Academic Radiology最新文献

{"title":"Mammography and Ultrasonography for Symptomatic Male Breast Lesions: A PRISMA-DTA Systematic Review and Meta-analysis.","authors":"Chao Wang, Xiaowen Liu, Chengxin Yan, Xiangguo Dang","doi":"10.1016/j.acra.2026.09.060","DOIUrl":"https://doi.org/10.1016/j.acra.2026.09.060","url":null,"abstract":"<p><strong>Rationale and objectives: </strong>Male breast disease is uncommon, but symptomatic men require reliable differentiation between benign conditions and lesions warranting biopsy or oncologic management. This PRISMA-DTA (Preferred Reporting Items for a Systematic Review and Meta-analysis of Diagnostic Test Accuracy Studies) systematic review and meta-analysis evaluated the diagnostic accuracy, clinical utility, and evidentiary limitations of mammography and ultrasonography for malignant versus benign male breast lesions.</p><p><strong>Materials and methods: </strong>PubMed, Embase, Web of Science Core Collection, and Scopus were searched from inception to May 30, 2026. The protocol was registered in PROSPERO before journal submission (CRD420261429183). Eligible reports evaluated mammography or ultrasonography in male breast disease and provided extractable 2 × 2 diagnostic data. Methodological quality was assessed using Quality Assessment of Diagnostic Accuracy Studies-2. Diagnostic accuracy was synthesized separately for each modality using bivariate random-effects modeling and summary receiver operating characteristic analysis.</p><p><strong>Results: </strong>Thirteen eligible reports were included. Seven mammography datasets and five ultrasonography datasets were available for quantitative synthesis. Mammography showed a pooled sensitivity of 0.840 (95% confidence interval [CI], 0.705-0.921) and a specificity of 0.951 (95% CI, 0.884-0.980), whereas ultrasonography showed a pooled sensitivity of 0.881 (95% CI, 0.775-0.941) and a specificity of 0.958 (95% CI, 0.869-0.987). Although pooled specificity estimates were high, their magnitude may have been influenced by heterogeneous diagnostic thresholds, differential verification, variable follow-up, and sparse malignant case counts. Nine of 13 reports were judged to have an overall high risk of bias, and consistent paired head-to-head data were lacking.</p><p><strong>Conclusion: </strong>Current evidence supports mammography and ultrasonography as complementary tools in risk-stratified assessment of symptomatic male breast lesions. Pooled estimates should be interpreted cautiously because the evidence base is small and methodologically limited. Current data do not establish direct modality superiority, do not justify biopsy avoidance when clinical-imaging discordance persists, and do not determine the incremental value of routine combined imaging.</p>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149197638","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":"Corrigendum to \"Application Effectiveness Analysis of Artificial Intelligence-assisted Teaching in Standardized Training for Medical Imaging Residents\" (Acad Radiol 33 [2026] 3619-3627).","authors":"Lu Zhang, Qiuying Chen, Zhang Bin, Jin Fang, Xiaokai Mo, Shuixing Zhang","doi":"10.1016/j.acra.2026.09.062","DOIUrl":"https://doi.org/10.1016/j.acra.2026.09.062","url":null,"abstract":"","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149197648","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":"Comparative Evaluation of Open-Weight Large Language Models for Pathological Brain MRI Reports Classification: From Prompts to a Fine-Tuned Model.","authors":"Claudio Demaria, Claudio Crema, Mario Verdicchio, Valentina Brancato, Vincenzo Alfano, Luca Basso, Giuseppe Pontillo, Damiano Archetti, Silvia Fostinelli, Silvia De Francesco, Michela Pievani, Cesare M Baronio, Alberto Boccali, Nicola Toschi, Arturo Brunetti, Alberto Redolfi, Marco Aiello","doi":"10.1016/j.acra.2026.09.053","DOIUrl":"https://doi.org/10.1016/j.acra.2026.09.053","url":null,"abstract":"<p><strong>Rationale and objectives: </strong>Automated semantic classification of radiology reports can enhance clinical workflows and research, but often relies on language- and task-specific models, limiting scalability. This study aims to evaluate and improve the 0-shot baseline performance of open-weight large language models (LLMs) in the semantic classification of free-text magnetic resonance imaging (MRI) reports.</p><p><strong>Materials and methods: </strong>A corpus of 1385 reports from a single neuroradiology center was used to assess four LLMs (GPT-OSS-20B, Magistral-Small-2509, Gemma-4-31B, Qwen3.5-35B-A3B) of comparable scale (∼20-35B parameters). The brain MRI reports were labeled as Pathological or Non-Pathological, with the Pathological ones further classified into five diagnostic categories (Neoformation, Vascular, Inflammation/Infection, Neurodegenerative, Other). The LLMs were evaluated using zero- and few-shot prompting with majority-vote aggregation. Quantized Low-Rank Adaptation (QLoRA) fine-tuning was then applied to the best-performing model across multiple configurations. Performance was assessed using Macro F1, Micro F1, Precision, Recall, and Exact Match Accuracy.</p><p><strong>Results: </strong>The binary classification task achieved a Macro F1 score of 0.84 on the best-performing Qwen3.5-35B-A3B model, while multilabel classification reached 0.81. Majority voting (MV) outperformed individual-set predictions. QLoRA fine-tuning yielded a Macro F1 score of 0.83 on the multilabel task; the improvement over the 0-shot baseline was statistically supported for Macro F1, Micro F1, Macro Precision, and Exact Match, whose bootstrap 95% confidence intervals excluded zero, but not for Macro Recall. Fine-tuning and few-shot prompting strategies are complementary, with neither consistently outperforming the other. Error analysis showed that the models struggled with borderline annotation cases and often missed secondary labels.</p><p><strong>Conclusion: </strong>Open-weight LLMs demonstrate strong capability for neuroradiology report classification without extensive task-specific training. Fine-tuning further improves performance, supporting integration into scalable semantic radiology data pipelines. However, the present evidence comes from a single Italian center without external validation, and the fine-tuning gain was estimated on a small held-out test set, for which the Macro F1 improvement, though statistically significant, has a confidence interval with a lower bound close to zero. In this light, the approach is best positioned as an internal indexing and semantic-retrieval tool, supporting faster report retrieval, structured cohort identification, internal audit, and scalable phenotyping for neuroradiology research and data warehousing, as well as efficient active-learning strategies, with broader generalizability requiring additional external multicentre validation before clinical deployment.</p>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149185829","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":"Development and Prospective Validation of an Interpretable Machine Learning Model Integrating Cardiac MRI Extracellular Volume for 1-Year MACE Prediction Post-STEMI.","authors":"Zhihao Zhao, Zeqing Zhang, Yu Yang, Jiafeng Lin","doi":"10.1016/j.acra.2026.09.054","DOIUrl":"https://doi.org/10.1016/j.acra.2026.09.054","url":null,"abstract":"<p><strong>Rationale and objectives: </strong>Extracellular volume (ECV) quantified by cardiac magnetic resonance (CMR) is a key indicator of myocardial fibrosis and strongly linked to outcomes after ST-segment elevation myocardial infarction (STEMI). This research focused on the development and validation of an interpretable machine learning (ML) model that incorporates ECV to predict 1-year major adverse cardiac events (MACE) in patients with STEMI following percutaneous coronary intervention.</p><p><strong>Materials and methods: </strong>MACE prediction models were developed using clinical variables combined with CMR-ECV parameters and validated in a prospective cohort. Feature selection was conducted through Boruta algorithms and least absolute shrinkage and selection operator algorithms, and eight ML algorithms were trained. Model performance was assessed by discrimination and calibration. The top performer was interpreted by Shapley Additive exPlanations (SHAP) analysis. An interactive online nomogram was deployed to facilitate personalized perioperative risk management.</p><p><strong>Results: </strong>Among 673 STEMI patients (derivation: n = 398; prospective validation: n = 275), five key predictors were identified: N-terminal pro-B-type natriuretic peptide (NT-proBNP), left ventricular ejection fraction (LVEF), myocardial infarction ECV (MI-ECV), age, and creatine kinase-MB. The random forest model achieved the highest discriminative performance (area under the receiver operating characteristic curve 0.825, 95% confidence interval: 0.763-0.876). SHAP analysis identified NT-proBNP as the most impactful feature, with LVEF class and MI-ECV ranking next.</p><p><strong>Conclusion: </strong>Integration of CMR-derived ECV into an interpretable machine learning framework enables accurate 1-year MACE risk prediction following STEMI. This approach combines robust discriminative performance with clinical transparency through SHAP-based explanations, offering a proof-of-concept tool for precision cardiovascular care.</p>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149185831","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":"Implementation of a Structured Peer Learning Program and Missed Critical Findings in a Large Integrated Radiology Practice: A Baseline Retrospective Analysis.","authors":"Zachary Ballenger, Thomas J Ward, Jason C Theadore, Ibrahim F Koury, Asif Bhimani, RoyLynn Taylor, Vanessa Loriga, Aramis Segarra, Felix M Gonzalez","doi":"10.1016/j.acra.2026.09.050","DOIUrl":"https://doi.org/10.1016/j.acra.2026.09.050","url":null,"abstract":"<p><strong>Rationale and objectives: </strong>The American College of Radiology (ACR) defines peer learning as nonpunitive, education-centered case review, separated from individual performance evaluation. We describe a program designed against ACR criteria in a large integrated practice and report first-cycle metrics, including missed critical findings-a narrow subset of major misses.</p><p><strong>Materials and methods: </strong>This retrospective analysis spanned 283 radiologists across more than 100 sites between March 30 and May 14, 2026. Cases entered through voluntary worklist submission, risk-event referral, and departmental escalation, and were adjudicated by a fellowship-trained coordinator. A missed critical finding (MCF) required both a finding on the institutional critical results list (provider notification within 30 minutes) and a 4B classification, with agreement between two independent reviewers. Exact (Poisson) 95% confidence intervals (CI) were calculated.</p><p><strong>Results: </strong>Of 702 submissions, 489 (69.7%) were Learning Opportunities, 187 (26.6%) Great Calls, and 26 (3.7%) Agrees. Of 283 radiologists, 182 (64.3%) submitted at least one case, and 213 (75.3%) received feedback; two conferences were held. Of 17 candidates, reviewers concurred on 7 at initial review; the 10 discordant were resolved by consensus, none meeting criteria. Across 409,122 interpreted studies, the confirmed MCF rate was 0.017 per 1000 (95% CI 0.007-0.035). Seven radiologists (2.5%) had one event; none had more. No confirmed event had been independently filed as a safety event.</p><p><strong>Conclusion: </strong>The MCF rate reflects a narrow, time-critical subset ascertained with unmeasured sensitivity against a diluted denominator. It is a conservative internal baseline, not a measure of diagnostic accuracy or individual performance.</p>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149171559","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":"Automated Structured Reporting of Neurodegenerative Imaging Markers Using Quantitative MRI and Large Language Models.","authors":"Ji Su Ko, Wooseok Jung, Pae Sun Suh, Hwon Heo, Woo Hyun Shim, Dong-Hee Kim, Seung Hyun Lee, Mijoo Kim, Nahyuk Lee, Saehyun Kim, Sang Joon Kim, Chong Hyun Suh","doi":"10.1016/j.acra.2026.09.056","DOIUrl":"https://doi.org/10.1016/j.acra.2026.09.056","url":null,"abstract":"<p><strong>Rationale and objectives: </strong>To evaluate the feasibility of a quantitative magnetic resonance imaging (MRI)-to-structured reporting pipeline that integrates quantitative MRI software with large language models (LLMs) for structured drafting of neurodegenerative imaging markers and generating clinician-facing report conclusions.</p><p><strong>Materials and methods: </strong>A total of 192 patients with 3D T1 and 2D or 3D FLAIR MRI images from the Alzheimer's Disease Neuroimaging Initiative database were randomly selected. A deep-learning model generated structured report findings based on nine indices related to brain atrophy, white matter hyperintensities, and cerebrospinal fluid (CSF) space abnormalities, and compared with neuroradiologist-generated reference reports. Four LLMs (GPT-OSS 20B, MedGemma 4B, Phi-3, and Llama 3.1 8B) generated clinician-facing conclusions from the structured findings. Generated conclusions were evaluated using a five-point Likert scale across four metrics (accuracy, relevance, clarity, and satisfaction). Output errors were assessed across four categories: language, omission, addition, and numeric errors.</p><p><strong>Results: </strong>Structured report findings demonstrated high accuracy for CSF space analysis (77.6%-97.9%). Performance for ordinal visual scales was lower under strict matching (27.1%-57.8%) but improved with a one-grade tolerance (91.7%-97.4%). Among LLMs, GPT achieved the highest scores across most metrics (median score 5) and exhibited the lowest output error rates. Smaller models (Phi-3 and Llama 3.1 8B) showed overall lower performance, with more language, omission, and addition errors. MedGemma demonstrated intermediate performance, but the highest rate of numeric errors.</p><p><strong>Conclusion: </strong>Our quantitative MRI-to-structured reporting pipeline, integrating quantitative MRI software with LLM-based conclusion generation, might be feasible for automated structured drafting of neurodegenerative imaging markers.</p>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149169330","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":"Comparison of Integrated Versus Sequential Artificial Intelligence-Based Computer-Aided Diagnosis Systems in Breast Ultrasound: A Prospective Pilot Study on Examination Time, Reproducibility, and Diagnostic Performance.","authors":"Jiyoon Ban, Haejung Kim, Sun-Young Baek, Boo-Kyung Han, Eun Sook Ko, Ji Soo Choi, Jeongmin Lee, Myoung Kyoung Kim, Eun Young Ko","doi":"10.1016/j.acra.2026.08.038","DOIUrl":"https://doi.org/10.1016/j.acra.2026.08.038","url":null,"abstract":"<p><strong>Rationale and objectives: </strong>To compare integrated and sequential artificial intelligence-based computer-aided diagnosis (AI-CAD) systems in terms of examination time, reproducibility, and diagnostic performance in breast ultrasound (US).</p><p><strong>Materials and methods: </strong>This prospective single-center study included women who underwent breast US at a breast imaging center between February and March 2025. Two real-time AI-CAD systems were evaluated: an integrated system providing simultaneous lesion detection and diagnosis during scanning, and a sequential system in which lesion detection and diagnostic assessment were performed separately. Examination time was assessed in patients who underwent paired breast US examinations with both AI-CAD systems during the same visit, with randomized examination order. Reproducibility was assessed by applying both systems twice to the same breast lesions. Diagnostic performance was evaluated using sensitivity, specificity, and area under the receiver operating characteristic curve (AUC).</p><p><strong>Results: </strong>Examination time analysis included 97 patients and showed longer examination time with sequential AI-CAD than with integrated AI-CAD (P < .001), regardless of examination order. Reproducibility analysis included 145 patients with 215 lesions and showed almost perfect agreement between the first and second acquisitions for both systems (integrated vs. sequential: κ = 0.81 vs. 0.83). For the comparison between the first and second acquisitions, κ values for the integrated versus sequential systems were 0.92 versus 0.76 for malignant lesions, 0.79 versus 0.84 for mass lesions, and 1.00 versus 0.75 for nonmass lesions. However, agreement patterns varied across other scan-plane comparisons. Diagnostic performance analysis included 242 patients. Integrated AI-CAD showed higher sensitivity (integrated vs. sequential: 87.6% vs. 76.9%; P < .001), whereas sequential AI-CAD showed higher specificity (integrated vs. sequential: 74.2% vs. 90.0%; P < .001). AUC did not differ between the two systems.</p><p><strong>Conclusion: </strong>Integrated and sequential AI-CAD systems showed complementary strengths in breast US. Integrated AI-CAD favored shorter examination time and higher sensitivity, whereas sequential AI-CAD favored higher specificity, with reproducibility differing according to lesion type.</p>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149185840","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":"Multicenter Comparative Validation of an nnU-Net-Based Automated CT Radiomics Workflow for Acute Pancreatitis Severity Prediction.","authors":"Boyao Hou, Wen Feng, Wenxue Li, Wenjing He, Junxian Gu, Yaxin Deng, Hong Deng","doi":"10.1016/j.acra.2026.09.040","DOIUrl":"https://doi.org/10.1016/j.acra.2026.09.040","url":null,"abstract":"<p><strong>Rationale and objectives: </strong>Computed tomography radiomics has shown potential for predicting acute pancreatitis (AP) severity, but conventional workflows depend on labor-intensive expert pancreatic segmentation. This study evaluated a no-new-Net (nnU-Net)-based automated segmentation-radiomics workflow against an expert manual-segmentation workflow, using AP severity classification as the downstream task, and assessed the automated workflow externally.</p><p><strong>Materials and methods: </strong>This retrospective multicenter study included patients with AP from Center A (n = 214), Center B (n = 56), and Center C (n = 42). Expert manual pancreatic masks served as reference annotations for nnU-Net training and as the comparator for the radiomics workflow. After selecting the optimal nnU-Net configuration, radiomics features were extracted separately from manual and automated regions of interest, and AP severity classifiers were developed using the same analytical framework. Discrimination between the two workflows was compared in the independent internal test cohort using the area under the curve (AUC) and the paired DeLong test, with Benjamini-Hochberg false discovery rate (FDR) and Bonferroni corrections applied to P values.</p><p><strong>Results: </strong>In the internal test cohort, the Light Gradient Boosting Machine classifier achieved AUCs of 0.799 (95% confidence interval (CI), 0.683-0.915) and 0.733 (95% CI, 0.593-0.873) with manual and automated segmentations, respectively. The absolute AUC difference was 0.066, and the paired DeLong comparison was not statistically significant after FDR correction (FDR-adjusted p = 0.360; Z = 1.1755, unadjusted p = 0.2398). The automated workflow achieved AUCs of 0.784 (95% CI, 0.653-0.915) in Center B and 0.715 (95% CI, 0.559-0.871) in Center C. The selected 2D nnU-Net achieved a mean Dice coefficient of 0.798 during five-fold cross-validation and 0.726 in the independent internal test cohort.</p><p><strong>Conclusion: </strong>The nnU-Net-based automated segmentation-radiomics workflow showed no significant difference in discrimination compared with the expert manual workflow in the internal test cohort. It also maintained moderate discrimination across two external cohorts. These findings support the feasibility of reducing dependence on manual pancreatic segmentation, although the present analysis does not establish formal equivalence or interchangeability.</p>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149185833","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":"Objective Structured Clinical Examinations in Radiology and Medical Imaging Education: Educational Effectiveness, Validity, and Implementation-A Systematic Review.","authors":"Yushi Jia, Lusi Liu","doi":"10.1016/j.acra.2026.09.044","DOIUrl":"https://doi.org/10.1016/j.acra.2026.09.044","url":null,"abstract":"<p><strong>Rationale and objectives: </strong>To synthesize evidence on objective structured clinical examinations (OSCEs) in radiology and medical imaging education, focusing on educational outcomes, validity evidence, reliability, scoring, feasibility, and implementation consequences.</p><p><strong>Materials and methods: </strong>PubMed/MEDLINE, ERIC, Embase, and Web of Science were searched from inception to July 20, 2026. Eligibility was limited to radiology, radiography, nuclear medicine, sonography, and medical student imaging curricula; clinician bedside POCUS courses outside radiology or medical imaging education were excluded. Educational-effect studies were appraised with MERSQI. Measurement evidence was appraised across seven validity and implementation domains. Meta-analysis was considered only for sufficiently comparable independent studies.</p><p><strong>Results: </strong>Forty-three reports met the narrowed scope: 22 measurement/implementation reports and 21 educational-effect reports. Educational studies comprised six randomized studies, nine nonrandomized comparisons, one pre-post study, and five post-course or implementation studies. Seven reports favored the intervention, one met a noninferiority criterion, one was mixed, six were uncertain, and six were not estimable as comparative effects. MERSQI scores ranged from 10.5 to 14.0. Feasibility was documented in all measurement reports, but internal-structure or reliability evidence was absent in ten, and consequence evidence was strong in only one.</p><p><strong>Conclusion: </strong>Imaging-focused OSCEs are feasible and useful for local assessment in radiology and medical imaging education, but evidence remains heterogeneous and incompletely supported by reliability, response-process, and consequence data.</p>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149185853","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":"Radhawk: An AI-driven Recommender for Pediatric Radiology Education.","authors":"Julian Lopez-Rippe, Manasa Reddy, Michael Welsh, Jennifer Vaughn, Raisa Amiruddin, Avin Khera, Ami Gokli, Michael Francavilla, Janet R Reid","doi":"10.1016/j.acra.2026.09.067","DOIUrl":"https://doi.org/10.1016/j.acra.2026.09.067","url":null,"abstract":"<p><strong>Rationale and objectives: </strong>The unprecedented growth of technology and unfiltered information has led to cognitive overload and educational gaps in radiology. Radiologists and trainees often require \"just-in-time\" educational content while interpreting unfamiliar or infrequently encountered examinations, yet identifying reliable and appropriately targeted resources within the time constraints of the reading room can be challenging. This need is particularly relevant in pediatric radiology, where normal imaging appearances vary with age and development. Radhawk was developed as a scalable, artificial intelligence-driven intelligent tutoring system that delivers personalized, real-time educational resources to learners and radiologists, creating a precision curriculum tailored to each user's needs and clinical context.</p><p><strong>Materials and methods: </strong>Radhawk integrates with existing picture archiving and communication systems and a custom learning management system (RADiology's Intelligent Adaptive Learning). It utilizes natural language processing and machine learning algorithms to analyze imaging study information and provides results based on user preferences. The system's scalable architecture is built on Amazon Web Services, ensuring adaptability across multiple hospital systems.</p><p><strong>Results: </strong>Radhawk successfully provides tailored recommendations for educational materials, including interactive courses, anatomical references, and case studies. Key features include: (1) Generation of personalized curricula based on learner performance and needs, (2) Presentation of age and gender-matched \"normal\" images for comparison, (3) Creation of user-specific heat maps to identify strengths and areas for improvement, and (4) Incorporation of expert subspecialist preferences to inform user choices.</p><p><strong>Conclusion: </strong>Radhawk replaces generic web searches with a knowledge-focused approach, providing instant access to high-quality, relevant resources. This precision education solution has the potential to improve diagnostic accuracy, enhance learning experiences, and address the shortage of specialized pediatric radiologists. Future plans include external testing across multiple pediatric radiology departments to validate Radhawk's scalability, acceptance, and value in real-world settings.</p>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2026-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149166880","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}