{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"bioRxiv"}],"indexed":{"date-parts":[[2026,9,27]],"date-time":"2026-09-27T14:57:56Z","timestamp":1790521076833,"version":"4.1.0"},"posted":{"date-parts":[[2026,9,22]]},"group-title":"Bioengineering","reference-count":75,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,9,22]],"date-time":"2026-09-22T00:00:00Z","timestamp":1790035200000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"award":["1RF1AG080371"],"award-info":[{"award-number":["1RF1AG080371"]}],"id":[{"id":"https:\/\/ror.org\/01cwqze88","id-type":"ROR","asserted-by":"publisher"}]},{"award":["1R01AG070988"],"award-info":[{"award-number":["1R01AG070988"]}],"id":[{"id":"https:\/\/ror.org\/01cwqze88","id-type":"ROR","asserted-by":"publisher"}]},{"award":["1UM1MH130981"],"award-info":[{"award-number":["1UM1MH130981"]}],"id":[{"id":"https:\/\/ror.org\/01cwqze88","id-type":"ROR","asserted-by":"publisher"}]},{"award":["1R21NS138995"],"award-info":[{"award-number":["1R21NS138995"]}],"id":[{"id":"https:\/\/ror.org\/01cwqze88","id-type":"ROR","asserted-by":"publisher"}]},{"award":["U19AG060909"],"award-info":[{"award-number":["U19AG060909"]}],"id":[{"id":"https:\/\/ror.org\/01cwqze88","id-type":"ROR","asserted-by":"publisher"}]},{"id":[{"id":"https:\/\/ror.org\/042nb2s44","id-type":"ROR","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,9,22]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:p>Brain age models can systematically mispredict age for specific demographic groups, risking biased estimates of neurological health. We present an approach to improve accuracy and reduce racial disparities using synthetic T1 images generated via style-transfer with SuperSynth, an open-source FreeSurfer tool that produces intensity harmonized isotropic images regardless of the input\u2019s contrast or resolution. We refer to the original scans as the real domain and their SuperSynth-derived counterparts as the synthetic domain. Because each synthetic image derives from the participant\u2019s own scan, comparisons between domains hold anatomy and demographic composition constant. We audited fairness by training a neural network across thirteen racial compositions on a diverse cohort (683 White, 605 Black, 431 Asian) in both domains. Our findings highlight three key insights. First, the synthetic domain enhanced both fairness and accuracy, and even models trained on a single demographic showed reduced racial disparity purely from switching domain. The same pattern appeared in two independently developed pretrained models. Second, models trained on synthetic data demonstrated superior out-of-distribution robustness in external clinical testing. Third, augmenting imbalanced datasets with synthetic minority images closed 61% of the fairness gap without requiring new data collection, and unlike explicitly supplying race labels, does not require race as a model input at deployment. Representational analyses indicated that appearance standardization changes which features drive age prediction rather than removing demographic information from the images. By utilizing synthetic data, our approach offers a robust pathway to more equitable and generalizable brain age models.<\/jats:p>","DOI":"10.64898\/2026.09.16.751796","type":"posted-content","created":{"date-parts":[[2026,9,25]],"date-time":"2026-09-25T16:40:18Z","timestamp":1790354418000},"source":"Crossref","is-referenced-by-count":0,"title":["Improving racial fairness in brain age models using style-transfer synthesis"],"prefix":"10.64898","author":[{"ORCID":"https:\/\/orcid.org\/0009-0009-2836-1098","authenticated-orcid":false,"given":"Vittoria","family":"Banchieri","sequence":"first","affiliation":[{"name":"Institute for Biomedical Engineering, University and ETH Zurich, Zurich, Switzerland"},{"name":"Martinos Center for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, United States of America"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Dina","family":"Zemlyanker","sequence":"additional","affiliation":[{"name":"Martinos Center for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, United States of America"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Juan Eugenio","family":"Iglesias","sequence":"additional","affiliation":[{"name":"Martinos Center for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, United States of America"},{"name":"Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Boston, Massachusetts, United States of America"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"name":"the Alzheimer\u2019s Disease Neuroimaging Initiative","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026092707462075000_2026.09.16.751796v1.1","doi-asserted-by":"publisher","DOI":"10.1101\/2025.09.18.25336117"},{"key":"2026092707462075000_2026.09.16.751796v1.2","doi-asserted-by":"publisher","DOI":"10.1016\/j.jacr.2022.06.008"},{"key":"2026092707462075000_2026.09.16.751796v1.3","doi-asserted-by":"publisher","DOI":"10.1162\/IMAG.a.1164"},{"key":"2026092707462075000_2026.09.16.751796v1.4","unstructured":"Alain, G. & Bengio, Y. (2016) Understanding intermediate layers using linear classifier probes."},{"key":"2026092707462075000_2026.09.16.751796v1.5","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-025-26189-5"},{"key":"2026092707462075000_2026.09.16.751796v1.6","doi-asserted-by":"publisher","DOI":"10.1016\/j.jacr.2023.06.025"},{"key":"2026092707462075000_2026.09.16.751796v1.7","doi-asserted-by":"publisher","DOI":"10.1016\/j.nicl.2019.102063"},{"key":"2026092707462075000_2026.09.16.751796v1.8","doi-asserted-by":"publisher","DOI":"10.1111\/j.2517-6161.1995.tb02031.x"},{"key":"2026092707462075000_2026.09.16.751796v1.9","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2023.102789"},{"key":"2026092707462075000_2026.09.16.751796v1.10","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.2216399120"},{"key":"2026092707462075000_2026.09.16.751796v1.11","doi-asserted-by":"publisher","DOI":"10.1101\/2025.10.19.25338298"},{"key":"2026092707462075000_2026.09.16.751796v1.12","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-023-39902-7"},{"key":"2026092707462075000_2026.09.16.751796v1.13","doi-asserted-by":"publisher","DOI":"10.1101\/2023.04.28.538724"},{"key":"2026092707462075000_2026.09.16.751796v1.14","doi-asserted-by":"publisher","DOI":"10.1016\/j.tins.2017.10.001"},{"key":"2026092707462075000_2026.09.16.751796v1.15","doi-asserted-by":"publisher","DOI":"10.1038\/mp.2017.62"},{"key":"2026092707462075000_2026.09.16.751796v1.16","doi-asserted-by":"publisher","DOI":"10.59275\/j.melba.2025-d1g3"},{"key":"2026092707462075000_2026.09.16.751796v1.17","doi-asserted-by":"publisher","DOI":"10.1002\/hbm.25837"},{"key":"2026092707462075000_2026.09.16.751796v1.18","doi-asserted-by":"publisher","DOI":"10.1038\/s42256-021-00338-7"},{"key":"2026092707462075000_2026.09.16.751796v1.19","doi-asserted-by":"publisher","DOI":"10.1101\/2023.12.05.23299222"},{"key":"2026092707462075000_2026.09.16.751796v1.20","doi-asserted-by":"publisher","DOI":"10.1016\/j.mri.2019.05.041"},{"key":"2026092707462075000_2026.09.16.751796v1.21","doi-asserted-by":"publisher","DOI":"10.1002\/hbm.26502"},{"key":"2026092707462075000_2026.09.16.751796v1.22","doi-asserted-by":"publisher","DOI":"10.1007\/978-3031-45249-9_10"},{"key":"2026092707462075000_2026.09.16.751796v1.23","doi-asserted-by":"publisher","DOI":"10.1038\/s41746-022-00720-8"},{"key":"2026092707462075000_2026.09.16.751796v1.24","doi-asserted-by":"publisher","DOI":"10.1177\/08982643221120743"},{"key":"2026092707462075000_2026.09.16.751796v1.25","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2012.01.021"},{"key":"2026092707462075000_2026.09.16.751796v1.26","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2010.01.005"},{"key":"2026092707462075000_2026.09.16.751796v1.27","doi-asserted-by":"publisher","DOI":"10.1038\/s42256-020-00257-z"},{"key":"2026092707462075000_2026.09.16.751796v1.28","doi-asserted-by":"publisher","DOI":"10.1016\/S2589-7500(22)00063-2"},{"key":"2026092707462075000_2026.09.16.751796v1.29","doi-asserted-by":"publisher","DOI":"10.1016\/j.patrec.2025.11.029"},{"key":"2026092707462075000_2026.09.16.751796v1.30","doi-asserted-by":"publisher","DOI":"10.1007\/s12021-024-09694-2"},{"key":"2026092707462075000_2026.09.16.751796v1.31","doi-asserted-by":"publisher","DOI":"10.1109\/MSP.2025.3590806"},{"key":"2026092707462075000_2026.09.16.751796v1.32","doi-asserted-by":"publisher","DOI":"10.1162\/imag_a_00497"},{"key":"2026092707462075000_2026.09.16.751796v1.33","doi-asserted-by":"publisher","DOI":"10.1126\/sciadv.add3607"},{"key":"2026092707462075000_2026.09.16.751796v1.34","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-17899-3_2"},{"key":"2026092707462075000_2026.09.16.751796v1.35","first-page":"448","article-title":"Batch normalization: Accelerating deep network training by reducing internal covariate shift","volume":"37","year":"2015","journal-title":"In: Proceedings of the 32nd International Conference on Machine Learning (ICML)"},{"key":"2026092707462075000_2026.09.16.751796v1.36","doi-asserted-by":"publisher","DOI":"10.1002\/jmri.21049"},{"key":"2026092707462075000_2026.09.16.751796v1.37","first-page":"551","article-title":"Enhancement of fairness in AI for chest x-ray classification","volume":"2024","year":"2025","journal-title":"AMIA Annual Symposium Proceedings"},{"key":"2026092707462075000_2026.09.16.751796v1.38","doi-asserted-by":"publisher","DOI":"10.1016\/j.atherosclerosis.2008.11.018"},{"key":"2026092707462075000_2026.09.16.751796v1.39","doi-asserted-by":"publisher","DOI":"10.1080\/01621459.1952.10483441"},{"key":"2026092707462075000_2026.09.16.751796v1.40","doi-asserted-by":"publisher","DOI":"10.1101\/2024.08.10.24311686"},{"key":"2026092707462075000_2026.09.16.751796v1.41","doi-asserted-by":"publisher","DOI":"10.1016\/j.socscimed.2014.07.022"},{"key":"2026092707462075000_2026.09.16.751796v1.42","doi-asserted-by":"publisher","DOI":"10.1002\/hbm.24588"},{"key":"2026092707462075000_2026.09.16.751796v1.43","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2017.07.005"},{"key":"2026092707462075000_2026.09.16.751796v1.44","doi-asserted-by":"publisher","DOI":"10.1002\/hbm.26422"},{"key":"2026092707462075000_2026.09.16.751796v1.45","doi-asserted-by":"publisher","DOI":"10.1093\/braincomms\/fcaf271"},{"key":"2026092707462075000_2026.09.16.751796v1.46","doi-asserted-by":"publisher","DOI":"10.1001\/jamafacial.2017.0743"},{"key":"2026092707462075000_2026.09.16.751796v1.47","doi-asserted-by":"publisher","DOI":"10.1016\/j.patcog.2011.04.006"},{"key":"2026092707462075000_2026.09.16.751796v1.48","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2020.101871"},{"key":"2026092707462075000_2026.09.16.751796v1.49","doi-asserted-by":"crossref","unstructured":"Pi\u00e7arra, C. & Glocker, B. Analysing race and sex bias in brain age prediction. In: Clinical Image-Based Procedures, Fairness of AI in Medical Imaging, and Ethical and Philosophical Issues in Medical Imaging. Vol. 14242 of Lecture Notes in Computer Science, 2023. Cham: Springer.","DOI":"10.1007\/978-3-031-45249-9_19"},{"key":"2026092707462075000_2026.09.16.751796v1.50","doi-asserted-by":"publisher","DOI":"10.1109\/MetroXRAINE62247.2024.10796114"},{"key":"2026092707462075000_2026.09.16.751796v1.51","doi-asserted-by":"publisher","DOI":"10.3389\/fcvm.2022.859310"},{"key":"2026092707462075000_2026.09.16.751796v1.52","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-022-32186-3"},{"key":"2026092707462075000_2026.09.16.751796v1.53","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-021-01595-0"},{"key":"2026092707462075000_2026.09.16.751796v1.54","doi-asserted-by":"publisher","DOI":"10.1146\/annurev-bioeng-071516-044442"},{"key":"2026092707462075000_2026.09.16.751796v1.55","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2019.06.017"},{"key":"2026092707462075000_2026.09.16.751796v1.56","unstructured":"Sundararajan, M. , Taly, A. & Yan, Q. (2017) Axiomatic attribution for deep networks. https:\/\/arxiv.org\/abs\/1703.01365"},{"key":"2026092707462075000_2026.09.16.751796v1.57","doi-asserted-by":"publisher","DOI":"10.3389\/fpsyt.2021.615754"},{"key":"2026092707462075000_2026.09.16.751796v1.58","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-023-47021-y"},{"key":"2026092707462075000_2026.09.16.751796v1.59","unstructured":"van Breugel, B. , Kyono, T. , Berrevoets, J. & van der Schaar, M. (2021) Decaf: Generating fair synthetic data using causally-aware generative networks. https:\/\/arxiv.org\/abs\/2110.12884"},{"key":"2026092707462075000_2026.09.16.751796v1.60","doi-asserted-by":"publisher","DOI":"10.1038\/s44222-024-00245-7"},{"key":"2026092707462075000_2026.09.16.751796v1.61","unstructured":"van Breugel, B. , Qian, Z. & Schaar, M.V.D. (2023) Synthetic data, real errors: How (not) to publish and use synthetic data. Proceedings of the 40th International Conference on Machine Learning, 34793\u201334808."},{"key":"2026092707462075000_2026.09.16.751796v1.62","doi-asserted-by":"crossref","unstructured":"van Breugel, B. , Seedat, N. , Imrie, F. & van der Schaar, M. (2023) Can you rely on your model evaluation? improving model evaluation with synthetic test data. https:\/\/arxiv.org\/abs\/2310.16524","DOI":"10.52202\/075280-0091"},{"key":"2026092707462075000_2026.09.16.751796v1.63","doi-asserted-by":"publisher","DOI":"10.5555\/1390681.1390693"},{"key":"2026092707462075000_2026.09.16.751796v1.64","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-019-0686-2"},{"key":"2026092707462075000_2026.09.16.751796v1.65","doi-asserted-by":"publisher","DOI":"10.1016\/j.ebiom.2024.105047"},{"key":"2026092707462075000_2026.09.16.751796v1.66","doi-asserted-by":"publisher","DOI":"10.1080\/01676830.2017.1310254"},{"key":"2026092707462075000_2026.09.16.751796v1.67","doi-asserted-by":"publisher","DOI":"10.1016\/j.jalz.2016.10.006"},{"key":"2026092707462075000_2026.09.16.751796v1.68","doi-asserted-by":"publisher","DOI":"10.2307\/3001968"},{"key":"2026092707462075000_2026.09.16.751796v1.69","doi-asserted-by":"crossref","unstructured":"Wu, Y. & He, K. Group normalization. In: Computer Vision \u2013 ECCV 2018, 2018. : Springer.","DOI":"10.1007\/978-3-030-01261-8_1"},{"key":"2026092707462075000_2026.09.16.751796v1.70","doi-asserted-by":"publisher","DOI":"10.1038\/s41746-024-01276-5"},{"key":"2026092707462075000_2026.09.16.751796v1.71","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-024-03113-4"},{"key":"2026092707462075000_2026.09.16.751796v1.72","doi-asserted-by":"publisher","DOI":"10.1016\/j.nicl.2023.103319"},{"key":"2026092707462075000_2026.09.16.751796v1.73","doi-asserted-by":"publisher","DOI":"10.1007\/s41060-017-0058-x"},{"key":"2026092707462075000_2026.09.16.751796v1.74","doi-asserted-by":"publisher","DOI":"10.1038\/s43856-025-01100-5"},{"key":"2026092707462075000_2026.09.16.751796v1.75","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2021.118569"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.09.16.751796","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,27]],"date-time":"2026-09-27T14:46:32Z","timestamp":1790520392000},"score":1,"resource":{"primary":{"URL":"http:\/\/biorxiv.org\/lookup\/doi\/10.64898\/2026.09.16.751796"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,22]]},"references-count":75,"URL":"https:\/\/doi.org\/10.64898\/2026.09.16.751796","relation":{},"subject":[],"published":{"date-parts":[[2026,9,22]]},"subtype":"preprint"}}