| name | qae-mri-anomaly-detection |
| description | Quantum autoencoder (QAE) for compression-driven anomaly detection in brain MRI data - angle encoding into quantum states, variational encoder-decoder with trash qubits, achieving 0.95 slice-level ROC-AUC. |
| category | quantum |
| created | 2026-07-06T00:00:00.000Z |
| source | arXiv:2606.27411 |
Compression-Driven Anomaly Detection in Brain MRI Using Quantum Autoencoder
Source
arXiv:2606.27411 - "Compression-Driven Anomaly Detection in Brain MRI Using an Interpretable Quantum Autoencoder" by Santanu Ganguly, Xing Liang, Dimitrios Makris (2026-06-25)
Overview
A quantum autoencoder (QAE) approach for compression-driven anomaly detection in brain MRI data. Leverages angle encoding to map image patches into quantum states, followed by variational encoder-decoder architecture.
Core Methodology
-
Angle Encoding: Map image patches into quantum states using angle encoding.
-
Variational Encoder-Decoder Architecture: Train to discard information via auxiliary trash qubits. The encoder compresses, the decoder reconstructs, and trash qubits absorb noise/irrelevant information.
-
Anomaly Scoring: Anomaly scores reflect the degree to which inputs resist compression relative to normal data. Higher scores = deviations from the learned normal manifold.
-
Interpretability: Analysis of learned parameters reveals encoder-decoder asymmetry where effective anomaly detection correlates with parameter patterns.
Key Results
- Slice-level ROC-AUC: ~0.95
- Patch-level ROC-AUC: ~0.813
- Outperforms: Classical autoencoder and PCA baselines
- Datasets: Publicly available brain MRI DICOM datasets
Applications
- Brain tumor detection via anomaly detection
- Neurological disease screening on MRI
- Medical imaging quality control
- Unsupervised pathology detection
Trigger Words
quantum autoencoder, brain MRI, anomaly detection, compression, trash qubits, angle encoding, variational quantum circuit, ROC-AUC, medical imaging
Activation
When working with:
- Quantum machine learning for medical imaging
- Anomaly detection on MRI or medical scans
- Quantum autoencoder architectures
- Unsupervised medical diagnosis
- Compression-based anomaly scoring
