{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,22]],"date-time":"2026-09-22T05:05:13Z","timestamp":1790053513151,"version":"4.0.1"},"reference-count":26,"publisher":"MDPI AG","issue":"18","license":[{"start":{"date-parts":[[2026,9,18]],"date-time":"2026-09-18T00:00:00Z","timestamp":1789689600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"Wonkwang University","id":[{"id":"https:\/\/ror.org\/006776986","id-type":"ROR","asserted-by":"crossref"}]},{"id":[{"id":"https:\/\/ror.org\/006776986","id-type":"ROR","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Diagnostics"],"abstract":"<jats:p>Background\/Objectives: Prototype-based few-shot segmentation is validated almost entirely on centimetre-scale abdominal organs; on pulmonary nodules, 70% of which are less than 10 mm in size, accuracy degrades sharply. We seek to determine how much of that degradation is geometric rather than methodological. Methods: No published method is trained or evaluated here. For any model whose prediction is a union of whole feature grid cells, the attainable Dice coefficient is bounded exactly by a single ratio \u03c1 of the lesion diameter to the grid cell extent; bilinear upsampling relaxes it rather than removing it. We measured this bound on the 50% consensus masks of 2614 LIDC-IDRI nodules, quantified lesion preservation under six mask-handling policies, and swept a training-free implementation across fifteen input geometries. Results: Using SSL-ALPNet\u2019s configuration, no whole-cell decision at that grid can exceed 19.0% Dice on the 5\u20138 mm band, which holds 44.6% of nodules, compared with 63\u201385% on abdominal organs. The governing size is the slice\u2019s equal-area diameter, i.e., a median of 69% of the nominal value, which overstates the bound by 11.5 points. A lung mask used to erase rather than to locate destroys the lesion it should help find; hard masking removes more than half of 11.7% of the 180 nodules (95% CI 7.8\u201317.2). The swept implementation tracks the ceiling (r = 0.74 per episode), attaining 69% of the ceiling at the derived 32 mm tile compared with 8% at full-chest input; at a fixed cell size, it peaks at the encoder\u2019s training resolution. Conclusions: Much of the reported deficit is bounded by representation geometry rather than by modelling quality. We provide a lesion-preserving formulation and derive the tile size from measurement rather than via a chosen margin.<\/jats:p>","DOI":"10.3390\/diagnostics16183027","type":"journal-article","created":{"date-parts":[[2026,9,21]],"date-time":"2026-09-21T12:50:03Z","timestamp":1789995003000},"page":"3027","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Quantifying the Resolution Ceiling of Prototype-Based Few-Shot Segmentation on Sub-Centimetre Pulmonary Nodules: A Training-Free Geometric Analysis"],"prefix":"10.3390","volume":"16","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-1533-7354","authenticated-orcid":false,"given":"Joon-Ho","family":"Cho","sequence":"first","affiliation":[{"name":"Department of Electronics Convergence Engineering, Wonkwang University, 460 Iksandae-ro, Iksan 54538, Jeollabuk-do, Republic of Korea"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1968","published-online":{"date-parts":[[2026,9,18]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"1837","DOI":"10.1109\/TMI.2022.3150682","article-title":"Self-supervised learning for few-shot medical image segmentation","volume":"41","author":"Ouyang","year":"2022","journal-title":"IEEE Trans. Med. Imaging"},{"key":"ref_2","first-page":"610","article-title":"Q-Net: Query-informed few-shot medical image segmentation","volume":"Volume 823","author":"Shen","year":"2024","journal-title":"Intelligent Systems and Applications (IntelliSys 2023)"},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"103412","DOI":"10.1016\/j.media.2024.103412","article-title":"Few-shot medical image segmentation with high-fidelity prototypes","volume":"100","author":"Tang","year":"2025","journal-title":"Med. Image Anal."},{"key":"ref_4","first-page":"271","article-title":"Few-shot medical image segmentation via a region-enhanced prototypical transformer","volume":"Volume 14223","author":"Zhu","year":"2023","journal-title":"Medical Image Computing and Computer Assisted Intervention (MICCAI 2023)"},{"key":"ref_5","doi-asserted-by":"crossref","unstructured":"Ayzenberg, L., Giryes, R., and Greenspan, H. (2024, January 27\u201330). DINOv2 based self supervised learning for few shot medical image segmentation. Proceedings of the 2024 IEEE International Symposium on Biomedical Imaging (ISBI), Athens, Greece.","DOI":"10.1109\/ISBI56570.2024.10635439"},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"11","DOI":"10.1145\/3746224","article-title":"Few-shot learning for medical image segmentation: A review and comparative study","volume":"58","author":"Dissanayake","year":"2026","journal-title":"ACM Comput. Surv."},{"key":"ref_7","doi-asserted-by":"crossref","unstructured":"Zhao, L., Chen, X., Chen, E.Z., Liu, Y., Chen, T., and Sun, S. (2024). Retrieval-augmented few-shot medical image segmentation with foundation models. arXiv.","DOI":"10.1109\/TNNLS.2025.3568479"},{"key":"ref_8","unstructured":"Wang, J., Xu, C., Yang, W., and Yu, L. (2021). A normalized Gaussian Wasserstein distance for tiny object detection. arXiv."},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"182","DOI":"10.1038\/s41592-023-02150-0","article-title":"Understanding metric-related pitfalls in image analysis validation","volume":"21","author":"Reinke","year":"2024","journal-title":"Nat. Methods"},{"key":"ref_10","doi-asserted-by":"crossref","first-page":"3290","DOI":"10.1038\/s41598-025-87829-4","article-title":"Multilevel support-assisted prototype optimization network for few-shot medical segmentation of lung lesions","volume":"15","author":"Tian","year":"2025","journal-title":"Sci. Rep."},{"key":"ref_11","unstructured":"Wang, K., Liew, J.H., Zou, Y., Zhou, D., and Feng, J. (November, January 27). PANet: Few-shot image semantic segmentation with prototype alignment. Proceedings of the IEEE\/CVF International Conference on Computer Vision (ICCV), Seoul, Republic of Korea."},{"key":"ref_12","doi-asserted-by":"crossref","first-page":"102385","DOI":"10.1016\/j.media.2022.102385","article-title":"Anomaly detection-inspired few-shot medical image segmentation through self-supervision with supervoxels","volume":"78","author":"Hansen","year":"2022","journal-title":"Med. Image Anal."},{"key":"ref_13","doi-asserted-by":"crossref","first-page":"2788","DOI":"10.1002\/mp.16933","article-title":"Few-shot segmentation framework for lung nodules via an optimized active contour model","volume":"51","author":"Yang","year":"2024","journal-title":"Med. Phys."},{"key":"ref_14","doi-asserted-by":"crossref","unstructured":"Liu, M.-Y., Tuzel, O., Ramalingam, S., and Chellappa, R. (2011, January 20\u201325). Entropy rate superpixel segmentation. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Colorado Springs, CO, USA.","DOI":"10.1109\/CVPR.2011.5995323"},{"key":"ref_15","doi-asserted-by":"crossref","first-page":"1465","DOI":"10.1109\/TPAMI.2015.2481406","article-title":"Supervised evaluation of image segmentation and object proposal techniques","volume":"38","author":"Marques","year":"2016","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"ref_16","unstructured":"Marin, D., He, Z., Vajda, P., Chatterjee, P., Tsai, S., Yang, F., and Boykov, Y. (November, January 27). Efficient segmentation: Learning downsampling near semantic boundaries. Proceedings of the IEEE\/CVF International Conference on Computer Vision (ICCV), Seoul, Republic of Korea."},{"key":"ref_17","doi-asserted-by":"crossref","unstructured":"Mojtahedi, R., Hamghalam, M., Do, R.K.G., and Simpson, A.L. (2022). Towards optimal patch size in vision transformers for tumor segmentation. Multiscale Multimodal Medical Imaging (MMMI 2022), Springer. Lecture Notes in Computer Science.","DOI":"10.1007\/978-3-031-18814-5_11"},{"key":"ref_18","unstructured":"Fu, S., Hamilton, M., Brandt, L., Feldman, A., Zhang, Z., and Freeman, W.T. (2024, January 7\u201311). FeatUp: A model-agnostic framework for features at any resolution. Proceedings of the International Conference on Learning Representations (ICLR), Vienna, Austria."},{"key":"ref_19","doi-asserted-by":"crossref","first-page":"50","DOI":"10.1186\/s41747-020-00173-2","article-title":"Automatic lung segmentation in routine imaging is primarily a data diversity problem, not a methodology problem","volume":"4","author":"Hofmanninger","year":"2020","journal-title":"Eur. Radiol. Exp."},{"key":"ref_20","first-page":"783","article-title":"Convolutional neural network architectures for texture classification of pulmonary nodules","volume":"Volume 11401","author":"Ferreira","year":"2019","journal-title":"Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications (CIARP 2018)"},{"key":"ref_21","doi-asserted-by":"crossref","first-page":"102027","DOI":"10.1016\/j.media.2021.102027","article-title":"LNDb challenge on automatic lung cancer patient management","volume":"70","author":"Pedrosa","year":"2021","journal-title":"Med. Image Anal."},{"key":"ref_22","doi-asserted-by":"crossref","first-page":"915","DOI":"10.1118\/1.3528204","article-title":"The Lung Image Database Consortium (LIDC) and Image Database Resource Initiative (IDRI): A completed reference database of lung nodules on CT scans","volume":"38","author":"Armato","year":"2011","journal-title":"Med. Phys."},{"key":"ref_23","unstructured":"Armato, S.G., McLennan, G., Bidaut, L., McNitt-Gray, M.F., Meyer, C.R., Reeves, A.P., Zhao, B., Aberle, D.R., Henschke, C.I., and Hoffman, E.A. (2015). Data From LIDC-IDRI, Version 4 [Data set]. Cancer Imaging Arch."},{"key":"ref_24","doi-asserted-by":"crossref","first-page":"1045","DOI":"10.1007\/s10278-013-9622-7","article-title":"The Cancer Imaging Archive (TCIA): Maintaining and operating a public information repository","volume":"26","author":"Clark","year":"2013","journal-title":"J. Digit. Imaging"},{"key":"ref_25","doi-asserted-by":"crossref","first-page":"044504","DOI":"10.1117\/1.JMI.3.4.044504","article-title":"Lung nodule malignancy classification using only radiologist-quantified image features as inputs to statistical learning algorithms","volume":"3","author":"Hancock","year":"2016","journal-title":"J. Med. Imaging"},{"key":"ref_26","unstructured":"Oquab, M., Darcet, T., Moutakanni, T., Vo, H., Szafraniec, M., Khalidov, V., Fernandez, P., Haziza, D., Massa, F., and El-Nouby, A. (2024). DINOv2: Learning robust visual features without supervision. arXiv."}],"container-title":["Diagnostics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/2075-4418\/16\/18\/3027\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,22]],"date-time":"2026-09-22T04:36:00Z","timestamp":1790051760000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/2075-4418\/16\/18\/3027"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,18]]},"references-count":26,"journal-issue":{"issue":"18","published-online":{"date-parts":[[2026,9]]}},"alternative-id":["diagnostics16183027"],"URL":"https:\/\/doi.org\/10.3390\/diagnostics16183027","relation":{},"ISSN":["2075-4418"],"issn-type":[{"value":"2075-4418","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,18]]}}}