Scalable next-scale autoregression for medical image generation across anatomical regions
Published in arXiv preprint arXiv:2602.14512 (v3), 2026
MedVAR uses next-scale autoregression to generate CT and MRI images across six anatomical regions, with separate semantic and structural controls. The updated preprint reports training on 438,905 slices from 40 datasets and evaluates generation quality, efficiency, downstream utility, and privacy.
Recommended citation: Zhicheng He, Yunpeng Zhao, Junde Wu, Ziwei Niu, Ziyue Wang, Bohan Li, Zijun Li, Lanfen Lin, Nan Liu, Yueming Jin. (2026). "Scalable next-scale autoregression for medical image generation across anatomical regions." arXiv preprint arXiv:2602.14512 (v3).
Download Paper
