Radiology-Artificial Intelligence最新文献

Purpose To develop a digitized integrated feature-based interpretable machine learning classification model to accurately recognize complex thyroid nodules while efficiently diagnosing conventional thyroid nodules (thyroid nodules with typical benign or malignant ultrasound features). Materials and Methods Thyroid ultrasound images with pathologically confirmed nodules were retrospectively collected from seven medical centers in China (January 2011 to December 2021). An interpretable classification model consisting of two independent masks was developed and defined as UltraMC. The front-end network was trained to identify conventional thyroid nodules using four digitized features, and the back-end network collected nodules classified as benign in the previous framework for secondary analysis to clarify their final diagnosis. UltraMC performance was evaluated using accuracy, sensitivity, specificity and confusion matrices. Results The total dataset included 73826 patients with thyroid ultrasound images (mean age, 45.56 ± [SD] 11.21years; 54398 female). Diagnostic accuracy of the front-end network for detecting conventional thyroid nodules was 92.9% (13718/14765), and accuracy of the back-end network for classifying mummified thyroid nodules (MTNs) was 88.5% (652/737). The overall diagnostic accuracy of the ultrasound MTN classification model (UltraMC) was 91.8% (14228/15502). The areas under the receiver operating characteristic curve of the front-end network and UltraMC in identifying conventional thyroid nodules were 0.98 (95% CI 0.98-0.98) and 0.96 (95% CI 0.96-0.97), respectively. Conclusion The proposed two-layer interpretable classification model achieved high diagnostic accuracy for both conventional and mummified thyroid nodules. These findings demonstrate that digitized ultrasound features integrated into a white-box framework can effectively support classification of complex thyroid nodules. ©RSNA, 2026.

The RSNA Screening Mammography Cancer Detection Challenge dataset contains mammograms with corresponding metadata and pathological outcomes. ©RSNA, 2026.

Purpose To develop a deep learning algorithm to automatically assess the posterior fossa on first-trimester US screening scans and identify open spina bifida (OSB) and cystic posterior fossa (CPF) anomalies. Materials and Methods This is the retrospective part of an international study involving 10 fetal medicine centers. Normal and abnormal (OSB, CPF anomaly) midsagittal fetal brain US images acquired between 11 and 14 weeks of gestation (July 2009-January 2024) with confirmed diagnosis at follow-up were evaluated. Images were manually annotated to delineate the posterior fossa. The dataset was split into a training/validation (70%) and internal test (30%) set. Three convolutional neural networks were trained via threefold cross-validation on the training/validation set, with predictions on the internal test set obtained by ensemble averaging across folds. Model performance in detecting OSB and CPF anomalies was evaluated for the whole cohort and for fetuses with OSB or CPF anomalies separately. Results Images from 251 fetuses were analyzed (mean gestational age, 12.7±0.65 weeks; 150 normal, 101 abnormal: 43 OSB, 58 CPF anomalies). On the internal test, the MobileNetV3 Large Weights achieved the best performance (area under the receiver operating characteristic curve, 0.94 [95% CI: 0.88, 0.99]; accuracy, 88% (67/76); recall, 81% (25/31); specificity, 93% (42/45); precision, 89% (25/28); NPV, 88% (42/48); and F1-score, 0.85). OSB was classified more accurately (93% (52/56) vs 88% (57/65), P = .38) and with higher recall (91% (10/11) versus 75% (15/20), P = .38 although the difference was not significant. Conclusion MobileNetV3 Large Weights accurately assessed the fetal posterior fossa between 11 and 14 weeks of gestation, distinguishing normal images from those showing OSB or CPF anomalies. ©RSNA, 2026.

The introduction of foundational models, specifically large language models, has promised a health care transformation. However, the field is rapidly evolving toward autonomous agent systems, defined as AI entities that perceive and react to their environment to achieve specific goals-representing a paradigm shift from passive information retrieval to proactive, goal-oriented clinical assistance. Agentic AI systems transcend static knowledge limitations through core capabilities including persistent memory systems that maintain context across patient encounters, knowledge retrieval tools connecting to medical repositories through retrieval-augmented generation techniques, and computer use functionality enabling navigation of clinical software interfaces. Agentic workflows introduce sophisticated coordination mechanisms including hierarchical, collaborative, and sequential patterns demonstrating superior performance compared with single-agent approaches. Multiagent systems can autonomously coordinate entire clinical workflows across the entire radiology lifecycle, from preacquisition protocol optimization through initial image analysis, specialized tool deployment, and preliminary report generation. However, successful clinical deployment requires systematic consideration of complexity thresholds, economic sustainability, cybersecurity frameworks, bias mitigation strategies, and appropriate governance structures. Critical challenges include managing the probabilistic nature of underlying models within deterministic clinical workflows, ensuring adequate human supervision, and preventing overcomplication of established processes. A structured four-phase implementation roadmap addresses these considerations through incremental progression from low-risk automation to comprehensive workflow orchestration while maintaining rigorous safety standards. As foundation models advance and interoperability standards mature, agentic AI will reshape radiology practice paradigms. Success depends on resolving stakeholder responsibility questions while orchestrating technological capabilities with clinical accountability, ensuring autonomous systems augment rather than replace professional judgment in pursuit of improved patient outcomes. ©RSNA, 2026.

The RSNA Lumbar Degenerative Imaging Spine Classification dataset is the largest publicly available adult MRI lumbar spine dataset for degenerative disease. The dataset includes multisequence, multiplanar MRIs from 2,697 patients through contributions from 8 institutions across 6 countries and 5 continents. ©RSNA, 2026.

Purpose To build extra-axial cerebrospinal fluid (eaCSF) growth charts that define key diagnostic criteria for benign enlargement of the subarachnoid space (BESS) by providing an age-related reference benchmark to aid in assessing atypical eaCSF development. Materials and Methods In this retrospective study, T1-weighted MRI scans from patients who underwent imaging at a pediatric health care system between January 2004 and December 2023 were accessed to form a clinical control group. Nine scans from patients diagnosed with BESS by a board-certified pediatric neuroradiologist were also reviewed. T1-weighted scans were segmented into various tissue types, including eaCSF. Growth charts of eaCSF were modeled using the clinical control group. The results of patients with confirmed BESS were then benchmarked against these charts to test the performance of the eaCSF growth charts. Generalized additive models of location, scale, and shape were used. Results The eaCSF measurements were obtained for 1205 patients (619 female; age range, 0.19-19.6 years). Measurements show that eaCSF evolved dynamically with age, steadily decreasing from birth to 2 years, then trending upward in childhood. Seven of the nine patients with a clinical diagnosis of BESS had eaCSF measurements above the 97.5th percentile for at least one measurement. Percentile scores distinguished patients with BESS from controls with areas under the receiver operating characteristic curve of greater than 0.95. Conclusion MRI-derived eaCSF measurements evolved dynamically throughout early life. Patients with atypical CSF development could be differentiated from clinical controls using computational measurements paired with normative modeling. Keywords: MRI, Brain/Brain Stem, Pediatrics, Benign Enlargement of Subarachnoid Space Supplemental material is available for this article. © The Author(s) 2025. Published by the Radiological Society of North America under a CC BY-NC-ND license.

Purpose To develop and validate a deep learning (DL) model for automated assessment of coronary stenosis in vessels with heavily calcified plaques at coronary CT angiography (CCTA), using quantitative coronary angiography as the reference standard. Materials and Methods A total of 10 101 CCTA examinations (June 2017-December 2020) from three tertiary hospitals in China were retrospectively collected for DL model development. External testing dataset 1 included 442 CCTA examinations (Agatston score > 300) from two independent hospitals (January 2021-May 2022) for performance evaluation. The separate external testing dataset 2 of 120 CCTA examinations was used for a reader study assessing whether DL assistance improved diagnostic accuracy among junior, attending, and senior radiologists. External testing dataset 3 included 150 prospectively collected CCTA examinations (June-July 2023) that were analyzed to compare model performance against clinical reports, simulating real-world deployment. Model diagnostic performance was assessed using receiver operating characteristic analysis, with quantitative coronary angiography as the reference. Results In external testing dataset 1, specificities for detecting 50% or more stenosis were 78%, 72%, and 48% and the areas under the receiver operating characteristic curve (AUC) were 0.89, 0.90, and 0.87 at the segment, vessel, and patient levels, respectively. In external testing dataset 2, DL assistance improved radiologist specificity by 7%-11% (P < .001) with improving AUC and increased interreader agreement (Δκ = 0.155-0.228; P < .05). In external testing dataset 3, the model demonstrated 53% specificity and a higher AUC versus clinical reports (0.91 vs 0.76; P < .001). Conclusion The proposed DL model accurately detected coronary stenosis of heavily calcified plaques at CCTA and improved diagnostic performance of radiologists. Keywords: CT Angiography, Cardiac, Heart, Arteriosclerosis, Calcifications, Calculi, Quantification, Diagnosis Supplemental material is available for this article. © The Author(s) 2025. Published by the Radiological Society of North America under a CC BY 4.0 license. See also commentary by Maiter and Alabed in this issue.

Purpose To calculate new pediatric age-specific normative values and percentiles for kidney length and volume through the use of a natural language processing (NLP) model. Materials and Methods In this cross-sectional study, 24 664 US reports from 18 769 children (birth to 18 years) conducted between January 2012 and December 2022 at a tertiary children's hospital in the northeastern United States were analyzed with an NLP model. Anthropometric data from 12 595 children were used to evaluate the effect of sex and body measurements on kidney length and volume through age-adjusted quantile regression models. Age-related percentiles were established after calibration, using the lambda-mu-sigma (LMS) method by age (year), with detailed subcategories for children younger than 1 year. Volume percentiles by body surface area were also generated using the LMS method. Results A total of 24 664 reports from 18 769 children were included (median age, 7 years [IQR, 11 years]; 10 134 female children). Normative value analysis showed that kidney growth was more pronounced in the 1st year of life (1.8-cm increase in length and 16.9-cm³ increase in volume). The large sample size resulted in standard errors that were 10%-30% less than previous normative values. Quantile regression models showed that body surface area was a better predictor of kidney volume than was age (R¹ = 0.57 [P < .001] vs 0.48 [P < .001]). Conclusion New LMS percentiles for kidney size were established using data from a large pediatric sample. Keywords: Kidney, Natural Language Processing, Pediatrics, Ultrasound Supplemental material is available for this article. © The Author(s) 2025. Published by the Radiological Society of North America under a CC BY 4.0 license See also the commentary by Sihlahla in this issue.

Purpose To develop a foundational pretraining method for digital mammography that extracts fine-grained visual-language representations from images and reports in label-limited settings. Materials and Methods A multiview mammogram-report pretraining framework for automated breast cancer analysis was developed using retrospectively collected data from January 2010 to December 2020. This framework provides visual explanations of the model's learning, allowing researchers to "visualize what you learn." The abnormality-aware technique was tailored to mammogram characteristics of dense fibroglandular tissue. The proposed framework was evaluated on downstream tasks from four external medical centers, involving label-efficient abnormality recognition in mammograms, including malignancy classification, segmentation, and localization. Statistical analyses were performed using the DeLong test and paired t test for area under the receiver operating characteristic curve and Dice scores, respectively. Results The visualization results, including abnormality-enhanced mammograms and abnormality-awareness maps, could explain that the developed model successfully captures relationships between multiview mammograms and corresponding reports. This reduces the false positives for breast cancer by 37% and enables zero-shot abnormality segmentation. Furthermore, the developed model consistently outperformed existing approaches in fine-tuning for both malignancy classification (area under the receiver operating characteristic curve, INbreast: 0.90 vs 0.78 [P < .001]; Curated Breast Imaging Subset of Digital Database for Screening Mammography [CBIS-DDSM]: 0.85 vs 0.79 [P < .01]; Chinese Mammography Database: 0.85 vs 0.78 [P < .001]; and Cohort of Screen-age Women-Case Control: 0.86 vs 0.77 [P < .001]) and segmentation and localization (Dice score, INbreast: 0.75 vs 0.63 [P < .001]; CBIS-DDSM: 0.76 vs 0.61 [P < .001]). Conclusion The proposed framework enhances interpretability and fine-grained multimodal foundational learning for multiview mammograms and reports. Keywords: Mammography, Breast, Segmentation, Feature Detection, Quantification, Diagnosis, Translation, Transfer Learning, Unsupervised Learning, Breast Cancer, Representation Learning, Visual-Language Foundation Model, Explainable AI Supplemental material is available for this article. © RSNA, 2025.