{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,15]],"date-time":"2026-06-15T14:32:13Z","timestamp":1781533933803,"version":"3.54.5"},"reference-count":22,"publisher":"MDPI AG","issue":"6","license":[{"start":{"date-parts":[[2026,5,27]],"date-time":"2026-05-27T00:00:00Z","timestamp":1779840000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"NIH","award":["R01NS121154"],"award-info":[{"award-number":["R01NS121154"]}]},{"name":"NIH","award":["U01AG070112"],"award-info":[{"award-number":["U01AG070112"]}]},{"award":["R01NS121154"],"award-info":[{"award-number":["R01NS121154"]}],"id":[{"id":"https:\/\/ror.org\/01cwqze88","id-type":"ROR","asserted-by":"publisher"}]},{"award":["U01AG070112"],"award-info":[{"award-number":["U01AG070112"]}],"id":[{"id":"https:\/\/ror.org\/01cwqze88","id-type":"ROR","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100004837","name":"Ministerio de Ciencia e Innovaci\u00f3n","doi-asserted-by":"publisher","award":["PID2023-146187OB-I00"],"award-info":[{"award-number":["PID2023-146187OB-I00"]}],"id":[{"id":"10.13039\/501100004837","id-type":"DOI","asserted-by":"publisher"}]},{"award":["PID2023-146187OB-I00"],"award-info":[{"award-number":["PID2023-146187OB-I00"]}],"id":[{"id":"https:\/\/ror.org\/05r0vyz12","id-type":"ROR","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100003741","name":"ICREA","doi-asserted-by":"crossref","id":[{"id":"10.13039\/501100003741","id-type":"DOI","asserted-by":"crossref"}]},{"DOI":"10.13039\/100008722","name":"Universitat de Girona","doi-asserted-by":"publisher","award":["IFUdG2022"],"award-info":[{"award-number":["IFUdG2022"]}],"id":[{"id":"10.13039\/100008722","id-type":"DOI","asserted-by":"publisher"}]},{"award":["IFUdG2022"],"award-info":[{"award-number":["IFUdG2022"]}],"id":[{"id":"https:\/\/ror.org\/01xdxns91","id-type":"ROR","asserted-by":"publisher"}]},{"name":"Penumbra, Inc."}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["MAKE"],"abstract":"<jats:p>Machine learning models for medical image analysis are vulnerable to hidden confounders, which can compromise generalization and clinical reliability. Existing detection strategies typically require explicit knowledge or labels of the confounder, which are often unavailable. In this work, we propose an ensemble-based framework to detect potential confounder-driven learning without explicitly defining the confounders, but only which samples might be affected. Our approach leverages the variability of model performance across ensembles to identify signatures of shortcut learning. Shortcut learning occurs when a model uses non-robust features or correlations rather than learning the true underlying task, and it is often observed when confounders are present. We generate controlled dataset variants with increasing confounding levels and analyze distributions of AUC (area under the ROC curve) scores across training, validation, and test splits, revealing converging performance and reduced variance as confounding intensifies. We validate our method on two clinically relevant tasks, diabetic retinopathy detection from retinal fundus images and tumor detection from brain MRI slices. Then, we further demonstrate its practical utility on another dataset and image modality with a stroke reperfusion prediction task with suspected hidden confounders. This work provides a practical, data-driven diagnostic tool to flag potential confounding and support the reliability assessment of machine learning models in medical imaging.<\/jats:p>","DOI":"10.3390\/make8060146","type":"journal-article","created":{"date-parts":[[2026,5,27]],"date-time":"2026-05-27T14:36:56Z","timestamp":1779892616000},"page":"146","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Ensemble Variability as a Signal of Confounding in Medical Imaging Models"],"prefix":"10.3390","volume":"8","author":[{"ORCID":"https:\/\/orcid.org\/0009-0006-8747-2036","authenticated-orcid":false,"given":"Uma M.","family":"Lal-Trehan Estrada","sequence":"first","affiliation":[{"name":"Research Institute of Computer Vision and Robotics, University of Girona, 17003 Girona, Spain"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0602-8509","authenticated-orcid":false,"given":"Sunil A.","family":"Sheth","sequence":"additional","affiliation":[{"name":"McGovern Medical School, University of Texas Health Science Center at Houston, Houston, TX 77030, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0115-0647","authenticated-orcid":false,"given":"Arnau","family":"Oliver","sequence":"additional","affiliation":[{"name":"Research Institute of Computer Vision and Robotics, University of Girona, 17003 Girona, Spain"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2777-3479","authenticated-orcid":false,"given":"Xavier","family":"Llad\u00f3","sequence":"additional","affiliation":[{"name":"Research Institute of Computer Vision and Robotics, University of Girona, 17003 Girona, Spain"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4862-2277","authenticated-orcid":false,"given":"Luca","family":"Giancardo","sequence":"additional","affiliation":[{"name":"Center for Precision Health, McWilliams School of Biomedical Informatics, University of Texas Health Science Center at Houston, Houston, TX 77030, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1968","published-online":{"date-parts":[[2026,5,27]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"29224","DOI":"10.1038\/s41598-024-79838-6","article-title":"The risk of shortcutting in deep learning algorithms for medical imaging research","volume":"14","author":"Hill","year":"2024","journal-title":"Sci. Rep."},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"286","DOI":"10.1038\/s41746-024-01276-5","article-title":"Addressing fairness issues in deep learning-based medical image analysis: A systematic review","volume":"7","author":"Xu","year":"2024","journal-title":"npj Digit. Med."},{"key":"ref_3","doi-asserted-by":"crossref","unstructured":"Puyol-Ant\u00f3n, E., Ruijsink, B., Piechnik, S.K., Neubauer, S., Petersen, S.E., Razavi, R., and King, A.P. (2021). Fairness in cardiac MR image analysis: An investigation of bias due to data imbalance in deep learning based segmentation. MICCAI 2021, Springer.","DOI":"10.1007\/978-3-030-87199-4_39"},{"key":"ref_4","doi-asserted-by":"crossref","first-page":"13","DOI":"10.1167\/tvst.10.2.13","article-title":"Addressing artificial intelligence bias in retinal diagnostics","volume":"10","author":"Burlina","year":"2021","journal-title":"Transl. Vis. Sci. Technol."},{"key":"ref_5","doi-asserted-by":"crossref","unstructured":"Wu, C., Yang, X., Gilkes, E.G., Cui, H., Choi, J., Sun, N., Liao, Z., Fan, B., Santillana, M., and Celi, L. (2023). De-identification and obfuscation of gender attributes from retinal scans. MICCAI Workshop on Clinical Image-Based Procedures, Springer.","DOI":"10.1007\/978-3-031-45249-9_9"},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"102723","DOI":"10.1016\/j.media.2022.102723","article-title":"Equitable modelling of brain imaging by counterfactual augmentation with morphologically constrained 3D deep generative models","volume":"84","author":"Pombo","year":"2023","journal-title":"Med. Image Anal."},{"key":"ref_7","doi-asserted-by":"crossref","unstructured":"Kim, B., Kim, H., Kim, K., Kim, S., and Kim, J. (2019). Learning not to learn: Training deep neural networks with biased data. Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition, IEEE.","DOI":"10.1109\/CVPR.2019.00922"},{"key":"ref_8","doi-asserted-by":"crossref","first-page":"6010","DOI":"10.1038\/s41467-020-19784-9","article-title":"Training confounder-free deep learning models for medical applications","volume":"11","author":"Zhao","year":"2020","journal-title":"Nat. Commun."},{"key":"ref_9","doi-asserted-by":"crossref","unstructured":"Bevan, P.J., and Atapour-Abarghouei, A. (2022). Detecting melanoma fairly: Skin tone detection and debiasing for skin lesion classification. MICCAI Workshop on Domain Adaptation and Representation Transfer, Springer.","DOI":"10.1007\/978-3-031-16852-9_1"},{"key":"ref_10","doi-asserted-by":"crossref","unstructured":"Vento, A., Zhao, Q., Paul, R., Pohl, K.M., and Adeli, E. (2022). A penalty approach for normalizing feature distributions to build confounder-free models. MICCAI 2022, Springer.","DOI":"10.1007\/978-3-031-16437-8_37"},{"key":"ref_11","doi-asserted-by":"crossref","first-page":"103529","DOI":"10.1016\/j.media.2025.103529","article-title":"AI-based association analysis for medical imaging using latent-space geometric confounder correction","volume":"102","author":"Liu","year":"2025","journal-title":"Med. Image Anal."},{"key":"ref_12","unstructured":"Pleiss, G., Raghavan, M., Wu, F., Kleinberg, J., and Weinberger, K.Q. (2017). On fairness and calibration. NeurIPS, Curran Associates Inc."},{"key":"ref_13","unstructured":"Komeyer, V., Eickhoff, S.B., Grefkes, C., Patil, K.R., and Raimondo, F. (2024). A framework for confounder considerations in AI-driven precision medicine. medRxiv."},{"key":"ref_14","unstructured":"Santa Cruz, B.G., Husch, A., and Hertel, F. (2022, January 10\u201312). The effect of dataset confounding on predictions of deep neural networks for medical imaging. Proceedings of the Northern Lights Deep Learning Workshop, Troms\u00f8, Norway."},{"key":"ref_15","doi-asserted-by":"crossref","unstructured":"Stanley, E.A.M., Wilms, M., and Forkert, N.D. (2023). A flexible framework for simulating and evaluating biases in deep learning-based medical image analysis. MICCAI 2023, Springer.","DOI":"10.1007\/978-3-031-43895-0_46"},{"key":"ref_16","doi-asserted-by":"crossref","first-page":"2613","DOI":"10.1093\/jamia\/ocae165","article-title":"Towards objective and systematic evaluation of bias in artificial intelligence for medical imaging","volume":"31","author":"Stanley","year":"2024","journal-title":"J. Am. Med. Inform. Assoc."},{"key":"ref_17","doi-asserted-by":"crossref","first-page":"105501","DOI":"10.1016\/j.ebiom.2024.105501","article-title":"Where, why, and how is bias learned in medical image analysis models? A study of bias encoding within convolutional networks using synthetic data","volume":"111","author":"Stanley","year":"2025","journal-title":"eBioMedicine"},{"key":"ref_18","unstructured":"Lakshminarayanan, B., Pritzel, A., and Blundell, C. (2017). Simple and scalable predictive uncertainty estimation using deep ensembles. Advances in Neural Information Processing Systems, Curran Associates Inc."},{"key":"ref_19","unstructured":"Ovadia, Y., Fertig, E., Ren, J., Nado, Z., Sculley, D., Nowozin, S., Dillon, J., Lakshminarayanan, B., and Snoek, J. (2019). Can you trust your model\u2019s uncertainty? Evaluating predictive uncertainty under dataset shift. Advances in Neural Information Processing Systems, Curran Associates Inc."},{"key":"ref_20","doi-asserted-by":"crossref","first-page":"231","DOI":"10.5566\/ias.1155","article-title":"Feedback on a publicly distributed image database: The Messidor database","volume":"33","author":"Zhang","year":"2014","journal-title":"Image Anal. Stereol."},{"key":"ref_21","unstructured":"Adewole, M., Rudie, J.D., Gbdamosi, A., Toyobo, O., Raymond, C., Zhang, D., Omidiji, O., Akinola, R., Suwaid, M.A., and Emegoakor, A. (2023). The brain tumor segmentation (BRATS) challenge 2023: Glioma segmentation in sub-Saharan Africa patient population (BRATS-Africa). arXiv."},{"key":"ref_22","doi-asserted-by":"crossref","first-page":"1790","DOI":"10.1161\/STROKEAHA.119.028891","article-title":"2b, 2c, or 3: What should be the angiographic target for endovascular treatment in ischemic stroke?","volume":"51","author":"LeCouffe","year":"2020","journal-title":"Stroke"}],"container-title":["Machine Learning and Knowledge Extraction"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/2504-4990\/8\/6\/146\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,3]],"date-time":"2026-06-03T04:11:10Z","timestamp":1780459870000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/2504-4990\/8\/6\/146"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,27]]},"references-count":22,"journal-issue":{"issue":"6","published-online":{"date-parts":[[2026,6]]}},"alternative-id":["make8060146"],"URL":"https:\/\/doi.org\/10.3390\/make8060146","relation":{},"ISSN":["2504-4990"],"issn-type":[{"value":"2504-4990","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,27]]}}}