A Controlled Benchmark of Quantum-Latent GAN Augmentation for Brain MRI
A new benchmark study has evaluated the effectiveness of quantum-latent generative adversarial networks (GANs) for augmenting brain MRI data. The research found that neither quantum nor classical generators, when matched for parameter count, significantly outperformed real-data-only training for medical image classification. The study suggests that any observed low-data benefit acts as regularization rather than faithful data expansion, with synthetic samples being off-distribution and mode-collapsed in scarce data regimes. The authors released their protocol to encourage rigorous evaluation of quantum generative augmentation in medical imaging. AI
- Hugging Face
- DagsHub
- alphaXiv
- CORE Recommender
- ScienceCast
- Gotit.pub
- magnetic resonance imaging of the brain
- CatalyzeX Code Finder for Papers
- IArxiv Recommender
- Syed Mujtaba Haider
- KL-regularized latent space
- Wasserstein GAN with gradient penalty
- variational quantum generator
- classical generator
- conditional Wasserstein GAN with gradient penalty