image classification
- Reducing operator dependency in CAG interpretation
- Lightweight quantum enhancement of existing CNN models
- Medical imaging where quantum advantage is plausible in classification layer
- Resource-constrained quantum hardware (few qubits available)
Implementation Steps
- Prepare dataset: CAG images with stenosis/severity labels
- Train/freeze CNN: Use pretrained ResNet, freeze early layers
- Design VQC: Match qubit count to compressed feature dimension
- Hybrid training: Backprop through CNN, parameter-shift for VQC
- Evaluate: Compare against CNN-only baseline and clinical expert performance
Error Handling
Feature Dimension Mismatch
- Use PCA/autoencoder to compress CNN features to exact qubit count
- Ensure compressed features retain discriminative information
Quantum Circuit Barren Plateaus
- Initialize with identity or classically-informed parameters
- Use local observables instead of global measurements
- Limit circuit depth to avoid vanishing gradients
Clinical Validation Gap
- Validate against expert cardiologist annotations
- Report both accuracy and clinical relevance metrics (sensitivity, specificity)
- Consider regulatory requirements for clinical deployment
Resources
- Paper: Entity ID 1809 in kg.db
- Related: Hybrid quantum-classical patterns from quantum-medical-diagnosis skill
- Related: VQC design from quantum-neural-architecture skill
