A controlled benchmark finds no significant performance gain from quantum generative models in brain MRI augmentation. Synthetic samples produced by quantum and classical GANs are statistically indistinguishable, with both showing mode collapse and off-distribution samples, especially at low data fractions. The study concludes that quantum augmentation does not outperform classical methods and acts more as regularization than data expansion.