Radiology-Artificial Intelligence最新文献

{"title":"Adaptive Dual-Task Deep Learning for Automated Thyroid Cancer Triaging at Screening US.","authors":"Shaohong Wu, Ming-De Li, Wen-Juan Tong, Yihao Liu, Rui Cui, Jinbo Hu, Mei-Qing Cheng, Wei-Ping Ke, Xinxin Lin, Jia-Yi Lv, Longzhong Liu, Jie Ren, Guangjian Liu, Hong Yang, Wei Wang","doi":"10.1148/ryai.240271","DOIUrl":"https://doi.org/10.1148/ryai.240271","url":null,"abstract":"<p><p><i>\"Just Accepted\" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content.</i> Purpose To develop an adaptive dual-task deep learning model (ThyNet-S) for detection and classification of thyroid lesions at US screening. Materials and Methods The retrospective study used a multicenter dataset comprising 35008 thyroid US images of 23294 individual examinations (mean age, 40.4 years ± 13.1[SD], 17587 female) from 7 medical centers during January 2009 and December 2021. Of these, 29004 images were used for model development and 6004 images for validation. The model determined cancer risk for each image and automatically triaged images with normal thyroid and benign nodules by dynamically integrating lesion detection through pixel-level feature analysis and lesion classification through deep semantic features analysis. Diagnostic performance of screening assisted by the model (ThyNet-S triaged screening) and traditional screening (radiologists alone) was assessed by comparing sensitivity, specificity, accuracy and AUC using McNemar's test and Delong test. The influence of ThyNet-S on radiologist workload and clinical decision-making was also assessed. Results ThyNet-S-assisted triaged screening achieved higher AUC than original screening in six senior and six junior radiologists (0.93 versus 0.91, and 0.92 versus 0.88, respectively, all <i>P</i> < .001). The model improved sensitivity for junior radiologists (88.2% versus 86.8%, <i>P</i> <.001). Notably, the model reduced radiologists' workload by triaging 60.4% of cases as not potentially malignant, which did not require further interpretation. The model simultaneously decreased unnecessary fine needle aspiration rate from 38.7% to 14.9% and 11.5% when used independently or in combination with Thyroid Imaging Reporting and Data System, respectively. Conclusion ThyNet-S improved efficiency of thyroid cancer screening and optimized clinical decision-making. ©RSNA, 2025.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e240271"},"PeriodicalIF":8.1,"publicationDate":"2025-04-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143812639","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":"Predicting Major Adverse Cardiac Events Using Deep Learning-based Coronary Artery Disease Analysis at CT Angiography.","authors":"Jin Young Kim, Kye Ho Lee, Ji Won Lee, Jiyong Park, Jinho Park, Pan Ki Kim, Kyunghwa Han, Song-Ee Baek, Dong Jin Im, Byoung Wook Choi, Jin Hur","doi":"10.1148/ryai.240459","DOIUrl":"https://doi.org/10.1148/ryai.240459","url":null,"abstract":"<p><p><i>\"Just Accepted\" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content.</i> Purpose To evaluate the predictive value of deep learning (DL)-based coronary artery disease (CAD) extent analysis for major adverse cardiac events (MACEs) in patients with acute chest pain presenting to the emergency department (ED). Materials and Methods This retrospective multicenter observational study included consecutive patients with acute chest pain who underwent coronary CT angiography (CCTA) at three institutional EDs from January 2018 to December 2022. Patients were classified as having no CAD, nonobstructive CAD, or obstructive CAD using a DL model. The primary outcome was MACEs during follow-up, defined as a composite of cardiac death, nonfatal myocardial infarction, and hospitalization for unstable angina. Cox proportional hazards regression models were used to evaluate the predictors of MACEs. Results The study included 408 patients (224 male; mean age, 59.4 ± 14.6 years). The DL model classified 162 (39.7%) patients as having no CAD, 94 (23%) as having nonobstructive CAD, and 152 (37.3%) as having obstructive CAD. Sixty-three (15.4%) patients experienced MACEs during follow-up. Patients with MACEs had a higher prevalence of obstructive CAD than those without (<i>P</i> < .001). In multivariate analysis model 1 (clinical risk factors), dyslipidemia (Hazard ratio [HR], 2.15 and elevated Troponin-T (HR 2.13) predicted MACEs (all <i>P</i> < .05). In model 2 (clinical risk factors + DL-based CAD extent), obstructive CAD detected by the DL model was the most significant independent predictor of MACEs (HR, 88.07, <i>P</i> < .001). Harrell's C-statistic showed that DL-based CAD extent enhanced the risk stratification beyond clinical risk factors (Harrell's C-statistics: 0.94 versus 0.80, <i>P</i> < .001). Conclusion DL-based detection of obstructive CAD demonstrated stronger predictive value than clinical risk factors for MACEs in patients with acute chest pain presenting to the ED. ©RSNA, 2025.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e240459"},"PeriodicalIF":8.1,"publicationDate":"2025-04-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143812641","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":"Unsupervised Deep Learning for Blood-Brain Barrier Leakage Detection in Diffuse Glioma Using Dynamic Contrast-enhanced MRI.","authors":"Joon Jang, Kyu Sung Choi, Junhyeok Lee, Hyochul Lee, Inpyeong Hwang, Jung Hyun Park, Jin Wook Chung, Seung Hong Choi, Hyeonjin Kim","doi":"10.1148/ryai.240507","DOIUrl":"https://doi.org/10.1148/ryai.240507","url":null,"abstract":"<p><p><i>\"Just Accepted\" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content.</i> Purpose To develop an unsupervised deep learning framework for generalizable blood-brain barrier (BBB) leakage detection using dynamic contrast-enhanced (DCE) MRI, without requiring pharmacokinetic (PK) models and arterial input function (AIF) estimation. Materials and Methods This retrospective study included data from patients who underwent DCE MRI between April 2010 and December 2020. An autoencoder-based anomaly detection (AEAD) identified 1D voxel-wise time-series abnormal signals through reconstruction residuals, separating them into residual leakage signals (RLS) and residual vascular signals (RVS). The RLS maps were evaluated and compared with the volume transfer constant (<i>K</i>^trans) using the structural similarity index (SSIM) and correlation coefficient (<i>r</i>). Generalizability was tested on subsampled data, and <i>IDH</i> status classification performance was assessed using areas under the receiver operating characteristic curves (AUCs). Results A total of 274 patients were included (164 male; mean age 54.23 ± [SD] 14.66 years). RLS showed high structural similarity (SSIM = 0.91 ± 0.02) and correlation (<i>r</i> = 0.56, <i>P</i> < .001) with <i>K</i>^trans. On subsampled data, RLS maps showed better correlation with RLS values from original data (0.89 versus 0.72, <i>P</i> < .001), higher PSNR (33.09 dB versus 28.94 dB, <i>P</i> < .001), and higher SSIM (0.92 versus 0.87, <i>P</i> < .001) compared with K^trans maps. RLS maps also outperformed <i>K</i>^trans maps in predicting <i>IDH</i> mutation status (AUC = 0.87 [95% CI: 0.83-0.91] versus 0.81 [95% CI: 0.76-0.85], <i>P</i> = .02). Conclusion The unsupervised framework effectively detected blood-brain barrier leakage without PK models and AIF. ©RSNA, 2025.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e240507"},"PeriodicalIF":8.1,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143764378","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":"Predicting Respiratory Disease Mortality Risk Using Open-source AI on Chest Radiographs in an Asian Health Screening Population.","authors":"Jong Hyuk Lee, Seung Ho Choi, Hugo J W L Aerts, Jakob Weiss, Vineet K Raghu, Michael T Lu, Jayoun Kim, Seungho Lee, Dongheon Lee, Hyungjin Kim","doi":"10.1148/ryai.240628","DOIUrl":"https://doi.org/10.1148/ryai.240628","url":null,"abstract":"<p><p><i>\"Just Accepted\" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content.</i> Purpose To assess the prognostic value of an open-source deep learning-based chest radiographs (CXR) algorithm, CXR-Lung-Risk, for stratifying respiratory disease mortality risk among an Asian health screening population using baseline and follow-up CXRs. Materials and Methods This single-center, retrospective study analyzed CXRs from individuals who underwent health screenings between January 2004 and June 2018. The CXR-Lung-Risk scores from baseline CXRs were externally tested for predicting mortality due to lung disease or lung cancer, using competing risk analysis, with adjustments made for clinical factors. The additional value of these risk scores beyond clinical factors was evaluated using the likelihood ratio test. An exploratory analysis was conducted on the CXR-Lung-Risk trajectory over a three-year follow-up period for individuals in the highest quartile of baseline respiratory disease mortality risk, using a time-series clustering algorithm. Results Among 36,924 individuals (median age, 58 years [interquartile range: 53-62 years]; 22,352 male), 264 individuals (0.7%) died of respiratory illness, over a median follow-up period of 11.0 years (interquartile range: 7.8- 12.7 years). CXR-Lung-Risk predicted respiratory disease mortality (adjusted hazard ratio [HR] per 5 years: 2.01, 95% CI: 1.76-2.39, <i>P</i> < .001), offering a prognostic improvement over clinical factors (<i>P</i> < .001). The trajectory analysis identified a subgroup with a continuous increase in CXR-Lung-Risk, which was associated with poorer outcomes (adjusted HR for respiratory disease mortality: 3.26, 95% CI: 1.20-8.81, <i>P</i> = .02) compared with the subgroup with a continuous decrease in CXR-Lung-Risk. Conclusion The open-source CXR-Lung-Risk model predicted respiratory disease mortality in an Asian cohort, enabling a two-layer risk stratification approach through an exploratory longitudinal analysis of baseline and follow-up CXRs. ©RSNA, 2025.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e240628"},"PeriodicalIF":8.1,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143765292","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":"Evaluating Performance of a Deep Learning Multilabel Segmentation Model to Quantify Acute and Chronic Brain Lesions at MRI after Stroke and Predict Prognosis.","authors":"Tianyu Tang, Ying Cui, Chunqiang Lu, Huiming Li, Jiaying Zhou, Xiaoyu Zhang, Yujie Zhou, Ying Zhang, Yi Zhang, Yuhao Xu, Yuefeng Li, Shenghong Ju","doi":"10.1148/ryai.240072","DOIUrl":"https://doi.org/10.1148/ryai.240072","url":null,"abstract":"<p><p><i>\"Just Accepted\" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content.</i> Purpose To develop and evaluate a multilabel deep learning network to identify and quantify acute and chronic brain lesions on multisequence MRI after acute ischemic stroke (AIS) and assess relationships between clinical and model-extracted radiologic features of the lesions and patient prognosis. Materials and Methods This retrospective study included AIS patients from multiple centers (September 2008- October 2022) who underwent MRI and thrombolysis or antiplatelets and/or anticoagulants treatment. A SegResNet-based deep learning model was developed to segment core infarcts and white matter hyperintensity (WMH) burdens on diffusion-weighted imaging and fluid-attenuated inversion recovery images. The model was trained, validated and tested with manual labels (<i>n</i> = 260, 60, and 40 patients in each dataset, respectively). Radiologic features extracted from the model, including regional infarct size and periventricular and deep WMH volumes and cluster numbers, combined with clinical variables, were used to predict favorable versus unfavorable patient outcomes at 7 days (modified Rankin scale [mRS] score). Mediation analyses explored associations between radiologic features and AIS outcomes within different treatment groups. Results A total of 1,008 patients (mean age, 67.0 ± 11.8 years; 686 male, 322 female) were included. The training and validation dataset comprised 702 patients with AIS, and the two external testing datasets included 206 and 100 patients, respectively. The prognostic model combining clinical and radiologic features achieved AUCs of 0.81 (95% CI: 0.74-0.88) and 0.77 (95% CI: 0.68-0.86) for predicting 7-day outcomes in the two external testing datasets, respectively. Mediation analyses revealed that deep WMH in patients treated with thrombolysis had a significant direct effect (17.7%, <i>P</i> = .01) and indirect effect (10.7%, <i>P</i> = .01) on unfavorable outcomes, as indicated by higher mRS scores, which was not observed in patients treated antiplatelets and/or anticoagulants. Conclusion The proposed deep learning model quantitatively analyzed radiologic features of acute and chronic brain lesions, and extracted radiologic features combined with clinical variables predicted short-term AIS outcomes. WMH burden, particularly deep WMH, emerged as a risk factor for poor outcomes in patients treated with thrombolysis. ©RSNA, 2025.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e240072"},"PeriodicalIF":8.1,"publicationDate":"2025-03-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143711426","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":"Pseudo-Contrast-Enhanced US via Enhanced Generative Adversarial Networks for Evaluating Tumor Ablation Efficacy.","authors":"Chen Chen, Jiabin Yu, Zhikang Xu, Changsong Xu, Zubang Zhou, Jindong Hao, Vicky Yang Wang, Jincao Yao, Lingyan Zhou, Chenke Xu, Mei Song, Qi Zhang, Xiaofang Liu, Lin Sui, Yuqi Yan, Tian Jiang, Yahan Zhou, Yingtianqi Wu, Binggang Xiao, Chenjie Xu, Hongmei Mi, Li Yang, Zhiwei Wu, Qingquan He, Jian Chen, Qi Liu, Dong Xu","doi":"10.1148/ryai.240370","DOIUrl":"https://doi.org/10.1148/ryai.240370","url":null,"abstract":"<p><p><i>\"Just Accepted\" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content.</i> Purpose To develop a methodology for creating pseudo-contrast-enhanced US (CEUS) using an enhanced generative adversarial network and evaluate its ability to assess tumor ablation effectiveness. Materials and Methods This retrospective study included 1,030 patients who underwent thyroid nodule ablation across seven centers from January 2020 to April 2023. A generative adversarial network-based model was developed for direct pseudo-CEUS generation from B-mode US and tested on thyroid, breast, and liver ablation datasets. The reliability of pseudo-CEUS was assessed using Structural Similarity Index (SSIM), Color Histogram Correlation (CHC), and Mean Absolute Percentage Error (MAPE) against real CEUS. Additionally, a subjective evaluation system was devised to validate its clinical value. The Wilcoxon signed-rank test was employed to analyze differences in the data. Results The study included 1,030 patients (mean age, 46.9 years ± 12.5; 799 females and 231 males). For internal test set 1, the mean SSIM was 0.89 ± 0.05, while across external test sets 1-6, mean SSIM values ranged from 0.84 ± 0.08 to 0.88 ± 0.04. Subjective assessments affirmed the method's stability and near-realistic performance in evaluating ablation effectiveness. The thyroid ablation datasets had an average identification score of 0.49 (0.5 indicates indistinguishability), while the similarity average score for all datasets was 4.75 out of 5. Radiologists' assessments of residual blood supply were nearly consistent, with no differences in defining ablation zones between real and pseudo-CEUS. Conclusion The pseudo-CEUS method demonstrated high similarity to real CEUS in evaluating tumor ablation effectiveness. Published under a CC BY 4.0 license.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e240370"},"PeriodicalIF":8.1,"publicationDate":"2025-03-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143711427","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":"Development and Validation of a Sham-AI Model for Intracranial Aneurysm Detection at CT Angiography.","authors":"Zhao Shi, Bin Hu, Mengjie Lu, Manting Zhang, Haiting Yang, Bo He, Jiyao Ma, Chunfeng Hu, Li Lu, Sheng Li, Shiyu Ren, Yonggao Zhang, Jun Li, Mayidili Nijiati, Jia-Ke Dong, Hao Wang, Zhen Zhou, Fan Dong Zhang, Chengwei Pan, Yizhou Yu, Zijian Chen, Chang Sheng Zhou, Yongyue Wei, Junlin Zhou, Long Jiang Zhang","doi":"10.1148/ryai.240140","DOIUrl":"https://doi.org/10.1148/ryai.240140","url":null,"abstract":"<p><p><i>\"Just Accepted\" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content.</i> Purpose To evaluate a Sham-AI model acting as a placebo control for a Standard-AI model for intracranial aneurysm diagnosis. Materials and Methods This retrospective crossover, blinded, multireader multicase study was conducted from November 2022 to March 2023. A Sham-AI model with near-zero sensitivity and similar specificity to a Standard-AI model was developed using 16,422 CT angiography (CTA) examinations. Digital subtraction angiography-verified CTA examinations from four hospitals were collected, half of which were processed by Standard-AI and the others by Sham-AI to generate Sequence A; Sequence B was generated reversely. Twenty-eight radiologists from seven hospitals were randomly assigned with either sequence, and then assigned with the other sequence after a washout period. The diagnostic performances of radiologists alone, radiologists with Standard-AI-assisted, and radiologists with Sham-AI-assisted were compared using sensitivity and specificity, and radiologists' susceptibility to Sham-AI suggestions was assessed. Results The testing dataset included 300 patients (median age, 61 (IQR, 52.0-67.0) years; 199 male), 50 of which had aneurysms. Standard-AI and Sham-AI performed as expected (sensitivity: 96.0% versus 0.0%, specificity: 82.0% versus 76.0%). The differences in sensitivity and specificity between Standard-AI-assisted and Sham-AIassisted readings were +20.7% (95%CI: 15.8%-25.5%, superiority) and 0.0% (95%CI: -2.0%-2.0%, noninferiority), respectively. The difference between Sham-AI-assisted readings and radiologists alone was-2.6% (95%CI: -3.8%--1.4%, noninferiority) for both sensitivity and specificity. 5.3% (44/823) of true-positive and 1.2% (7/577) of false-negative results of radiologists alone were changed following Sham-AI suggestions. Conclusion Radiologists' diagnostic performance was not compromised when aided by the proposed Sham-AI model compared with their unassisted performance. Published under a CC BY 4.0 license.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e240140"},"PeriodicalIF":8.1,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143658885","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":"Open-Weight Language Models and Retrieval Augmented Generation for Automated Structured Data Extraction from Diagnostic Reports: Assessment of Approaches and Parameters.","authors":"Mohamed Sobhi Jabal, Pranav Warman, Jikai Zhang, Kartikeye Gupta, Ayush Jain, Maciej Mazurowski, Walter Wiggins, Kirti Magudia, Evan Calabrese","doi":"10.1148/ryai.240551","DOIUrl":"https://doi.org/10.1148/ryai.240551","url":null,"abstract":"<p><p><i>\"Just Accepted\" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content.</i> Purpose To develop and evaluate an automated system for extracting structured clinical information from unstructured radiology and pathology reports using open-weights language models (LMs) and retrieval augmented generation (RAG) and to assess the effects of model configuration variables on extraction performance. Materials and Methods This retrospective study utilized two datasets: 7,294 radiology reports annotated for Brain Tumor Reporting and Data System (BT-RADS) scores and 2,154 pathology reports annotated for <i>IDH</i> mutation status (January 2017 to July 2021). An automated pipeline was developed to benchmark the performance of various LMs and RAG configurations for structured data extraction accuracy from reports. The impact of model size, quantization, prompting strategies, output formatting, and inference parameters on model accuracy was systematically evaluated. Results The best performing models achieved up to 98% accuracy in extracting BT-RADS scores from radiology reports and over 90% for <i>IDH</i> mutation status extraction from pathology reports. The best model was medical finetuned llama3. Larger, newer, and domain fine-tuned models consistently outperformed older and smaller models (mean accuracy, 86% versus 75%; <i>P</i> < .001). Model quantization had minimal impact on performance. Few-shot prompting significantly improved accuracy (mean increase: 32% ± 32%, <i>P</i> = .02). RAG improved performance for complex pathology reports +48% ± 11% (<i>P</i> = .001), but not for shorter radiology reports-8% ± 31% (<i>P</i> = .39). Conclusion This study demonstrates the potential of open LMs in automated extraction of structured clinical data from unstructured clinical reports with local privacy-preserving application. Careful model selection, prompt engineering, and semiautomated optimization using annotated data are critical for optimal performance. ©RSNA, 2025.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e240551"},"PeriodicalIF":8.1,"publicationDate":"2025-03-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143606547","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":"Enhancing Large Language Models with Retrieval-augmented Generation: A Radiology-specific Approach.","authors":"Dane A Weinert, Andreas M Rauschecker","doi":"10.1148/ryai.240313","DOIUrl":"https://doi.org/10.1148/ryai.240313","url":null,"abstract":"<p><p><i>\"Just Accepted\" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content.</i> Retrieval-augmented generation (RAG) is a strategy to improve performance of large language models (LLMs) by providing the LLM with an updated corpus of knowledge that can be used for answer generation in real-time. RAG may improve LLM performance and clinical applicability in radiology by providing citable, up-to-date information without requiring model fine-tuning. In this retrospective study, a radiology-specific RAG was developed using a vector database of 3,689 <i>RadioGraphics</i> articles published from January 1999 to December 2023. Performance of 5 LLMs with and without RAG on a 192-question radiology examination was compared. RAG significantly improved examination scores for GPT-4 (81.2% versus 75.5%, <i>P</i> = .04) and Command R+ (70.3% versus 62.0%, <i>P</i> = .02), but not for Claude Opus, Mixtral, or Gemini 1.5 Pro. RAG-System performed significantly better than pure LLMs on a 24-question subset directly sourced from <i>RadioGraphics</i> (85% versus 76%, <i>P</i> = .03). The RAG-System retrieved 21/24 (87.5%, <i>P</i> < .001) relevant <i>RadioGraphics</i> references cited in the examination's answer explanations and successfully cited them in 18/21 (85.7%, <i>P</i> < .001) outputs. The results suggest that RAG is a promising approach to enhance LLM capabilities for radiology knowledge tasks, providing transparent, domain-specific information retrieval. ©RSNA, 2025.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e240313"},"PeriodicalIF":8.1,"publicationDate":"2025-03-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143606543","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":"A Pipeline for Automated Quality Control of Chest Radiographs.","authors":"Ian A Selby, Eduardo González Solares, Anna Breger, Michael Roberts, Lorena Escudero Sánchez, Judith Babar, James H F Rudd, Nicholas A Walton, Evis Sala, Carola-Bibiane Schönlieb, Jonathan R Weir-McCall","doi":"10.1148/ryai.240003","DOIUrl":"https://doi.org/10.1148/ryai.240003","url":null,"abstract":"<p><p><i>\"Just Accepted\" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content.</i> This article presents a suite of quality control tools for chest radiographs based on traditional and artificial intelligence methods, developed and tested with data from 39 centers in 7 countries. Published under a CC BY 4.0 license.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e240003"},"PeriodicalIF":8.1,"publicationDate":"2025-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143558237","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}