{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,30]],"date-time":"2026-04-30T16:12:55Z","timestamp":1777565575913,"version":"3.51.4"},"reference-count":31,"publisher":"Wiley","issue":"7","license":[{"start":{"date-parts":[[2025,4,24]],"date-time":"2025-04-24T00:00:00Z","timestamp":1745452800000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/onlinelibrary.wiley.com\/termsAndConditions#vor"}],"content-domain":{"domain":["aapm.onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Medical Physics"],"published-print":{"date-parts":[[2025,7]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>In CT\u2010based medical image segmentation, the choice of loss function profoundly impacts the training efficacy of deep neural networks. Traditional loss functions like cross entropy (CE), Dice, Boundary, and TopK each have unique strengths and limitations, often introducing biases when used individually.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Purpose<\/jats:title>\n                    <jats:p>This study aims to enhance segmentation accuracy by optimizing ensemble loss functions, thereby addressing the biases and limitations of single loss functions and their linear\u00a0combinations.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>We implemented a comprehensive evaluation of loss function combinations by integrating CE, Dice, Boundary, and TopK loss functions through both loss\u2010level linear combination and model\u2010level ensemble methods. Our approach utilized two state\u2010of\u2010the\u2010art 3D segmentation architectures, Attention U\u2010Net (AttUNet) and SwinUNETR, to test the impact of these methods. The study was conducted on two large CT dataset cohorts: an institutional dataset containing pelvic organ segmentations, and a public dataset consisting of multiple organ segmentations. All the models were trained from scratch with different loss settings, and performance was evaluated using Dice similarity coefficient (DSC), Hausdorff distance (HD), and average surface distance (ASD). In the ensemble approach, both static averaging and learnable dynamic weighting strategies were employed to combine the outputs of models trained with different loss\u00a0functions.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>Extensive experiments revealed the following: (1) the linear combination of loss functions achieved results comparable to those of single loss\u2010driven methods; (2) compared to the best non\u2010ensemble methods, ensemble\u2010based approaches resulted in a 2%\u20137% increase in DSC scores, along with notable reductions in HD (e.g., a 19.1% reduction for rectum segmentation using SwinUNETR) and ASD (e.g., a 49.0% reduction for prostate segmentation using AttUNet); (3) the learnable ensemble approach with optimized weights produced finer details in predicted masks, as confirmed by qualitative analyses; and (4) the learnable ensemble consistently outperforms the static ensemble across most metrics (DSC, HD, ASD) for both AttUNet and SwinUNETR architectures.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>Our findings support the efficacy of using ensemble models with optimized weights to improve segmentation accuracy, highlighting the potential for broader applications in automated medical image\u00a0analysis.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1002\/mp.17848","type":"journal-article","created":{"date-parts":[[2025,4,25]],"date-time":"2025-04-25T00:36:19Z","timestamp":1745541379000},"update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":3,"title":["Enhancing CT image segmentation accuracy through ensemble loss function optimization"],"prefix":"10.1002","volume":"52","author":[{"given":"Chengyin","family":"Li","sequence":"first","affiliation":[{"name":"Department of Computer Science Wayne State University Detroit Michigan USA"},{"name":"Department of Radiation Oncology Henry Ford Health Detroit Michigan USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Rafi Ibn","family":"Sultan","sequence":"additional","affiliation":[{"name":"Department of Computer Science Wayne State University Detroit Michigan USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hassan","family":"Bagher\u2010Ebadian","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Henry Ford Health Detroit Michigan USA"},{"name":"Department of Radiology Michigan State University E. Lansing Michigan USA"},{"name":"Department of Osteopathic Michigan State University E. Lansing Michigan USA"},{"name":"Department of Physics Oakland University Rochester Michigan USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yao","family":"Qiang","sequence":"additional","affiliation":[{"name":"Department of Computer Science Wayne State University Detroit Michigan USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Kundan","family":"Thind","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Henry Ford Health Detroit Michigan USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Dongxiao","family":"Zhu","sequence":"additional","affiliation":[{"name":"Department of Computer Science Wayne State University Detroit Michigan USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Indrin J.","family":"Chetty","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Cedars Sinai Medical Center Los Angeles California USA"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"311","published-online":{"date-parts":[[2025,4,24]]},"reference":[{"key":"e_1_2_9_2_1","first-page":"623","volume-title":"Abdomen and Thoracic Imaging: An Engineering & Clinical Perspective","author":"Acosta O","year":"2013"},{"key":"e_1_2_9_3_1","doi-asserted-by":"publisher","DOI":"10.1002\/jmrs.64"},{"key":"e_1_2_9_4_1","doi-asserted-by":"publisher","DOI":"10.1109\/TMI.2016.2519264"},{"key":"e_1_2_9_5_1","doi-asserted-by":"publisher","DOI":"10.1088\/0031-9155\/61\/17\/6553"},{"key":"e_1_2_9_6_1","doi-asserted-by":"publisher","DOI":"10.1111\/1754-9485.13286"},{"key":"e_1_2_9_7_1","doi-asserted-by":"crossref","unstructured":"LiC QiangY SultanRI et\u00a0al.FocalUNETR: a focal transformer for boundary\u2010aware prostate segmentation using CT images. In:International Conference on Medical Image Computing and Computer\u2010Assisted Intervention.Springer;2023:592\u2010602.","DOI":"10.1007\/978-3-031-43898-1_57"},{"key":"e_1_2_9_8_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.16750"},{"key":"e_1_2_9_9_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.cmpb.2022.106874"},{"key":"e_1_2_9_10_1","doi-asserted-by":"crossref","unstructured":"RonnebergerO FischerP BroxT.U\u2010net: Convolutional networks for biomedical image segmentation. In:Medical Image Computing and Computer\u2010Assisted Intervention\u2013MICCAI 2015: 18th International Conference Munich Germany October 5\u20109 2015 Proceedings Part III 18.Springer;2015:234\u2010241.","DOI":"10.1007\/978-3-319-24574-4_28"},{"key":"e_1_2_9_11_1","unstructured":"OktayO et\u00a0al.Attention U\u2010Net: Learning Where to Look for the Pancreas. In:Medical Imaging with Deep Learning.2018."},{"key":"e_1_2_9_12_1","doi-asserted-by":"crossref","unstructured":"TangY YangD LiW et\u00a0al.Self\u2010supervised pre\u2010training of Swin transformers for 3D medical image analysis. In:Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition.IEEE;2022:20730\u201020740.","DOI":"10.1109\/CVPR52688.2022.02007"},{"key":"e_1_2_9_13_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.neunet.2017.06.003"},{"key":"e_1_2_9_14_1","doi-asserted-by":"publisher","DOI":"10.1088\/1361-6560\/aaf11c"},{"key":"e_1_2_9_15_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-020-01008-z"},{"key":"e_1_2_9_16_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46976-8_19"},{"key":"e_1_2_9_17_1","doi-asserted-by":"crossref","unstructured":"MilletariF NavabN AhmadiS.V\u2010net: fully convolutional neural netw for volumetric medical image segmentation. In:2016 Fourth International Conference on 3D Vision (3DV).IEEE;2016:565\u2010571.","DOI":"10.1109\/3DV.2016.79"},{"key":"e_1_2_9_18_1","unstructured":"SudreCH LiW VercauterenT OurselinS CardosoMJ.Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations. In:Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: Third International Workshop DLMIA 2017 and 7th International Workshop ML\u2010CDS 2017 Held in Conjunction with MICCAI 2017 Qu\u00e9bec City QC Canada September 14 Proceedings 3.Springer;2017:240\u2010248."},{"key":"e_1_2_9_19_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-020-19449-7"},{"key":"e_1_2_9_20_1","unstructured":"KervadecH BouchtibaJ DesrosiersC GrangerE DolzJ AyedIB.Boundary loss for highly unbalanced segmentation. In:International Conference on Medical Imaging With Deep Learning.PMLR;2019:285\u2010296."},{"key":"e_1_2_9_21_1","unstructured":"WuZ ShenC HengelAvd.Bridging category\u2010level and instance\u2010level semantic image segmentation.arXiv preprint arXiv:1605.06885(2016)."},{"key":"e_1_2_9_22_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2021.102035"},{"key":"e_1_2_9_23_1","doi-asserted-by":"crossref","unstructured":"KendallA GalY CipollaR Multi\u2010task learning using uncertainty to weigh losses for scene geometry and semantics. In:Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition IEEE;2018:7482\u20107491.","DOI":"10.1109\/CVPR.2018.00781"},{"key":"e_1_2_9_24_1","doi-asserted-by":"publisher","DOI":"10.1007\/s11704-019-8208-z"},{"key":"e_1_2_9_25_1","doi-asserted-by":"publisher","DOI":"10.1002\/widm.1249"},{"key":"e_1_2_9_26_1","doi-asserted-by":"publisher","DOI":"10.1049\/ipr2.12419"},{"key":"e_1_2_9_27_1","doi-asserted-by":"crossref","unstructured":"LiuZ LinY CaoY et\u00a0al.Swin transformer: Hierarchical vision transformer using shifted windows. In:Proceedings of the IEEE\/CVF International Conference on Computer Vision.IEEE;2021:10012\u201010022.","DOI":"10.1109\/ICCV48922.2021.00986"},{"key":"e_1_2_9_28_1","doi-asserted-by":"publisher","DOI":"10.1214\/aop\/1176996454"},{"key":"e_1_2_9_29_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijrobp.2019.03.017"},{"key":"e_1_2_9_30_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41597-020-00715-8"},{"key":"e_1_2_9_31_1","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0261307"},{"key":"e_1_2_9_32_1","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2022.3182399"}],"container-title":["Medical Physics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/mp.17848","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,12,8]],"date-time":"2025-12-08T10:36:25Z","timestamp":1765190185000},"score":1,"resource":{"primary":{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/10.1002\/mp.17848"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,4,24]]},"references-count":31,"journal-issue":{"issue":"7","published-print":{"date-parts":[[2025,7]]}},"alternative-id":["10.1002\/mp.17848"],"URL":"https:\/\/doi.org\/10.1002\/mp.17848","archive":["Portico"],"relation":{},"ISSN":["0094-2405","2473-4209"],"issn-type":[{"value":"0094-2405","type":"print"},{"value":"2473-4209","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,4,24]]},"assertion":[{"value":"2024-08-17","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-04-04","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-04-24","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e17848"}}