{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,26]],"date-time":"2026-09-26T03:11:41Z","timestamp":1790392301865,"version":"4.1.0"},"reference-count":25,"publisher":"Wiley","issue":"10","license":[{"start":{"date-parts":[[2024,7,12]],"date-time":"2024-07-12T00:00:00Z","timestamp":1720742400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/onlinelibrary.wiley.com\/termsAndConditions#vor"}],"funder":[{"DOI":"10.13039\/501100002910","name":"F\u00e9d\u00e9ration Wallonie-Bruxelles","doi-asserted-by":"publisher","award":["7289"],"award-info":[{"award-number":["7289"]}],"id":[{"id":"10.13039\/501100002910","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["aapm.onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Medical Physics"],"published-print":{"date-parts":[[2024,10]]},"abstract":"<jats:title>Abstract<\/jats:title><jats:sec><jats:title>Background<\/jats:title><jats:p>The reliable and efficient estimation of uncertainty in artificial intelligence (AI) models poses an ongoing challenge in many fields such as radiation therapy. AI models are intended to automate manual steps involved in the treatment planning workflow. We focus in this study on dose prediction models that predict an optimal dose trade\u2010off for each new patient for a specific treatment modality. They can guide physicians in the optimization, be part of automatic treatment plan generation or support decision in treatment\u00a0indication.<\/jats:p><jats:p>Most common uncertainty estimation methods are based on Bayesian approximations, like Monte Carlo dropout (MCDO) or Deep ensembling (DE). These two techniques, however, have a high inference time (i.e., require multiple inference passes) and might not work for detecting out\u2010of\u2010distribution (OOD) data (i.e., overlapping uncertainty estimate for in\u2010distribution (ID) and OOD).<\/jats:p><\/jats:sec><jats:sec><jats:title>Purpose<\/jats:title><jats:p>In this study, we present a direct uncertainty estimation method and apply it for a dose prediction U\u2010Net architecture. It can be used to flag OOD data and give information on the quality of the dose\u00a0prediction.<\/jats:p><\/jats:sec><jats:sec><jats:title>Methods<\/jats:title><jats:p>Our method consists in the addition of a branch decoding from the bottleneck which reconstructs the CT scan given as input. The input reconstruction error can be used as a surrogate of the model uncertainty. For the proof\u2010of\u2010concept, our method is applied to proton therapy dose prediction in head and neck cancer patients. A dataset of 60 oropharyngeal patients was used to train the network using a nested cross\u2010validation approach with 11 folds (training: 50 patients, validation: 5 patients, test: 5 patients). For the OOD experiment, we used 10 extra patients with a different head and neck sub\u2010location. Accuracy, time\u2010gain, and OOD detection are analyzed for our method in this particular application and compared with the popular MCDO and\u00a0DE.<\/jats:p><\/jats:sec><jats:sec><jats:title>Results<\/jats:title><jats:p>The additional branch did not reduce the accuracy of the dose prediction model. The median absolute error is close to zero for the target volumes and less than 1% of the dose prescription for organs at risk. Our input reconstruction method showed a higher Pearson correlation coefficient with the prediction error (0.620) than DE (0.447) and MCDO (between 0.599 and 0.612). Moreover, our method allows an easier identification of OOD (no overlap for ID and OOD data and a Z\u2010score of 34.05). The uncertainty is estimated simultaneously to the regression task, therefore requires less time and computational\u00a0resources.<\/jats:p><\/jats:sec><jats:sec><jats:title>Conclusions<\/jats:title><jats:p>This study shows that the error in the CT scan reconstruction can be used as a surrogate of the uncertainty of the model. The Pearson correlation coefficient with the dose prediction error is slightly higher than state\u2010of\u2010the\u2010art techniques. OOD data can be more easily detected and the uncertainty metric is computed simultaneously to the regression task, therefore faster than MCDO or DE. The code and pretrained model are available on the gitlab repository: <jats:ext-link xmlns:xlink=\"http:\/\/www.w3.org\/1999\/xlink\" xlink:href=\"https:\/\/gitlab.com\/ai4miro\/ct-reconstruction-for-uncertainty-quatification-of-hdunet\">https:\/\/gitlab.com\/ai4miro\/ct\u2010reconstruction\u2010for\u2010uncertainty\u2010quatification\u2010of\u2010hdunet<\/jats:ext-link><\/jats:p><\/jats:sec>","DOI":"10.1002\/mp.17287","type":"journal-article","created":{"date-parts":[[2024,7,12]],"date-time":"2024-07-12T14:03:27Z","timestamp":1720793007000},"page":"7369-7377","update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":5,"title":["Can input reconstruction be used to directly estimate uncertainty of a dose prediction U\u2010Net model?"],"prefix":"10.1002","volume":"51","author":[{"given":"Margerie","family":"Huet\u2010Dastarac","sequence":"first","affiliation":[{"name":"Molecular Imaging, Radiation and Oncology (MIRO) Laboratory Institut de Recherche Exp\u00e9rimentale et Clinique (IREC) UCLouvain Belgium"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Dan","family":"Nguyen","sequence":"additional","affiliation":[{"name":"Medical Artificial Intelligence and Automation (MAIA) Laboratory Department of Radiation Oncology UT Southwestern Medical Center Dallas USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Eleonore","family":"Longton","sequence":"additional","affiliation":[{"name":"Department of Radiotherapy Cliniques Universitaires Saint\u2010Luc Brussels Belgium"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Steve","family":"Jiang","sequence":"additional","affiliation":[{"name":"Medical Artificial Intelligence and Automation (MAIA) Laboratory Department of Radiation Oncology UT Southwestern Medical Center Dallas USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"John","family":"Lee","sequence":"additional","affiliation":[{"name":"Molecular Imaging, Radiation and Oncology (MIRO) Laboratory Institut de Recherche Exp\u00e9rimentale et Clinique (IREC) UCLouvain Belgium"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ana Barrag\u00e1n","family":"Montero","sequence":"additional","affiliation":[{"name":"Molecular Imaging, Radiation and Oncology (MIRO) Laboratory Institut de Recherche Exp\u00e9rimentale et Clinique (IREC) UCLouvain Belgium"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"311","published-online":{"date-parts":[[2024,7,12]]},"reference":[{"key":"e_1_2_8_2_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.semradonc.2022.06.010"},{"key":"e_1_2_8_3_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.ctro.2023.100616"},{"key":"e_1_2_8_4_1","doi-asserted-by":"publisher","DOI":"10.1002\/acm2.14065"},{"key":"e_1_2_8_5_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.16431"},{"key":"e_1_2_8_6_1","doi-asserted-by":"publisher","DOI":"10.1088\/1361\u20106560\/ac678a"},{"key":"e_1_2_8_7_1","volume-title":"Advances in Neural Information Processing Systems","author":"Lakshminarayanan B","year":"2017"},{"key":"e_1_2_8_8_1","first-page":"1050","article-title":"Dropout as a Bayesian approximation: representing model uncertainty in deep learning","volume":"48","author":"Gal Y","year":"2016","journal-title":"Proc Mach Learn Res"},{"key":"e_1_2_8_9_1","article-title":"The peril of popular deep learning uncertainty estimation methods","author":"Liu Y","year":"2021","journal-title":"NeurIPS"},{"key":"e_1_2_8_10_1","article-title":"DEUP: Direct epistemic uncertainty prediction","author":"Lahlou S","year":"2023","journal-title":"Transactions on Machine Learning Research"},{"key":"e_1_2_8_11_1","doi-asserted-by":"crossref","unstructured":"KushibarK CampelloV GarruchoL LinardosA RadevaP LekadirK.Layer ensembles: a single\u2010pass uncertainty estimation in deep learning for segmentation. In:MICCAI.Springer;2022:514\u2010524.","DOI":"10.1007\/978-3-031-16452-1_49"},{"key":"e_1_2_8_12_1","doi-asserted-by":"crossref","unstructured":"TilmanR Huet\u2010DastaracM Barrag\u00e1n\u2010MonteroAM LeeJA.Single\u2010pass uncertainty estimation with layer ensembling for regression: application to proton therapy dose prediction for head and neck cancer. In:31st European Symposium on Artificial Neural Networks Computational Intelligence and Machine Learning {ESANN} 2023 Bruges Belgium October 4\u20136 2023.2023.https:\/\/doi.org\/10.14428\/esann\/2023.ES2023\u2010115","DOI":"10.14428\/esann\/2023.ES2023-115"},{"key":"e_1_2_8_13_1","first-page":"457","article-title":"Transformer\u2010based out\u2010of\u2010distribution detection for clinically safe segmentation","volume":"172","author":"Graham M","year":"2022","journal-title":"Proc Mach Learn Res"},{"key":"e_1_2_8_14_1","doi-asserted-by":"publisher","DOI":"10.1109\/TAI.2022.3159510"},{"key":"e_1_2_8_15_1","doi-asserted-by":"publisher","DOI":"10.1088\/1361-6560\/ab039b"},{"key":"e_1_2_8_16_1","doi-asserted-by":"crossref","unstructured":"RonnebergerO FischerP BroxT.U\u2010Net: convolutional networks for biomedical image segmentation. In:MICCAI.Springer;2015:234\u2010241.","DOI":"10.1007\/978-3-319-24574-4_28"},{"key":"e_1_2_8_17_1","doi-asserted-by":"crossref","unstructured":"HuangG LiuZ Van Der MaatenL WeinbergerKQ.Densely connected convolutional networks. In:CVPR.IEEE;2017.","DOI":"10.1109\/CVPR.2017.243"},{"key":"e_1_2_8_18_1","doi-asserted-by":"crossref","unstructured":"LiuR BaiJ ZhouJ ZhangK NiC.A feasibility study for predicting 3D radiotherapy dose distribution of lung VMAT Patients. In:ICTAI.IEEE;2020.","DOI":"10.1109\/ICTAI50040.2020.00197"},{"key":"e_1_2_8_19_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.13597"},{"key":"e_1_2_8_20_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.13749"},{"key":"e_1_2_8_21_1","doi-asserted-by":"crossref","unstructured":"HeK ZhangX RenS SunJ.Delving deep into rectifiers: surpassing human\u2010level performance on ImageNet classification. In:ICCV.IEEE;2015.","DOI":"10.1109\/ICCV.2015.123"},{"key":"e_1_2_8_22_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4020-5614-7_3826"},{"key":"e_1_2_8_23_1","doi-asserted-by":"publisher","DOI":"10.1002\/acm2.14095"},{"key":"e_1_2_8_24_1","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v36i7.20757"},{"key":"e_1_2_8_25_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2021.102062"},{"key":"e_1_2_8_26_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.inffus.2021.05.008"}],"container-title":["Medical Physics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/mp.17287","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,8,22]],"date-time":"2025-08-22T16:33:22Z","timestamp":1755880402000},"score":1,"resource":{"primary":{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/10.1002\/mp.17287"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,7,12]]},"references-count":25,"journal-issue":{"issue":"10","published-print":{"date-parts":[[2024,10]]}},"alternative-id":["10.1002\/mp.17287"],"URL":"https:\/\/doi.org\/10.1002\/mp.17287","archive":["Portico"],"relation":{},"ISSN":["0094-2405","2473-4209"],"issn-type":[{"value":"0094-2405","type":"print"},{"value":"2473-4209","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,7,12]]},"assertion":[{"value":"2023-11-22","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2024-06-23","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2024-07-12","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}