{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,10]],"date-time":"2026-03-10T00:30:15Z","timestamp":1773102615045,"version":"3.50.1"},"reference-count":26,"publisher":"Wiley","issue":"2","license":[{"start":{"date-parts":[[2026,2,20]],"date-time":"2026-02-20T00:00:00Z","timestamp":1771545600000},"content-version":"vor","delay-in-days":19,"URL":"http:\/\/onlinelibrary.wiley.com\/termsAndConditions#vor"},{"start":{"date-parts":[[2026,2,1]],"date-time":"2026-02-01T00:00:00Z","timestamp":1769904000000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/doi.wiley.com\/10.1002\/tdm_license_1.1"}],"content-domain":{"domain":["aapm.onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Medical Physics"],"published-print":{"date-parts":[[2026,2]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Despite the high precision of deep learning models on internal tests on CT, their effectiveness often drops on external validation due to artifacts caused by patient motion and implants like metal or silicone that were not accounted for in the carefully curated training data. Potential wide categories of \u201cunknown\u201d anomalies within a CT exam makes training a supervised model for out\u2010of\u2010distribution (OOD) identification impractical, especially when considering unseen external data.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Purpose<\/jats:title>\n                    <jats:p>To develop an artificial intelligence (AI) model to detect and identify anomalies\/OOD data in abdominal\u2013pelvis CT exams for the purpose of improving the performance of downstream applications.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>Our proposed 2D and 3D generative architecture receives the third lumbar vertebra (L3) slice (slice\u2010level model) or all the slices from the series (series\u2010level model), generates a reconstruction and the secondary part of our architecture\u2014anomaly score computation block, computes the anomalies pixels\/voxels (slice\u2010level\/series\u2010level) to identify anomalous L3\u2010slices\/volumes (slice\u2010level\/series\u2010level). We trained on data from over 2850 abdominal\u2013pelvis CT volumes from adults over age 50 years collected throughout multiple Mayo Clinic campuses (60% female; mean age: 66.9, 92.4% non\u2010Hispanic White) and tested on a prospective test set of 544 CTs from July 2024 (47.3% female; mean age: 70.9, 94% non\u2010Hispanic White) as well as an external test set.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>We found that while traditional methods show moderate success, our generative models\u2014Vector Quantized Variational Autoencoder (VQVAE) and Vision Transformer\u2010Masked Autoencoder (VIT\u2010MAE)\u2014deliver excellent results with negligible false positives (FPs) and are also superior in identifying varied types of OOD samples. Prospective analysis showed the model was able to handle the under\u2010documentation of anomaly in radiology reports with 86.11% true positive (TP) rate. We also performed external validation using the publicly available AbdominalCT\u20101k dataset, which contains 1062 CT scans compiled from several existing benchmark datasets. The model achieved a 75.26% TP rate, while the 24.7% FP rate was primarily triggered by anomalies located outside the body.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>\n                      The proposed method can be leveraged to detect both intra\u2010 and interclass OOD data from abdominal CT images and can assess the quality of CT datasets to provide actionable insights. This workflow is particularly valuable for nonshareable healthcare collaborations, where it can be deployed as a service within local firewalls for automated dataset curation without prior knowledge about the OOD types. The implementation of the algorithm is available in the GitHub:\n                      <jats:ext-link xmlns:xlink=\"http:\/\/www.w3.org\/1999\/xlink\" xlink:href=\"https:\/\/github.com\/gokul-ramasamy\/implant_detection.git\">https:\/\/github.com\/gokul\u2010ramasamy\/implant_detection.git<\/jats:ext-link>\n                      .\n                    <\/jats:p>\n                  <\/jats:sec>","DOI":"10.1002\/mp.70339","type":"journal-article","created":{"date-parts":[[2026,2,20]],"date-time":"2026-02-20T16:15:46Z","timestamp":1771604146000},"update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Self\u2010supervised out\u2010of\u2010distribution detection\u2014Metal implants and other anomaly"],"prefix":"10.1002","volume":"53","author":[{"given":"Gokul","family":"Ramasamy","sequence":"first","affiliation":[{"name":"AI &amp; Informatics Mayo Clinic Phoenix Arizona USA"},{"name":"School of Computing and Augmented Intelligence Arizona State University Tempe Arizona USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Amara","family":"Tariq","sequence":"additional","affiliation":[{"name":"AI &amp; Informatics Mayo Clinic Phoenix Arizona USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Samuel J.","family":"Fahrenholtz","sequence":"additional","affiliation":[{"name":"Division of Medical Physics Mayo Clinic Phoenix Arizona USA"},{"name":"Department of Radiology Mayo Clinic Phoenix Arizona USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"William F.","family":"Sensakovic","sequence":"additional","affiliation":[{"name":"Division of Medical Physics Mayo Clinic Phoenix Arizona USA"},{"name":"Department of Radiology Mayo Clinic Phoenix Arizona USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Bhavik N.","family":"Patel","sequence":"additional","affiliation":[{"name":"School of Computing and Augmented Intelligence Arizona State University Tempe Arizona USA"},{"name":"Department of Radiology Mayo Clinic Phoenix Arizona USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Imon","family":"Banerjee","sequence":"additional","affiliation":[{"name":"AI &amp; Informatics Mayo Clinic Phoenix Arizona USA"},{"name":"School of Computing and Augmented Intelligence Arizona State University Tempe Arizona USA"},{"name":"Department of Radiology Mayo Clinic Phoenix Arizona USA"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"311","published-online":{"date-parts":[[2026,2,20]]},"reference":[{"key":"e_1_2_8_2_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.crad.2021.01.010"},{"key":"e_1_2_8_3_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.cmpb.2016.10.007"},{"key":"e_1_2_8_4_1","doi-asserted-by":"publisher","DOI":"10.1016\/S0140-6736(18)31645-3"},{"key":"e_1_2_8_5_1","doi-asserted-by":"publisher","DOI":"10.1148\/ryai.210064"},{"key":"e_1_2_8_6_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.16230"},{"key":"e_1_2_8_7_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.jacr.2024.10.003"},{"key":"e_1_2_8_8_1","unstructured":"KanavatiF IslamS AboagyeEO RockallAG.Automatic L3 slice detection in 3D CT images using fully\u2010convolutional networks.arXiv preprint arXiv:1811.09244.2018."},{"key":"e_1_2_8_9_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.jot.2019.11.004"},{"key":"e_1_2_8_10_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.csbj.2020.04.005"},{"key":"e_1_2_8_11_1","doi-asserted-by":"crossref","unstructured":"DengJ DongW SocherR LiLJ LiK Fei\u2010FeiL.ImageNet: a large\u2010scale hierarchical image database. In:2009 IEEE Conference on Computer Vision and Pattern Recognition.IEEE2009:248\u2010255.","DOI":"10.1109\/CVPR.2009.5206848"},{"key":"e_1_2_8_12_1","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-024-02117-4"},{"key":"e_1_2_8_13_1","unstructured":"Van DenOordA VinyalsO KavukcuogluK.Neural discrete representation learning. In:Advances in Neural Information Processing Systems. Vol30.2017."},{"key":"e_1_2_8_14_1","unstructured":"AlexeyD BeyerL KolesnikovA et\u00a0al.An image is worth 16x16 words: transformers for image recognition at scale.ICLR.2021."},{"key":"e_1_2_8_15_1","doi-asserted-by":"crossref","unstructured":"BaiY WangZ XiaoJ et\u00a0al.Masked autoencoders enable efficient knowledge distillers. In:Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition.2023:24256\u201024265.","DOI":"10.1109\/CVPR52729.2023.02323"},{"key":"e_1_2_8_16_1","unstructured":"LucasJ TuckerG GrosseRB NorouziM.Don't blame the ELBO! A linear VAE perspective on posterior collapse. In:Advances in Neural Information Processing Systems. Vol.32.2019."},{"key":"e_1_2_8_17_1","doi-asserted-by":"crossref","unstructured":"HeK ChenX XieS LiY Doll\u00e1rP GirshickR.Masked autoencoders are scalable vision learners.2022 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR) New Orleans LA USA 2021 pp.15979\u201015988.https:\/\/doi.org\/10.1109\/CVPR52688.2022.01553","DOI":"10.1109\/CVPR52688.2022.01553"},{"key":"e_1_2_8_18_1","doi-asserted-by":"publisher","DOI":"10.1117\/1.JMI.11.6.064003"},{"key":"e_1_2_8_19_1","unstructured":"van denOordA VinyalsO KavukcuogluK.Neural discrete representation learning. InProceedings of the 31st International Conference on Neural Information Processing Systems (NIPS'17). Curran Associates Inc. Red Hook NY USA 6309\u20106318."},{"key":"e_1_2_8_20_1","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2021.3100536"},{"key":"e_1_2_8_21_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2022.102680"},{"key":"e_1_2_8_22_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2020.101821"},{"key":"e_1_2_8_23_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-022-30695-9"},{"key":"e_1_2_8_24_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.ejrad.2018.08.010"},{"key":"e_1_2_8_25_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.jacr.2025.03.001"},{"key":"e_1_2_8_26_1","doi-asserted-by":"publisher","DOI":"10.1007\/s00261\u2010023\u201003800\u20109"},{"key":"e_1_2_8_27_1","doi-asserted-by":"publisher","DOI":"10.1007\/s11914\u2010022\u201000764\u20105"}],"container-title":["Medical Physics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/mp.70339","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/full-xml\/10.1002\/mp.70339","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/mp.70339","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,9]],"date-time":"2026-03-09T13:35:07Z","timestamp":1773063307000},"score":1,"resource":{"primary":{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/10.1002\/mp.70339"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2]]},"references-count":26,"journal-issue":{"issue":"2","published-print":{"date-parts":[[2026,2]]}},"alternative-id":["10.1002\/mp.70339"],"URL":"https:\/\/doi.org\/10.1002\/mp.70339","archive":["Portico"],"relation":{},"ISSN":["0094-2405","2473-4209"],"issn-type":[{"value":"0094-2405","type":"print"},{"value":"2473-4209","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2]]},"assertion":[{"value":"2024-11-26","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-01-16","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-02-20","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e70339"}}