{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"medRxiv"}],"indexed":{"date-parts":[[2026,6,20]],"date-time":"2026-06-20T10:52:42Z","timestamp":1781952762188,"version":"3.54.5"},"posted":{"date-parts":[[2026,6,17]]},"group-title":"Radiology and Imaging","reference-count":20,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,6,17]],"date-time":"2026-06-17T00:00:00Z","timestamp":1781654400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"funder":[{"name":"National Cancer Institute","award":["R01CA264987"],"award-info":[{"award-number":["R01CA264987"]}]},{"name":"National Cancer Institute","award":["P01CA281826"],"award-info":[{"award-number":["P01CA281826"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,6,17]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:p>Fine-tuning can adapt pretrained medical imaging models to new clinical datasets, but device-specific domain shifts may limit generalizability. We evaluated Mirai, a mammography-based deep learning model for breast cancer risk prediction, in a large screening cohort containing Hologic and General Electric (GE) full-field digital mammography systems, including GE Premium View (GE PV) and Tissue Equalization (GE TE) post-processing software. Native Mirai showed lower performance on TE images than on Hologic or PV images. Fine-tuning on TE images improved TE performance, particularly for short-term risk prediction, but substantially reduced performance on Hologic images, consistent with catastrophic forgetting. To mitigate this effect, we developed a device-invariant model using interleaved multi-device sampling and conditional adversarial training. This approach largely restored Hologic performance while maintaining improved TE performance, providing better robustness across heterogeneous imaging platforms. Comparison of cumulative and annual risk AUCs over a five-year time horizon further showed that performance gains were driven mainly by short- and intermediate-term predictions. These findings highlight both the value and dangers of device-specific fine-tuning and support balanced domain-adaptation strategies for deploying mammography-based risk models across diverse clinical imaging environments.<\/jats:p>","DOI":"10.64898\/2026.06.15.26355696","type":"posted-content","created":{"date-parts":[[2026,6,18]],"date-time":"2026-06-18T00:50:20Z","timestamp":1781743820000},"source":"Crossref","is-referenced-by-count":0,"title":["Cross-Device Adaptation of Mirai for Mammography-Based Breast Cancer Risk Prediction"],"prefix":"10.64898","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-0211-3629","authenticated-orcid":false,"given":"Adriana","family":"Sistig","sequence":"first","affiliation":[{"name":"Department of Artificial Intelligence and Human Health, Icahn School of Medicine at Mount Sinai, New York, NY, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8618-1660","authenticated-orcid":false,"given":"Joseph H.","family":"Rothstein","sequence":"additional","affiliation":[{"name":"Department of Epidemiology, University of Texas MD Anderson Cancer Center, Houston, TX, USA"},{"name":"Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, NY, USA"},{"name":"Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0005-8338-0997","authenticated-orcid":false,"given":"Tejomay","family":"Gadgil","sequence":"additional","affiliation":[{"name":"Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ninah","family":"Achacoso","sequence":"additional","affiliation":[{"name":"Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8988-6976","authenticated-orcid":false,"given":"Stacey E.","family":"Alexeeff","sequence":"additional","affiliation":[{"name":"Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0009-7463-5846","authenticated-orcid":false,"given":"Lawrence D.","family":"Gerstley","sequence":"additional","affiliation":[{"name":"Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3539-5391","authenticated-orcid":false,"given":"Robert J.","family":"Klein","sequence":"additional","affiliation":[{"name":"Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, NY, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7037-7570","authenticated-orcid":false,"given":"Laurie R.","family":"Margolies","sequence":"additional","affiliation":[{"name":"Department of Diagnostic, Molecular and Interventional Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Albert","family":"Pu","sequence":"additional","affiliation":[{"name":"Kaiser Permanente Information Technology, Oakland, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0006-9858-8886","authenticated-orcid":false,"given":"Cara L. Smith","family":"Gueye","sequence":"additional","affiliation":[{"name":"Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Marvella","family":"Villase\u00f1or","sequence":"additional","affiliation":[{"name":"Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mark","family":"Westley","sequence":"additional","affiliation":[{"name":"Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Laurel A.","family":"Habel","sequence":"additional","affiliation":[{"name":"Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4315-3106","authenticated-orcid":false,"given":"Vignesh A.","family":"Arasu","sequence":"additional","affiliation":[{"name":"Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA"},{"name":"Department of Radiology, Kaiser Permanente Northern California, Vallejo, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0085-1190","authenticated-orcid":false,"given":"Weiva","family":"Sieh","sequence":"additional","affiliation":[{"name":"Department of Epidemiology, University of Texas MD Anderson Cancer Center, Houston, TX, USA"},{"name":"Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, NY, USA"},{"name":"Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5190-2851","authenticated-orcid":false,"given":"Li","family":"Shen","sequence":"additional","affiliation":[{"name":"Department of Artificial Intelligence and Human Health, Icahn School of Medicine at Mount Sinai, New York, NY, USA"},{"name":"Department of Neuroscience, Icahn School of Medicine at Mount Sinai, New York, NY, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026062003400580000_2026.06.15.26355696v1.1","unstructured":"Arasu VA GT , Rothstein JH , Alexeeff SE , Achacoso NS , Bhattacharya A , Cord JB , Esserman LJ , Galbraith W , Gerstley LD , Head S , Hylton NM , Kushi LH , Lee C , Leimpeter AD , Lewis DA , Liu V , Marafino BJ , Margolies LR , Navarro DA , Pu A , Sakoda LC , Shan J , Shieh Y , Sistig A , Gueye CS , Veer LV , Villasenor M , Westley M , Wisner DJ , Tice JA , Shen L , Habel LA , Sieh W. . Comparative 10-Year Performance of Mammography Artificial Intelligence, Polygenic, and Clinical Breast Cancer Risk Models in the Kaiser Permanente Research Bank. . JNCI: Journal of the National Cancer Institute. e-Pub 2026."},{"key":"2026062003400580000_2026.06.15.26355696v1.2","doi-asserted-by":"publisher","DOI":"10.1002\/sim.1668"},{"key":"2026062003400580000_2026.06.15.26355696v1.3","doi-asserted-by":"publisher","DOI":"10.1093\/jnci\/81.24.1879"},{"key":"2026062003400580000_2026.06.15.26355696v1.4","doi-asserted-by":"crossref","unstructured":"Yala A , Mikhael PG , Strand F , Lin G , Smith K , Wan YL , et al. Toward robust mammography-based models for breast cancer risk. Sci Transl Med. 2021;13(578).","DOI":"10.1126\/scitranslmed.aba4373"},{"key":"2026062003400580000_2026.06.15.26355696v1.5","doi-asserted-by":"crossref","unstructured":"Meltzer C , Skaane P. Mammography screening. Breast imaging: diagnosis and intervention: Springer; 2022. p. 43\u201368.","DOI":"10.1007\/978-3-030-94918-1_3"},{"key":"2026062003400580000_2026.06.15.26355696v1.6","doi-asserted-by":"publisher","DOI":"10.1007\/s00330-011-2320-2"},{"key":"2026062003400580000_2026.06.15.26355696v1.7","doi-asserted-by":"crossref","unstructured":"Anisuzzaman D , Malins JG , Jackson JI , Lee E , Naser JA , Rostami B , et al. A View-Agnostic Deep Learning Framework for Comprehensive Analysis of 2D-Echocardiography. medRxiv. 2025:2025.07. 10.25331304.","DOI":"10.1101\/2025.07.10.25331304"},{"key":"2026062003400580000_2026.06.15.26355696v1.8","doi-asserted-by":"crossref","unstructured":"Lowry KP , Jeong HE , Kim KH , Hughes KS , Lee CI , Yala A , et al. Current state of mammography-based artificial intelligence for future breast cancer risk prediction: a systematic review. JNCI: Journal of the National Cancer Institute. 2026:djag002.","DOI":"10.1093\/jnci\/djag002"},{"key":"2026062003400580000_2026.06.15.26355696v1.9","doi-asserted-by":"publisher","DOI":"10.1093\/jnci\/djac142"},{"issue":"1","key":"2026062003400580000_2026.06.15.26355696v1.10","doi-asserted-by":"crossref","first-page":"244","DOI":"10.1186\/s13244-024-01808-3","article-title":"Validation of the Mirai model for predicting breast cancer risk in Mexican women","volume":"15","year":"2024","journal-title":"Insights into Imaging"},{"issue":"2","key":"2026062003400580000_2026.06.15.26355696v1.11","doi-asserted-by":"crossref","first-page":"319","DOI":"10.1016\/j.jacr.2023.10.018","article-title":"Artificial intelligence-driven mammography-based future breast cancer risk prediction: a systematic review","volume":"21","year":"2024","journal-title":"Journal of the American College of Radiology"},{"issue":"1","key":"2026062003400580000_2026.06.15.26355696v1.12","doi-asserted-by":"crossref","first-page":"e250391","DOI":"10.1148\/radiol.250391","article-title":"Evaluation of a mammography-based deep learning model for breast cancer risk prediction in a triennial screening program","volume":"317","year":"2025","journal-title":"Radiology"},{"key":"2026062003400580000_2026.06.15.26355696v1.13","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.1611835114"},{"issue":"1","key":"2026062003400580000_2026.06.15.26355696v1.14","doi-asserted-by":"crossref","first-page":"73","DOI":"10.1016\/j.ejrad.2009.05.010","article-title":"Comparison of tissue equalization, and premium view post-processing methods in full field digital mammography","volume":"76","year":"2010","journal-title":"European journal of radiology"},{"key":"2026062003400580000_2026.06.15.26355696v1.15","doi-asserted-by":"publisher","DOI":"10.1259\/bjr\/96554696"},{"key":"2026062003400580000_2026.06.15.26355696v1.16","doi-asserted-by":"crossref","unstructured":"DeLong ER , DeLong DM , Clarke-Pearson DL . Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988:837\u201345.","DOI":"10.2307\/2531595"},{"key":"2026062003400580000_2026.06.15.26355696v1.17","doi-asserted-by":"publisher","DOI":"10.1109\/LSP.2014.2337313"},{"key":"2026062003400580000_2026.06.15.26355696v1.18","doi-asserted-by":"crossref","unstructured":"Wickramasinghe OE, Rajan N , Wanniarachchi C , Karunanayaka K , Wimalaratne P , editors. Evolution of formation control algorithms for unmanned aerial vehicles. 2024 IEEE 33rd International Symposium on Industrial Electronics (ISIE); 2024: IEEE.","DOI":"10.1109\/ISIE54533.2024.10595718"},{"key":"2026062003400580000_2026.06.15.26355696v1.19","doi-asserted-by":"crossref","unstructured":"Su Z , Guo J , Yao K , Yang X , Wang Q , Huang K , editors. Unraveling batch normalization for realistic test-time adaptation. Proceedings of the AAAI conference on artificial intelligence; 2024.","DOI":"10.1609\/aaai.v38i13.29436"},{"key":"2026062003400580000_2026.06.15.26355696v1.20","unstructured":"Lim H , Kim B , Choo J , Choi S. Ttn: A domain-shift aware batch normalization in test-time adaptation. arXiv preprint arXiv:230205155. 2023."}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.06.15.26355696","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,20]],"date-time":"2026-06-20T10:40:19Z","timestamp":1781952019000},"score":1,"resource":{"primary":{"URL":"http:\/\/medrxiv.org\/lookup\/doi\/10.64898\/2026.06.15.26355696"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,17]]},"references-count":20,"URL":"https:\/\/doi.org\/10.64898\/2026.06.15.26355696","relation":{},"subject":[],"published":{"date-parts":[[2026,6,17]]},"subtype":"preprint"}}