European Radiology最新文献

{"title":"Deep learning model for noninvasive prediction of Ki-67 expression and prognostic stratification in breast cancer: a multicenter retrospective study.","authors":"Weilu Yu, Pei Chen, Suwan Chai, Wentong Ding, Lin Zhang","doi":"10.1007/s00330-025-12203-x","DOIUrl":"https://doi.org/10.1007/s00330-025-12203-x","url":null,"abstract":"<p><strong>Objective: </strong>Ki-67 correlates with prognosis for patients with breast cancer. However, the evaluation of Ki-67expression relies on pathological analysis and invasive biopsy, which hinders its wide adoption. This work sought to develop a noninvasive Ki-67 prediction model for breast cancer patients through ultrasound and clinical information and evaluate model performance in risk stratification of lymph metastasis and prognosis.</p><p><strong>Materials and methods: </strong>Clinical, ultrasound, pathological, and prognostic information were collected from four centers to develop a deep learning (DL) model. Ultrasound features were extracted by ResNet-50 and integrated with clinical information through logistic regression. Class activation mapping and nomograms were used to visualize the prediction process. Area under curve (AUC), confusion matrices, calibration curves, and decision curve analysis were used to evaluate model performance on Ki-67. Prognostic relevance was evaluated with lymph node metastasis and recurrence-free survival (RFS).</p><p><strong>Results: </strong>From January 2021 to December 2024, 456 patients from three centers were collected as training (n = 264), validation (n = 96), and internal test (n = 96) sets, 204 patients from an independent center were collected as an external test set. In the external set, the combined model achieved satisfactory performance on Ki-67 (AUC = 0.828, 95% CI: 0.761-0.890). High Ki-67 group showed higher lymph metastasis rates (67.7% vs 16.2%, p < 0.001) and worse RFS (p = 0.041) than the low Ki-67 group. The combined model achieved the best predictive ability on recurrence in the first 6 months after operation (AUC = 0.820).</p><p><strong>Conclusion: </strong>This noninvasive model could predict Ki-67 status, classify the risk of lymph metastasis, and provide prognostic insights. Its wide application would contribute to the formulation of individualized treatment plans and follow-up strategies.</p><p><strong>Key points: </strong>Question Ki-67 is an important pathological information in breast cancer and is meaningful for lymph metastasis and prognosis. However, it can currently only be evaluated through invasive testing. Findings We developed a non-invasive model based on ultrasound and clinical information, which can predict Ki-67 status (accuracy = 0.828), classify lymph metastasis (accuracy = 0.765), and provide prognostic insights (p = 0.041). Clinical relevance This model enabled preoperative prediction of Ki-67 status and lymph metastasis for patients with breast cancer, thereby informing surgical planning. Furthermore, it demonstrated prognostic utility, facilitating the development of personalized patient follow-up strategies.</p>","PeriodicalId":12076,"journal":{"name":"European Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147835349","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Commercial AI for CT lung cancer screening: product capabilities, coverage of nodule management tasks and supporting evidence.","authors":"Noa Antonissen, Steven Schalekamp, Horst K Hahn, Kicky G van Leeuwen, Colin Jacobs","doi":"10.1007/s00330-026-12580-x","DOIUrl":"https://doi.org/10.1007/s00330-026-12580-x","url":null,"abstract":"<p><strong>Objectives: </strong>To characterize the capabilities of CE-marked AI products for lung nodule analysis in lung cancer screening (LCS), quantify their coverage of tasks defined in nodule management recommendations, and assess their peer-reviewed evidence.</p><p><strong>Materials and methods: </strong>Six core tasks in LCS (nodule detection, classification, measurement, growth assessment, malignancy risk estimation, and structured management) were derived from 4 nodule management recommendations: Lung-RADS 2022, British Thoracic Society (BTS) guidelines, European Union Position Statement (EUPS), and European Society of Thoracic Imaging (ESTI). Products were identified through www.healthairegister.com . Vendors confirmed capabilities using a standardized questionnaire; public documentation supplemented non-responders. Task coverage was calculated as the percentage of functional overlap (0-100%) per recommendation. Peer-reviewed evidence was evaluated using a six-level efficacy framework and assessed for study characteristics.</p><p><strong>Results: </strong>In total, 16 products from 16 vendors were included; 10 vendors completed questionnaires. Analysis showed that 14 products detect and measure solid and subsolid nodules, 12 support growth assessment, and 9 provide malignancy risk estimation (PanCan in 5, AI-based scores in 4). No product provides support for endobronchial or cystic lesions. High task coverage (> 75%) was observed in 10 products for EUPS and 4 for BTS, whereas no product achieved high coverage for Lung-RADS or ESTI. Overall, 60 peer-reviewed studies were identified; 7% were prospective and evidence clustered at lower efficacy levels: 70% assessed diagnostic accuracy, while none reported patient outcomes or societal impact.</p><p><strong>Conclusion: </strong>Numerous CE-certified AI products could support CT-based lung cancer screening, but gaps in task coverage and predominantly lower-level evidence necessitate cautious, monitored implementation.</p><p><strong>Key points: </strong>Question Do commercially available AI products for lung nodule analysis functionally cover international nodule management recommendation-defined tasks, and what peer-reviewed clinical evidence supports them? Findings AI products support standard nodule detection and measurement in line with management recommendations but lack support for endobronchial or cystic lesions and high-level clinical evidence. Clinical relevance CE-marked AI products can assist radiologists with core lung cancer screening tasks, but capability gaps exist. Limited high-level clinical evidence complicates integrating AI into guidelines, securing reimbursement, and formulating recommendations for its use in lung cancer screening programs.</p>","PeriodicalId":12076,"journal":{"name":"European Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147835396","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Why aren't the ENZIAN and # ENZIAN classification systems used more widely in Europe?","authors":"Susan Barter","doi":"10.1007/s00330-026-12605-5","DOIUrl":"https://doi.org/10.1007/s00330-026-12605-5","url":null,"abstract":"","PeriodicalId":12076,"journal":{"name":"European Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147835472","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Photon counting detector CT contrast agent-reduced transcatheter aortic valve reconstruction planning: a comparative study.","authors":"Yannik C Layer, Alexander Isaak, Narine Mesropyan, Patrick A Kupczyk, Dmitrij Kravchenko, Marilia Voigt, Tatjana Dell, Julian A Luetkens, Daniel Kuetting","doi":"10.1007/s00330-026-12613-5","DOIUrl":"https://doi.org/10.1007/s00330-026-12613-5","url":null,"abstract":"<p><strong>Objectives: </strong>Continuous efforts are made to reduce contrast media, improving patient safety, reducing environmental risks, and addressing recurring supply shortages. The aim of this study was to evaluate contrast agent-reduced CT protocols for transcatheter aortic valve reconstruction (TAVR) planning in photon counting detector CT (PCDCT).</p><p><strong>Materials and methods: </strong>162 BMI-matched examinations with standard dose contrast media (SCD; 80 mL; Iohexol 300 mg/mL; 81 examinations) and reduced contrast media dose (RCD; 50 mL; 81 examinations) for TAVR planning on a PCDCT were included in this retrospective monocentric study. Virtual monoenergetic reconstructions (VMI) at 70 keV, 60 keV and 50 keV of contrast agent-reduced examinations were compared with polyenergetic images. Quantitatively, regions-of-interest (ROIs) were placed in the abdominal aorta, iliac bifurcation, femoral artery, left ventricle and trapezius muscles. Signal-to-noise-ratio (SNR) and contrast-to-noise-ratio (CNR) were calculated. Qualitatively, diagnostic quality and contrast were assessed on a visual grading scale of 1 (non-diagnostic) - 5 (excellent) and contrast agent dose was estimated.</p><p><strong>Results: </strong>Averaged, SNR and CNR decreased by 8.71% and 16.78%, respectively, on PCDCT with reduced contrast dose (RCD vs. SCD; both p < 0.001). VMI50keV increased SNR by 44.10% (p < 0.001) and CNR by 52.73% (p < 0.001) compared with SCD. In the ascending aorta, SNR increased from 19.80 ± 6.24 (SCD) to 35.78 ± 13.20 (RCD_VMI50keV) and CNR from 18.84 ± 7.78 to 29.77 ± 16.70. Median contrast intensity was 5 for SCD, 4 for RCD_CR, and 5 for RCD_VMI50keV.</p><p><strong>Conclusion: </strong>The diagnostic efficacy of TAVR planning assessment using PCDCT with minimized contrast agent dosing is preserved, presenting a practical approach to conserve contrast media.</p><p><strong>Key points: </strong>Question The aim of the study was to implement a PCDCT-adapted contrast media dose protocol to reduce contrast agent volume at sufficient diagnostic quality. Findings PCDCT enables substantial contrast dose reduction for TAVR planning with maintained diagnostic image quality. Low-keV virtual monoenergetic image reconstructions compensate for the reduced iodine concentration. Clinical relevance The study demonstrates the potential of contrast media reduction of PCD-CT in clinical routine. This can benefit patients with renal impairment, for example, and reduce the negative effects of iodinated contrast media on the environment.</p>","PeriodicalId":12076,"journal":{"name":"European Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147835478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Letter to the Editor: Deep learning TMJ MRI-reader-level equivalence is a foundation, not a finish line.","authors":"Carlos Fernando Mourão, Antonio Coutinho, Luiz Eduardo Juliasse, Rodrigo Dos Santos Pereira","doi":"10.1007/s00330-026-12606-4","DOIUrl":"https://doi.org/10.1007/s00330-026-12606-4","url":null,"abstract":"","PeriodicalId":12076,"journal":{"name":"European Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147835480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Correction: High-frequency ultrasound combined with microbubbles for preoperative lymphatic mapping for lymphedema with a non-linear pattern in indocyanine green lymphography.","authors":"ShuFang Yuan, LiPing Chen, LanJing Wu, ChenYang Zhao, JingJing Wen, Long Biao Yu, ZheGang Zhou, DeSheng Sun, ZhengMing Hu","doi":"10.1007/s00330-026-12509-4","DOIUrl":"https://doi.org/10.1007/s00330-026-12509-4","url":null,"abstract":"","PeriodicalId":12076,"journal":{"name":"European Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147835399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CT quantification of interstitial lung abnormalities: a prospective comparison between supine and prone positions.","authors":"Yura Ahn, Sang Min Lee, Jung Im Kim, Kyung-Hyun Do, Jang Ho Lee, Ho Cheol Kim, Joon Beom Seo","doi":"10.1007/s00330-026-12595-4","DOIUrl":"https://doi.org/10.1007/s00330-026-12595-4","url":null,"abstract":"<p><strong>Objective: </strong>To prospectively evaluate positional variability in quantitative CT (qCT)-derived measurements of interstitial lung abnormality (ILA) using the same-day supine and prone CT.</p><p><strong>Materials and methods: </strong>In this prospective study (February 2024-February 2025), participants with ILA underwent sequential non-contrast supine and prone CT scans using identical acquisition parameters. A commercially available deep learning-based software quantified fibrotic (reticulation and honeycombing) and nonfibrotic (ground-glass opacity) ILA components in accordance with Fleischner Society definitions. qCT differences between supine and prone measurements were assessed using paired t tests, Bland-Altman analysis with 95% limits of agreement (LOA), and concordance correlation coefficients (CCC).</p><p><strong>Results: </strong>Of 47 consented participants, 38 (mean age, 70.9 ± 6.4 years; 27 men) were included in the final analysis. Mean total ILA extent was greater on supine than on prone scans (1.76% vs 1.39%, p = 0.02), largely due to a greater extent of fibrotic ILAs on supine images (1.43% vs 1.12%, p = 0.007). The 95% LOA between supine and prone scans were -1.49% to 2.23% for total ILA extent, -0.97% to 1.58% for the fibrotic component, and -0.89% to 1.03% for the nonfibrotic component. Agreement between positions was moderate for the fibrotic component (CCC = 0.770), and poor for the nonfibrotic component (CCC = 0.431), with the lowest reproducibility observed in dependent lung zones.</p><p><strong>Conclusion: </strong>The measurement variability of qCT result of ILA between supine and prone scans was approximately 1.9%, with moderate agreement for the fibrotic component but poor agreement for the nonfibrotic component between the same-day supine and prone CT scans.</p><p><strong>Key points: </strong>Question Whether qCT-derived measurements of ILA are interchangeable between supine and prone CT scans remains unknown. Findings qCT measurements showed approximately 1.9% positional variability, with moderate agreement for fibrotic components but poor agreement for nonfibrotic components, especially in dependent lung zones. Clinical relevance Supine and prone scans should not be used interchangeably for qCT analysis of ILA. Consistent patient positioning is essential to ensure accurate longitudinal assessment.</p>","PeriodicalId":12076,"journal":{"name":"European Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147835421","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Reply to the Letter to the Editor: Deep learning TMJ MRI-reader-level equivalence is a foundation, not a finish line.","authors":"Gyu-Dong Jo, Kug Jin Jeon, Yoon Joo Choi, Chena Lee, Sang-Sun Han","doi":"10.1007/s00330-026-12607-3","DOIUrl":"https://doi.org/10.1007/s00330-026-12607-3","url":null,"abstract":"","PeriodicalId":12076,"journal":{"name":"European Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147835483","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"High risk score of breast cancer by artificial intelligence (AI) on screening mammograms: a review of negative and cancer cases.","authors":"Marit A Martiniussen, Marie B Bergan, Merete U Kristiansen, Nataliia Moshina, Anne Sofie F Larsen, Marthe Larsen, Fredrik A Dahl, Solveig Hofvind","doi":"10.1007/s00330-026-12579-4","DOIUrl":"https://doi.org/10.1007/s00330-026-12579-4","url":null,"abstract":"<p><strong>Objectives: </strong>To investigate mammographic features associated with high artificial intelligence (AI) risk scores as provided by two AI models applied to screening mammograms.</p><p><strong>Materials and methods: </strong>This retrospective study included 130,031 screening mammograms from 42,371 women attending BreastScreen Norway, 2008-2018. Two AI models (A and B) developed for cancer detection on screening mammograms were applied. An informed radiological review was conducted for mammograms within the highest 5% of AI risk scores by both models in two study samples: (1) High AI risk score, but no breast cancer detected within 6 years (n = 120), and (2) High AI risk score in mammograms with screen-detected cancers (n = 120). Mammographic density (BI-RADS a-d), features (mass, spiculated mass, asymmetry, architectural distortion, calcification alone, and density with calcification), and radiologists' interpretation scores (1-5) were analyzed descriptively.</p><p><strong>Results: </strong>Mammographic density was higher in sample 1 compared to sample 2 (BI-RADS d: 11% vs 3%, respectively). In sample 1, calcifications alone were the most frequent AI-marked feature (model A: 72%; model B: 68%), predominantly with amorphous morphology and a cluster distribution, and 76% were interpreted as benign by the radiologists (interpretation score 1). In sample 2, a spiculated mass was the most frequent mammographic feature among the screen-detected cancers (29%).</p><p><strong>Conclusion: </strong>Mammograms assigned high AI risk scores exhibit distinct features depending on screening outcome. Systematic characterization of these features may help refine AI thresholds, improve specificity, reduce AI false-positive findings, and decrease the recall rate in breast cancer screening.</p><p><strong>Key points: </strong>Question Knowledge about mammographic features associated with high AI risk scores is essential for distinguishing cancer from non-cancer cases. Findings Calcifications were the dominant feature in non-cancers in screening mammograms with high AI risk score, whereas spiculated mass was the most frequent feature among cancers. Clinical relevance Calcifications in non-cancer screening mammograms with a high AI risk score were frequently interpreted as benign or probably benign by radiologists. This knowledge may help refine AI thresholds and thereby improve specificity and reduce false-positive results in mammographic screening.</p>","PeriodicalId":12076,"journal":{"name":"European Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147835520","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Open-weight language models for procedural impression generation.","authors":"Xuerong Jia, Wusheng Zhu, Kangmo Huang","doi":"10.1007/s00330-026-12612-6","DOIUrl":"https://doi.org/10.1007/s00330-026-12612-6","url":null,"abstract":"","PeriodicalId":12076,"journal":{"name":"European Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147835505","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}