Radiology-Artificial Intelligence最新文献

{"title":"More than Meets the Eye: Exposure Parameters Cause Shortcut Learning in Radiography Deep Learning Models.","authors":"Bastien Le Guellec, Guillaume Chassagnon","doi":"10.1148/ryai.260590","DOIUrl":"https://doi.org/10.1148/ryai.260590","url":null,"abstract":"","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":"8 4","pages":"e260590"},"PeriodicalIF":20.1,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148550537","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Clinical Implementation of AI for Pulmonary Embolism Detection in over 30 000 CT Pulmonary Angiography Examinations.","authors":"Shlomit Goldberg-Stein, Amir Gandomi, Matthew A Barish, Stuart L Cohen, David Hirschorn, Rakesh Shah, Ritesh D Patel, Elizabeth Y Rula, Jason Naidich, Pina C Sanelli","doi":"10.1148/ryai.250017","DOIUrl":"10.1148/ryai.250017","url":null,"abstract":"<p><p>Purpose To quantify postimplementation concordance between a U.S. Food and Drug Administration-cleared artificial intelligence (AI) tool and AI-informed radiologists for pulmonary embolism (PE) detection at CT pulmonary angiography, with real-time adjudication of discordances. Materials and Methods A commercial PE AI tool was retrospectively implemented in the clinic across an integrated network (August 9, 2021-February 20, 2023). Adult CT pulmonary angiographic acquisitions underwent real-time AI analysis and radiologist interpretation. Radiologist-AI disagreements triggered adjudication by thoracic radiologists via the AI quality oversight process. Adjudicator diagnosis served as the reference standard for discordant cases. Concordance was measured and diagnostic performance of radiologists and AI was compared using adjudication for discordant cases. Results A total of 32 501 CT pulmonary angiographic acquisitions obtained from 29 492 patients (mean age, 62.4 years ± 18.6 [SD], 17 424 female patients) were evaluated. PE positivity was 9.93% (3226 of 32 501). Overall concordance was 97.79% (95% CI: 97.62, 97.94) and was higher for AI-negative than for AI-positive examinations (98.18% vs 93.75%; <i>P</i> < .001). Expert adjudication favored the radiologist in 88.73% of discordances. The rate of unique diagnosis by the interpreting radiologist (483 of 3226 [14.97%]) was approximately 19 times that of the AI tool alone (26 of 3226 [0.81%]). Concordance varied by PE features: acute versus chronic (87.34% vs 60.12%; <i>P</i> < .001) and location (central, 95.79%; lobar and/or segmental, 83.81%; subsegmental, 58.62%; all <i>P</i> < .001). Conclusion In large-scale deployment, AI showed high concordance with radiologists and made meaningful contributions in discordant reviews while expert oversight confirmed complementary roles and highlighted scenarios of radiologist-AI divergence. <b>Keywords:</b> Embolism, Thrombosis, Thorax, Pulmonary Arteries <i>Supplemental material is available for this article.</i> © RSNA, 2026.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250017"},"PeriodicalIF":20.1,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147934051","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Toward Personalized Care of Intracranial Aneurysms.","authors":"Ahmed Maiter, Saminderjit Kular","doi":"10.1148/ryai.260509","DOIUrl":"10.1148/ryai.260509","url":null,"abstract":"","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":"8 4","pages":"e260509"},"PeriodicalIF":20.1,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148309679","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Alignment of Policy, Practice, and Patient Safety for Trustworthy AI in Radiology.","authors":"Florence X Doo, Melissa A Davis, Jason Poff, Yvonne W Lui, Kevin Haines, Alexander J Towbin","doi":"10.1148/ryai.250982","DOIUrl":"10.1148/ryai.250982","url":null,"abstract":"<p><p>Artificial intelligence (AI) has progressed from technical research to routine clinical use, reaching an inflection point where technological capabilities may exceed current regulatory and oversight frameworks. These systems are becoming more complex, progressing from narrow, task-specific algorithms to foundation models and early agentic prototypes. This progression has redistributed risk, responsibility, and clinical judgment, requiring radiologists and health care leaders to understand how policy choices affect patient safety and clinical innovation advancement. This special report provides a roadmap for aligning policy with clinical practice through a practical, lifecycle-based framework centered on patient safety. <i>Translational bialignment</i> is a concept that pairs regulatory science requirements (what AI systems should deliver to clinicians and patients) with implementation science capabilities (what institutions should provide for safe deployment of AI). This framework addresses the full AI lifecycle, from data stewardship and model development to validation, deployment, and monitoring, and articulates shared responsibilities for vendors, institutions, and clinicians grounded in trustworthy AI principles. The analysis focuses on U.S. regulatory frameworks, particularly Food and Drug Administration policies governing medical AI, with relevant highlights from international approaches. Concrete opportunities for radiologists to engage in policy formation, participate in oversight, and collaborate with industry and policymakers are provided to help shape a trustworthy and sustainable AI ecosystem. The alignment of policy, practice, and patient safety will enable medical AI to have a lasting impact on clinical care and public trust. The analysis and recommendations provided represent the authors' perspectives and do not necessarily reflect the official positions of the Radiological Society of North America. <b>Keywords:</b> Artificial Intelligence, Food and Drug Administration, Health Policy, Implementation Science, Large Language Models, Machine Learning, Patient Safety, Regulatory Science, Translational Bialignment © RSNA, 2026.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250982"},"PeriodicalIF":20.1,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13349408/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148035933","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"End-to-End Autonomous Quantification of Brain Aneurysm and Parent Artery Morphology at CT Angiography.","authors":"Samuel D Pettersson, Jean Filo, Paulina Skrzypkowska, Thomas B Fodor, Kamil Siedlecki, Peter Liaw, Ciprian N Ionita, Pratham B Bhatt, Jason Luo, Wolfgang Marik, Piotr Zieliński, Elad I Levy, Adnan H Siddiqui, Carlos E Baccin, Yu-Ming Chang, Christopher S Ogilvy","doi":"10.1148/ryai.251093","DOIUrl":"10.1148/ryai.251093","url":null,"abstract":"<p><p>Purpose To develop and validate an end-to-end autonomous platform for the quantification and visualization of brain aneurysm and parent artery morphology at CT angiography (CTA). Materials and Methods: A total of 2980 CTA scans performed between 2004 and 2025 containing 2585 aneurysms from 2980 patients obtained from five international high-volume stroke centers were included. The model was trained using expert hand-annotated vascular segmentations spanning the cervical internal carotid and vertebral arteries through the A4, M3, and P3 segments. Internal prospective and multicenter external testing assessed aneurysm detection performance and compared morphology measurements with expert-derived values obtained with digital subtraction angiography-verified CTA. Statistical analysis used paired <i>t</i> tests or Wilcoxon signed rank tests, with sensitivity, specificity, and 95% CIs reported. A public web-based platform was developed to allow further validation. Results Internal and multicenter external testing yielded a patient-level sensitivity of 87.9% (290 of 330; 95% CI: 83.6, 91.3) and specificity of 86.6% (395 of 456; 95% CI: 83.3, 89.4). Physician-performed morphology extraction required 24.3 minutes ± 6.6 per scan, whereas the model completed the task autonomously in 90 seconds ± 12 (<i>P</i> < .001). At the aggregate level, no statistically significant differences were observed for aneurysm volume, neck diameter, parent artery diameter, flow angle, aspect ratio, size ratio, height-width ratio, undulating index, ellipticity index, or nonsphericity index. Dome height (mean difference, 0.1 ± 1.1; <i>P</i> = .03) and surface area (mean difference, -5.3 mm² ± 29.5; <i>P</i> = .04) differed but had minimal effect sizes. Conclusion The developed autonomous system for rupture-related morphology metric acquisition substantially reduced physician workload. <b>Keywords:</b> Intracranial Aneurysm, CT Angiography, Artificial Intelligence, Morphology, Rupture Risk <i>Supplemental material is available for this article.</i> © RSNA, 2026 See also commentary by Maiter and Kular in this issue.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e251093"},"PeriodicalIF":20.1,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147975985","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Externally Tested AI Models for Malignancy Classification of Lung Nodules at CT: A Systematic Review and Meta-Analysis.","authors":"Oke Dimas Asmara, Eline G M Steenhuis, Kim de Jong, Anita Joseph, Marjan Timmer, Eric Daniel Tenda, E Christiaan Boerma, Marjolein A Heuvelmans, Wouter H van Geffen","doi":"10.1148/ryai.250331","DOIUrl":"10.1148/ryai.250331","url":null,"abstract":"<p><p>Purpose To evaluate the pooled diagnostic accuracy of externally tested artificial intelligence (AI) models for malignancy classification of lung nodules at chest CT. Materials and Methods A systematic search of PubMed, Embase, Web of Science, Cumulative Index to Nursing and Allied Health Literature, and the Cochrane Library was performed in March 2023 and updated on January 12, 2025, to identify studies evaluating AI models for malignancy classification of lung nodules at chest CT using pathology and/or at least 2-year follow-up as reference standards. Risk of bias was assessed using the Quality Assessment of Diagnostic Accuracy Studies-2 tool, and pooled sensitivity and specificity were estimated using bivariate random-effects models. Results Twenty-one studies including 7454 nodules were analyzed, with lung cancer prevalence ranging from 5.7% (17 of 297) to 91.5% (214 of 234). All models were based on deep learning; 17 of the 21 studies (81%) involved Asian populations, 16 (76%) used nonscreening populations, 14 (67%) reported two-dimensional or three-dimensional convolutional neural network (CNN) architectures, and eight (38%) specified predefined malignancy thresholds. High risk of bias was identified in five studies for patient selection and in two for index testing. Pooled sensitivity was 88%, pooled specificity was 75%, positive likelihood ratio was 3.55, negative likelihood ratio was 0.16, area under the receiver operating characteristic curve was 0.89, and the diagnostic odds ratio was 22.4. Heterogeneity was high (<i>I</i>² > 90%). Model architecture was associated with specificity, with higher values in studies reporting two-dimensional or three-dimensional CNNs compared with those without reported architecture (82%-83% vs 58%, <i>P</i> = .03; meta-regression <i>P</i> = .02); other subgroup analyses showed no evidence of differences. Conclusion Externally tested AI models demonstrated high sensitivity but moderate specificity for malignancy classification of lung nodules at chest CT, supporting a potential role in rule-out strategies. However, substantial heterogeneity, inconsistent reporting, and risk of bias limit interpretation. <b>Keywords:</b> Lung Cancer, Artificial Intelligence, Malignancy Classification, External Validation, Meta-Analysis <i>Supplemental material is available for this article.</i> © RSNA, 2026 See also commentary by Bressem and Kim in this issue.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250331"},"PeriodicalIF":20.1,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148151723","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Artificial Intelligence as a Triage Partner in Breast Cancer Screening.","authors":"Michael S Yao, Allison Chae","doi":"10.1148/ryai.260493","DOIUrl":"10.1148/ryai.260493","url":null,"abstract":"","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":"8 4","pages":"e260493"},"PeriodicalIF":20.1,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13430027/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148266412","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Artificial Intelligence for Digital Breast Tomosynthesis Screening with and without Prior Examinations in BreastScreen Norway.","authors":"Nataliia Moshina, Marthe Larsen, Åsne S Holen, Hildegunn S Aase, Kerstin Obst-Gleditsch, Hauke Bartsch, Leif Oltedal, Yan Chen, Solveig Hofvind","doi":"10.1148/ryai.250988","DOIUrl":"10.1148/ryai.250988","url":null,"abstract":"<p><p>Purpose To evaluate the performance of an artificial intelligence (AI) model for digital breast tomosynthesis (DBT) with and without prior screening examinations by using independent double reading by radiologists as the reference standard. Materials and Methods This retrospective study included women screened in BreastScreen Norway, 2018-2019. The AI model provided continuous malignancy risk and categorical scores (ie, 1-10) with and without the inclusion of prior examinations for 24 315 women with prior examinations. The area under the receiver operating characteristic curve (AUC) was calculated for AI using screening-detected and interval cancers as true-positives. The positive predictive value for a score of 10 was defined as the proportion of screening-detected cancers among all examinations scored 10. Results Among 30 724 DBT examinations of women (mean age ± SD, 61.3 years ± 5.0), the sensitivity of radiologists was 86.0% (95% CI: 81.3, 89.7). Among comparisons that included prior examinations, the AUC of the AI model was 0.93 (95% CI: 0.91, 0.95), with 81.0% (95% CI: 75.8, 85.3) scored 10 (screening-detected: 88.3% [95% CI: 83.4, 91.9] vs interval cancers: 36.1% [95% CI: 22.5, 52.4]). The positive predictive value was 10.2% (95% CI: 8.9, 11.6). Among cases without prior examinations, the AUC was 0.93 (95% CI: 0.91, 0.95; <i>P</i> = .13), with 82.2% (95% CI: 77.0, 86.4) of all cancer cases scored 10 (screening-detected: 88.7% [95% CI: 83.9, 92.3] vs interval cancers: 41.7% [95% CI: 27.1, 57.8]). The positive predictive value was 8.9% (95% CI: 7.8, 10.1). Conclusion The performance of the AI model was similar to that of double reading by radiologists. Although prior examinations provided limited additional benefit, they offered slightly improved specificities. <b>Keywords:</b> Digital Breast Tomosynthesis, Mammography, Breast Cancer, Screening <i>Supplemental material is available for this article.</i> © RSNA, 2026 See also commentary by Yao and Chae in this issue.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250988"},"PeriodicalIF":20.1,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147933985","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Cracking the Registration Conundrum in Breast MRI: Preserving the Tumor Signal to Reveal True Treatment Change.","authors":"Fan Zhang","doi":"10.1148/ryai.260465","DOIUrl":"10.1148/ryai.260465","url":null,"abstract":"","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":"8 4","pages":"e260465"},"PeriodicalIF":20.1,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148309692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"ReclAIm: A Multiagent Framework for Monitoring and Correcting Performance Decline in Medical Imaging AI.","authors":"Eleftherios Tzanis, Michail E Klontzas","doi":"10.1148/ryai.250923","DOIUrl":"10.1148/ryai.250923","url":null,"abstract":"<p><p>Purpose To develop and evaluate a multiagent framework (ReclAIm) for automated monitoring, detection, and correction of performance decline in medical image classification models. Materials and Methods ReclAIm is a large language model-based multiagent system that operates through natural language interaction. A master agent coordinating three task-specific agents executed performance evaluation and triggered fine-tuning when substantial performance declines were detected. The fine-tuning workflow incorporated data augmentation, class imbalance handling, and a parameter-anchoring regularization strategy to limit catastrophic forgetting. The system was benchmarked using multiple imaging datasets, including brain MRI, chest CT, and chest radiography, partitioned into model development, inference (performance monitoring), and fine-tuning subsets (60%:20%:20%). Results ReclAIm successfully orchestrated training, evaluation, and performance monitoring across all datasets. Performance discrepancies between test and inference data were detected in eight of 18 models, prompting fine-tuning workflows that reduced performance gaps. In cases with declines of up to 40.6% (cardiomegaly dataset, InceptionV3), fine-tuning restored performance metrics to within ±2% of baseline values. Conclusion ReclAIm provides a prototype framework for automated monitoring and targeted fine-tuning of medical image classification models, with a natural language interface designed to support accessibility in research and potential clinical applications. <b>Keywords:</b> Diagnosis, CT, MR Imaging, Conventional Radiography, Convolutional Neural Network, Agents, AI, Continual Learning, Medical Image Classification, Large Language Model, Fine-tuning, Postmarket Surveillance, Monitoring, Performance <i>Supplemental material is available for this article.</i> © RSNA, 2026.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250923"},"PeriodicalIF":20.1,"publicationDate":"2026-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148151733","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}