{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,23]],"date-time":"2026-09-23T07:02:03Z","timestamp":1790146923561,"version":"4.0.1"},"reference-count":38,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,9,23]],"date-time":"2026-09-23T00:00:00Z","timestamp":1790121600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100000265","name":"Medical Research Council","doi-asserted-by":"publisher","award":["MR\/W006804\/1"],"award-info":[{"award-number":["MR\/W006804\/1"]}],"id":[{"id":"10.13039\/501100000265","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Med. Technol."],"abstract":"<jats:p>\n                    Segmentation of ultrasound images is complex, and although deep learning has improved automation, trust remains limited, restricting adoption. Interpretability methods may help reveal model behaviours and failure modes, supporting more trustworthy AI systems. A U-Net with a ResNet-18 backbone was trained to segment pipes in images of the Edinburgh Pipe Phantom acquired using ten ultrasound probes (\n                    <jats:italic>N<\/jats:italic>\n                    \u2009=\u2009162). Performance was assessed using five-fold cross-validation. Predictive entropy and variance across 20 Monte Carlo dropout forward passes were used as approximations of data-related and model-related uncertainty. Calibration was assessed using the Brier score and expected calibration error (ECE). Input-gradient saliency and Grad-CAM maps from four encoder stages and the final decoder block were evaluated. For exploratory prospective quality control, complete empty-mask failures were assessed directly from model output, while non-empty predictions were classified retrospectively using reference-based performance criteria and reference-free uncertainty and morphological features were evaluated as candidate indicators. The model achieved a Dice score of 0.81\u2009\u00b1\u20090.03 and an HD95 of 11.98\u2009\u00b1\u20093.39. Seven segmentation errors were identified: three complete failures in which no foreground mask was predicted and four non-empty predictions in which the incorrect pipe was segmented. Dice score was associated with the Brier score (Spearman\n                    <jats:italic>\u03c1<\/jats:italic>\n                    \u2009=\u2009\u22120.96) and ECE (\n                    <jats:italic>\u03c1<\/jats:italic>\n                    \u2009=\u2009\u22120.72), whereas predictive entropy and Monte Carlo dropout variance showed weak associations with segmentation performance (|\n                    <jats:italic>\u03c1<\/jats:italic>\n                    |\u2009&amp;lt;\u20090.4). In selected failure examples, uncertainty and activation sometimes overlapped the reference pipe region despite incorrect segmentation. Complete empty-mask failures were deterministically identifiable from the model output, whereas the evaluated reference-free indicators showed limited ability to identify poor-quality non-empty segmentations. Within this phantom dataset and model configuration, calibration metrics were more closely associated with segmentation quality than the evaluated uncertainty approximations. Attention and uncertainty maps provided complementary descriptive information about selected errors, but indicated spatial sensitivity rather than causal model reasoning. Simple output checks enabled deterministic identification of complete segmentation failures; however, the evaluated reference-free indicators were insufficient for validated prospective identification of subtler non-empty errors. These findings support multimodal model evaluation for EPP segmentation while emphasising the need for validation on independent datasets, devices and architectures in future studies.\n                  <\/jats:p>","DOI":"10.3389\/fmedt.2026.1918432","type":"journal-article","created":{"date-parts":[[2026,9,23]],"date-time":"2026-09-23T06:20:57Z","timestamp":1790144457000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Beyond segmentation accuracy: understanding model reliability and failure modes in ultrasound image analysis"],"prefix":"10.3389","volume":"8","author":[{"given":"Agathe","family":"Bricout","sequence":"first","affiliation":[{"name":"Institute for Neurological and Cardiovascular Research, University of Edinburgh","place":["Edinburgh, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"David","family":"Hardman","sequence":"additional","affiliation":[{"name":"Centre for Medical Informatics, University of Edinburgh","place":["Edinburgh, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Scott","family":"Inglis","sequence":"additional","affiliation":[{"name":"Medical Physics, NHS Lothian","place":["Edinburgh, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Stephen","family":"Pye","sequence":"additional","affiliation":[{"name":"Institute for Neurological and Cardiovascular Research, University of Edinburgh","place":["Edinburgh, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Carmel M.","family":"Moran","sequence":"additional","affiliation":[{"name":"Institute for Neurological and Cardiovascular Research, University of Edinburgh","place":["Edinburgh, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,9,23]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"251","DOI":"10.1016\/J.ZEMEDI.2023.05.005","article-title":"Medical ultrasound: time-honored method or emerging research frontier?","volume":"33","author":"Ruiter","year":"2023","journal-title":"Z Med Phys"},{"key":"B2","doi-asserted-by":"publisher","first-page":"657","DOI":"10.1097\/RLI.0000000000000188","article-title":"Ultrasound in radiology: from anatomic, functional, molecular imaging to drug delivery and image-guided therapy","volume":"50","author":"Klibanov","year":"2015","journal-title":"Invest Radiol"},{"key":"B3","article-title":"Ultrasound physics and instrumentation","volume-title":"StatPearls","author":"Grogan","year":"2023"},{"key":"B4","doi-asserted-by":"publisher","first-page":"8","DOI":"10.1177\/1742271X13511805","article-title":"BMUS Guidelines for the regular quality assurance testing of ultrasound scanners by sonographers","volume":"22","author":"Dudley","year":"2014","journal-title":"Ultrasound"},{"key":"B5","doi-asserted-by":"publisher","first-page":"654","DOI":"10.1038\/s41467-024-44824-z","article-title":"Segment anything in medical images","volume":"15","author":"Ma","year":"2024","journal-title":"Nat Commun"},{"key":"B6","doi-asserted-by":"publisher","first-page":"317","DOI":"10.1007\/S00701-024-06179-8","article-title":"Challenges with segmenting intraoperative ultrasound for brain tumours","volume":"166","author":"Weld","year":"2024","journal-title":"Acta Neurochir (Wien"},{"key":"B7","doi-asserted-by":"publisher","first-page":"103768","DOI":"10.1016\/J.INFFUS.2025.103768","article-title":"Ultrasound image segmentation: a systematic review of deformable models from classical techniques to intelligent advancements","volume":"127","author":"Sharma","year":"2026","journal-title":"Inf Fus"},{"key":"B8","doi-asserted-by":"publisher","first-page":"234","DOI":"10.1007\/978-3-319-24574-4_28","article-title":"U-net: convolutional networks for biomedical image segmentation","volume":"9351","author":"Ronneberger","year":"2015","journal-title":"Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)"},{"key":"B9","first-page":"770","article-title":"Deep residual learning for image recognition","author":"He","year":"2016"},{"key":"B10","doi-asserted-by":"publisher","first-page":"203","DOI":"10.1038\/s41592-020-01008-z","article-title":"nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation","volume":"18","author":"Isensee","year":"2021","journal-title":"Nat Methods"},{"key":"B11","doi-asserted-by":"publisher","first-page":"52","DOI":"10.3390\/TOMOGRAPHY11050052","article-title":"Medical image segmentation: a comprehensive review of deep learning-based methods","volume":"11","author":"Gao","year":"2025","journal-title":"Tomography"},{"key":"B12","doi-asserted-by":"publisher","first-page":"100115","DOI":"10.1016\/J.WFUMBO.2026.100115","article-title":"Artificial intelligence in ultrasound imaging: inevitable, promising, and in need of responsible stewardship","author":"Dighe","year":"2026","journal-title":"WFUMB Ultrasound Open"},{"key":"B13","doi-asserted-by":"publisher","first-page":"685","DOI":"10.3390\/EJIHPE14030045","article-title":"Artificial-Intelligence-Based clinical decision support systems in primary care: a scoping review of current clinical implementations","volume":"14","author":"Gomez-Cabello","year":"2024","journal-title":"Eur J Investig Health Psychol Educ"},{"key":"B14","doi-asserted-by":"publisher","first-page":"195","DOI":"10.1186\/S12916-019-1426-2","article-title":"Key challenges for delivering clinical impact with artificial intelligence","volume":"17","author":"Kelly","year":"2019","journal-title":"BMC Med"},{"key":"B15","doi-asserted-by":"publisher","first-page":"e53207","DOI":"10.2196\/53207","article-title":"How explainable artificial intelligence can increase or decrease Clinicians\u2019 trust in AI applications in health care: systematic review","volume":"3","author":"Rosenbacke","year":"2024","journal-title":"JMIR AI"},{"key":"B16","doi-asserted-by":"publisher","first-page":"2019","DOI":"10.1162\/99608F92.5A8A3A3D","article-title":"Why are we using black box models in AI when we don\u2019t need to? A lesson from an explainable AI competition","volume":"1","author":"Rudin","year":"2019","journal-title":"Harv Data Sci Rev"},{"key":"B17","doi-asserted-by":"publisher","first-page":"2204","DOI":"10.3390\/DIAGNOSTICS15172204\/S1","article-title":"Enhancing clinician trust in AI diagnostics: a dynamic framework for confidence calibration and transparency","volume":"15","author":"Yu","year":"2025","journal-title":"Diagnostics"},{"key":"B18","first-page":"318","article-title":"Estimating uncertainty in deep learning for reporting confidence to clinicians when segmenting nuclei image data","author":"Ghoshal","year":"2019"},{"key":"B19","doi-asserted-by":"publisher","first-page":"336","DOI":"10.1007\/s11263-019-01228-7","article-title":"Grad-CAM: visual explanations from deep networks via gradient-based localization","volume":"128","author":"Selvaraju","year":"2020","journal-title":"Int J Comput Vis"},{"key":"B20","doi-asserted-by":"publisher","first-page":"16887","DOI":"10.1038\/s41598-023-42946-w","article-title":"Ground truth based comparison of saliency maps algorithms","volume":"13","author":"Szczepankiewicz","year":"2023","journal-title":"Sci Rep"},{"key":"B21","first-page":"80","article-title":"Explaining explanations: an overview of interpretability of machine learning","author":"Gilpin","year":"2018"},{"key":"B22","doi-asserted-by":"publisher","first-page":"187","DOI":"10.1088\/1742-6596\/1\/1\/041","article-title":"Medical ultrasound: a new metric of performance for grey-scale imaging","volume":"1","author":"Pye","year":"2004","journal-title":"J Phys Conf Ser"},{"key":"B23","doi-asserted-by":"publisher","first-page":"393","DOI":"10.1055\/A-1194-3818","article-title":"The imaging performance of diagnostic ultrasound scanners using the Edinburgh pipe phantom to measure the resolution integral-15 years of experience","volume":"43","author":"Moran","year":"2022","journal-title":"Ultraschall Med"},{"key":"B24","first-page":"17","article-title":"Resnet versus U-net for measurements of imaging performance of ultrasound scanners","author":"Bricout","year":"2025"},{"key":"B25","author":"Bricout","year":"2025"},{"key":"B26","first-page":"2276","article-title":"The VIA annotation software for images, audio and video","author":"Dutta","year":"2019"},{"key":"B27","doi-asserted-by":"publisher","first-page":"5067","DOI":"10.1088\/0031-9155\/55\/17\/012","article-title":"The resolution integral: visual and computational approaches to characterizing ultrasound images","volume":"55","author":"MacGillivray","year":"2010","journal-title":"Phys Med Biol"},{"key":"B28","doi-asserted-by":"publisher","first-page":"127","DOI":"10.11648\/J.AJTAS.20241305.13","article-title":"Comparative analysis of cross-validation techniques: LOOCV, K-folds cross-validation, and repeated K-folds cross-validation in machine learning models","volume":"13","author":"Lumumba","year":"2024","journal-title":"Am J Theor Appl Stat"},{"key":"B29","author":"Iakubovskii","year":"2019"},{"key":"B30","first-page":"929","article-title":"PyTorch 2: faster machine learning through dynamic Python bytecode transformation and graph compilation","volume-title":"Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems","author":"Ansel","year":"2024"},{"key":"B31","first-page":"248","article-title":"Imagenet: a large-scale hierarchical image database","author":"Deng","year":"2009"},{"key":"B32","first-page":"1050","article-title":"Dropout as a Bayesian approximation: representing model uncertainty in deep learning","author":"Gal","year":"2016"},{"key":"B33","doi-asserted-by":"publisher","first-page":"297","DOI":"10.2307\/1932409","article-title":"Measures of the amount of ecologic association between Species","volume":"26","author":"Dice","year":"1945","journal-title":"Ecology"},{"key":"B34","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1186\/S41747-020-00200-2\/FIGURES\/2","article-title":"On the usage of average Hausdorff distance for segmentation performance assessment: hidden error when used for ranking","volume":"5","author":"Aydin","year":"2021","journal-title":"Eur Radiol Exp"},{"key":"B35","year":""},{"key":"B36","doi-asserted-by":"publisher","first-page":"457","DOI":"10.1007\/s10994-021-05946-3","article-title":"Aleatoric and epistemic uncertainty in machine learning: an introduction to concepts and methods","volume":"110","author":"H\u00fcllermeier","year":"2021","journal-title":"Mach Learn"},{"key":"B37","doi-asserted-by":"publisher","first-page":"100242","DOI":"10.1016\/J.GLOEPI.2025.100242","article-title":"On misconceptions about the brier score in binary prediction models","volume":"11","author":"Hoessly","year":"2026","journal-title":"Glob Epidemiol"},{"key":"B38","author":"Pavlovic","year":"2025"}],"container-title":["Frontiers in Medical Technology"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmedt.2026.1918432\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,23]],"date-time":"2026-09-23T06:20:58Z","timestamp":1790144458000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmedt.2026.1918432\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,23]]},"references-count":38,"alternative-id":["10.3389\/fmedt.2026.1918432"],"URL":"https:\/\/doi.org\/10.3389\/fmedt.2026.1918432","relation":{},"ISSN":["2673-3129"],"issn-type":[{"value":"2673-3129","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,23]]},"article-number":"1918432"}}