Metadata-conditioned generative models to synthesize anatomically-plausible 3D brain MRIs

Wei Peng, Tomas M. Bosschieter, Jiahong Ouyang, Robert Paul, Edith V. Sullivan, Adolf Pfefferbaum, Ehsan Adeli, Qingyu Zhao, Kilian M. Pohl

Abstract: Highlights•A generative model to synthesize high-quality MRIs conditioned on metadata.•A method to quantify anatomical plausibility of synthetic MRIs and aging/sex effects.•Anatomical accuracy of synthetic MRIs varies with regional geometric complexity.•Synthetic MRIs can greatly enhance the training of brain-age predictive models.