{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,11]],"date-time":"2026-06-11T11:00:55Z","timestamp":1781175655533,"version":"3.54.1"},"reference-count":53,"publisher":"Wiley","license":[{"start":{"date-parts":[[2026,6,11]],"date-time":"2026-06-11T00:00:00Z","timestamp":1781136000000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2026,6,11]],"date-time":"2026-06-11T00:00:00Z","timestamp":1781136000000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/doi.wiley.com\/10.1002\/tdm_license_1.1"}],"content-domain":{"domain":["onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Computer Graphics Forum"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Deep neural networks are widely used for image segmentation, also in sensitive applications such as medical imaging or autonomous driving. However, few explainable AI methods are available that help developers understand such networks beyond classification. We demonstrate that studying how segmentation networks behave in scale space, i.e., when increasingly blurring their input, conveys relevant insight into whether detections require features such as sharp edges or high\u2010frequency textures, and what priors have been learned implicitly. We introduce a visual analytics framework that supports a systematic qualitative and quantitative investigation of how segmentations evolve across scale space. In particular, summary visualizations for individual images support the formation of hypotheses that are subsequently formalized via customizable feature representations and evaluated on larger sets of test images using interactive embeddings and quantitative plots. On a cardiac MRI dataset, we demonstrate that this framework reveals important differences between a convolutional and a transformer\u2010based neural network that, at first glance, appear to produce similar results. We confirm the practical relevance of those insights on out\u2010of\u2010distribution images from an MR scanner from which no images were included in the training data. Beyond this primary analysis, we provide extended results on the Adverse Conditions Dataset with Correspondences (ACDC) for semantic driving scene understanding.<\/jats:p>","DOI":"10.1111\/cgf.70470","type":"journal-article","created":{"date-parts":[[2026,6,11]],"date-time":"2026-06-11T10:42:56Z","timestamp":1781174576000},"update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Visualizing Image Segmentation Network Behavior Through the Lens of Scale Space Analysis"],"prefix":"10.1111","author":[{"ORCID":"https:\/\/orcid.org\/0009-0008-0585-0553","authenticated-orcid":false,"given":"A. C.","family":"Mikliss","sequence":"first","affiliation":[{"name":"b\u2010it and Institute for Computer Science University of Bonn  Bonn Germany"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1200-7248","authenticated-orcid":false,"given":"T.","family":"Schultz","sequence":"additional","affiliation":[{"name":"b\u2010it and Institute for Computer Science University of Bonn  Bonn Germany"},{"name":"Lamarr Institute for Machine Learning and Artificial Intelligence  Bonn Germany"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"311","published-online":{"date-parts":[[2026,6,11]]},"reference":[{"key":"e_1_2_9_2_2","doi-asserted-by":"publisher","DOI":"10.1109\/access.2021.3127960"},{"key":"e_1_2_9_3_2","doi-asserted-by":"publisher","DOI":"10.1134\/s1064230724700138"},{"key":"e_1_2_9_4_2","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0130140"},{"key":"e_1_2_9_5_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.inffus.2024.102608"},{"key":"e_1_2_9_6_2","doi-asserted-by":"crossref","unstructured":"BajajC. L. PascucciV. SchikoreD.: The contour spectrum. InProc. IEEE Vis(1997) pp.167\u2013174. doi:10.1109\/VISUAL.1997.663875. 5","DOI":"10.1109\/VISUAL.1997.663875"},{"key":"e_1_2_9_7_2","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.1986.4767749"},{"key":"e_1_2_9_8_2","doi-asserted-by":"publisher","DOI":"10.1109\/trpms.2023.3265863"},{"key":"e_1_2_9_9_2","doi-asserted-by":"publisher","DOI":"10.1109\/MCG.2025.3533806"},{"key":"e_1_2_9_10_2","doi-asserted-by":"publisher","DOI":"10.1109\/tpami.2023.3261988"},{"key":"e_1_2_9_11_2","unstructured":"CardosoM. J. LiW. BrownR. MaN. KerfootE. WangY. MurreyB. MyronenkoA. ZhaoC. YangD. NathV. HeY. XuZ. HatamizadehA. ZhuW. LiuY. ZhengM. TangY. YangI. ZephyrM. HashemianB. AlleS. DarestaniM. Z. BuddC. ModatM. VercauterenT. WangG. LiY. HuY. FuY. GormanB. JohnsonH. GenereauxB. ErdalB. S. GuptaV. Diaz-PintoA. DoursonA. MaierHeinL. JaegerP. F. BaumgartnerM. Kalpathy-CramerJ. FloresM. KirbyJ. CooperL. A. D. RothH. R. XuD. BericatD. FlocaR. ZhouS. K. ShuaibH. FarahaniK. Maier-HeinK. H. AylwardS. DograP. OurselinS. FengA.:MONAI: An open-source framework for deep learning in healthcare. Tech. Rep. 2211.02701 arXiv 2022. doi:10.48550\/ARXIV.2211.02701. 3"},{"key":"e_1_2_9_12_2","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-006-8711-1"},{"key":"e_1_2_9_13_2","doi-asserted-by":"crossref","unstructured":"FongR. VedaldiA.: Interpretable explanations of black boxes by meaningful perturbation. In2017 IEEE International Conference on Computer Vision (ICCV)(2017) pp.3449\u20133457. doi:10.1109\/ICCV.2017.371. 2","DOI":"10.1109\/ICCV.2017.371"},{"key":"e_1_2_9_14_2","doi-asserted-by":"publisher","DOI":"10.1186\/s13014-022-02035-0"},{"key":"e_1_2_9_15_2","doi-asserted-by":"publisher","DOI":"10.1109\/tbme.2021.3117407"},{"key":"e_1_2_9_16_2","unstructured":"GeirhosR. RubischP. MichaelisC. BethgeM. WichmannF. A. BrendelW.: ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness. InInt'l Conf. on Learning Representations (ICLR)(2019). 2"},{"key":"e_1_2_9_17_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.icte.2024.09.008"},{"key":"e_1_2_9_18_2","doi-asserted-by":"publisher","DOI":"10.1111\/cgf.14302"},{"key":"e_1_2_9_19_2","unstructured":"HoyerL. MunozM. KatiyarP. KhorevaA. FischerV.: Grid saliency for context explanations of semantic segmentation. InAdvances in Neural Information Processing Systems(2019) pp.6459\u20136470. 2"},{"key":"e_1_2_9_20_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-63797-1_9"},{"key":"e_1_2_9_21_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-08999-2_22"},{"key":"e_1_2_9_22_2","doi-asserted-by":"crossref","unstructured":"HasanyS. N. PetitjeanC. M\u00e9riaudeauF.: Seg-XRes-CAM: Explaining spatially local regions in image segmentation. InIEEE\/CVF Conf. on Computer Vision and Pattern Recognition Workshops (CVPRW)(2023) pp.3733\u20133738. doi:10.1109\/CVPRW59228.2023.00384. 2","DOI":"10.1109\/CVPRW59228.2023.00384"},{"key":"e_1_2_9_23_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41592-020-01008-z"},{"key":"e_1_2_9_24_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72114-4_47"},{"key":"e_1_2_9_25_2","doi-asserted-by":"publisher","DOI":"10.3390\/jimaging11090322"},{"key":"e_1_2_9_26_2","doi-asserted-by":"publisher","DOI":"10.1002\/9780470050118.ecse609"},{"key":"e_1_2_9_27_2","unstructured":"LiuZ. LinY. CaoY. HuH. WeiY. ZhangZ. LinS. GuoB.: Swin transformer: Hierarchical vision transformer using shifted windows. InProc. Int'l Conf. on Computer Vision (ICCV)(2021) pp.10012\u201310022. 9"},{"key":"e_1_2_9_28_2","doi-asserted-by":"publisher","DOI":"10.1111\/cgf.14733"},{"key":"e_1_2_9_29_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2016.2598831"},{"key":"e_1_2_9_30_2","unstructured":"McInnesL. HealyJ. MelvilleJ.:UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. Tech. Rep. 1802.03426 arXiv 2018. doi:10.48550\/ARXIV.1802.03426. 6"},{"key":"e_1_2_9_31_2","doi-asserted-by":"publisher","DOI":"10.1109\/JBHI.2023.3267857"},{"key":"e_1_2_9_32_2","doi-asserted-by":"publisher","DOI":"10.1007\/s42979-021-00735-0"},{"key":"e_1_2_9_33_2","doi-asserted-by":"publisher","DOI":"10.1117\/1.jmi.10.5.054001"},{"key":"e_1_2_9_34_2","first-page":"23296","volume":"34","author":"Naseer M. M.","year":"2021","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_2_9_35_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2017.2744358"},{"key":"e_1_2_9_36_2","unstructured":"ParkN. KimS.: How do vision transformers work? InInternational Conference on Learning Representations(2022). URL:https:\/\/openreview.net\/forum?id=D78Go4hVcxO. 9 10"},{"key":"e_1_2_9_37_2","doi-asserted-by":"crossref","unstructured":"PopaA. M. PrasadV. JaspersT. J. van derSommenF. VilanovaA.: SEG-RobustEye: understanding medical image segmentation models. InProc. IEEE Visualization and Visual Analytics (VIS)(2025) pp.76\u201380. doi:10.1109\/vis60296.2025.00021. 2","DOI":"10.1109\/VIS60296.2025.00021"},{"key":"e_1_2_9_38_2","unstructured":"PrinceS. J. D.:Computer Vision: Models Learning and Inference. Cambridge University Press 2012. 1"},{"key":"e_1_2_9_39_2","doi-asserted-by":"publisher","DOI":"10.1109\/tvcg.2022.3219248"},{"key":"e_1_2_9_40_2","doi-asserted-by":"publisher","DOI":"10.1109\/tvcg.2023.3301722"},{"key":"e_1_2_9_41_2","doi-asserted-by":"publisher","DOI":"10.1111\/cgf.14195"},{"key":"e_1_2_9_42_2","first-page":"234","volume-title":"Medical Image Computing and Computer-Assisted Intervention (MICCAI)","author":"Ronneberger O.","year":"2015"},{"key":"e_1_2_9_43_2","doi-asserted-by":"crossref","unstructured":"R\u00e4ukerT. HoA. CasperS. Hadfield-MenellD.: Toward transparent AI: a survey on interpreting the inner structures of deep neural networks. InIEEE Conf. on Secure and Trustworthy Machine Learning (SaTML)(2023). doi:10.1109\/satml54575.2023.00039. 2 10","DOI":"10.1109\/SaTML54575.2023.00039"},{"key":"e_1_2_9_44_2","first-page":"12116","volume":"34","author":"Raghu M.","year":"2021","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_2_9_45_2","doi-asserted-by":"publisher","DOI":"10.1007\/S11263-019-01228-7"},{"key":"e_1_2_9_46_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.01059"},{"key":"e_1_2_9_47_2","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr42600.2020.01428"},{"key":"e_1_2_9_48_2","doi-asserted-by":"publisher","DOI":"10.21037\/qims-24-723"},{"key":"e_1_2_9_49_2","doi-asserted-by":"publisher","DOI":"10.3390\/sci4010003"},{"key":"e_1_2_9_50_2","doi-asserted-by":"crossref","unstructured":"VinogradovaK. DibrovA. MyersG.: Towards interpretable semantic segmentation via gradient-weighted class activation mapping. InAAAI (student abstracts)(2020). 2","DOI":"10.1609\/aaai.v34i10.7244"},{"key":"e_1_2_9_51_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2022.102470"},{"key":"e_1_2_9_52_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2019.2903943"},{"key":"e_1_2_9_53_2","first-page":"1019","volume":"2","author":"Witkin A. P.","year":"1983","journal-title":"Proc. Int'l Joint Conf. on Artificial Intelligence (IJCAI)"},{"key":"e_1_2_9_54_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-10590-1_53"}],"container-title":["Computer Graphics Forum"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1111\/cgf.70470","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/full-xml\/10.1111\/cgf.70470","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1111\/cgf.70470","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,11]],"date-time":"2026-06-11T10:43:10Z","timestamp":1781174590000},"score":1,"resource":{"primary":{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/10.1111\/cgf.70470"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,11]]},"references-count":53,"alternative-id":["10.1111\/cgf.70470"],"URL":"https:\/\/doi.org\/10.1111\/cgf.70470","archive":["Portico"],"relation":{},"ISSN":["0167-7055","1467-8659"],"issn-type":[{"value":"0167-7055","type":"print"},{"value":"1467-8659","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,6,11]]},"assertion":[{"value":"2026-06-11","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e70470"}}