Radiology最新文献

{"title":"Diagnostic Accuracy and Clinical Value of a Domain-specific Multimodal Generative AI Model for Chest Radiograph Report Generation.","authors":"Eun Kyoung Hong, Jiyeon Ham, Byungseok Roh, Jawook Gu, Beomhee Park, Sunghun Kang, Kihyun You, Jihwan Eom, Byeonguk Bae, Jae-Bock Jo, Ok Kyu Song, Woong Bae, Ro Woon Lee, Chong Hyun Suh, Chan Ho Park, Seong Jun Choi, Jai Soung Park, Jae-Hyeong Park, Hyun Jeong Jeon, Jeong-Ho Hong, Dosang Cho, Han Seok Choi, Tae Hee Kim","doi":"10.1148/radiol.241476","DOIUrl":"https://doi.org/10.1148/radiol.241476","url":null,"abstract":"<p><p>Background Generative artificial intelligence (AI) is anticipated to alter radiology workflows, requiring a clinical value assessment for frequent examinations like chest radiograph interpretation. Purpose To develop and evaluate the diagnostic accuracy and clinical value of a domain-specific multimodal generative AI model for providing preliminary interpretations of chest radiographs. Materials and Methods For training, consecutive radiograph-report pairs from frontal chest radiography were retrospectively collected from 42 hospitals (2005-2023). The trained domain-specific AI model generated radiology reports for the radiographs. The test set included public datasets (PadChest, Open-i, VinDr-CXR, and MIMIC-CXR-JPG) and radiographs excluded from training. The sensitivity and specificity of the model-generated reports for 13 radiographic findings, compared with radiologist annotations (reference standard), were calculated (with 95% CIs). Four radiologists evaluated the subjective quality of the reports in terms of acceptability, agreement score, quality score, and comparative ranking of reports from <i>(a)</i> the domain-specific AI model, <i>(b)</i> radiologists, and <i>(c)</i> a general-purpose large language model (GPT-4Vision). Acceptability was defined as whether the radiologist would endorse the report as their own without changes. Agreement scores from 1 (clinically significant discrepancy) to 5 (complete agreement) were assigned using RADPEER; quality scores were on a 5-point Likert scale from 1 (very poor) to 5 (excellent). Results A total of 8 838 719 radiograph-report pairs (training) and 2145 radiographs (testing) were included (anonymized with respect to sex and gender). Reports generated by the domain-specific AI model demonstrated high sensitivity for detecting two critical radiographic findings: 95.3% (181 of 190) for pneumothorax and 92.6% (138 of 149) for subcutaneous emphysema. Acceptance rate, evaluated by four radiologists, was 70.5% (6047 of 8680), 73.3% (6288 of 8580), and 29.6% (2536 of 8580) for model-generated, radiologist, and GPT-4Vision reports, respectively. Agreement scores were highest for the model-generated reports (median = 4 [IQR, 3-5]) and lowest for GPT-4Vision reports (median = 1 [IQR, 1-3]; <i>P</i> < .001). Quality scores were also highest for the model-generated reports (median = 4 [IQR, 3-5]) and lowest for the GPT-4Vision reports (median = 2 [IQR, 1-3]; <i>P</i> < .001). From the ranking analysis, model-generated reports were most frequently ranked the highest (60.0%; 5146 of 8580), and GPT-4Vision reports were most frequently ranked the lowest (73.6%; 6312 of 8580). Conclusion A domain-specific multimodal generative AI model demonstrated potential for high diagnostic accuracy and clinical value in providing preliminary interpretations of chest radiographs for radiologists. © RSNA, 2025 <i>Supplemental material is available for this article.</i> See also the editorial by Little in this issue.</p>","PeriodicalId":20896,"journal":{"name":"Radiology","volume":"314 3","pages":"e241476"},"PeriodicalIF":12.1,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143701323","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Harnessing the Power of Generative AI to Enhance Radiologist Efficiency and Accuracy.","authors":"Paul S Babyn, Scott J Adams","doi":"10.1148/radiol.250339","DOIUrl":"https://doi.org/10.1148/radiol.250339","url":null,"abstract":"","PeriodicalId":20896,"journal":{"name":"Radiology","volume":"314 3","pages":"e250339"},"PeriodicalIF":12.1,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143606266","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"US Liver Imaging Reporting and Data System Version 2017: A Systematic Review and Meta-Analysis.","authors":"Sunyoung Lee, Ja Kyung Yoon, Jaeseung Shin, Hyejung Shin, Anum Aslam, Aya Kamaya, Shuchi K Rodgers, Claude B Sirlin, Victoria Chernyak","doi":"10.1148/radiol.240450","DOIUrl":"10.1148/radiol.240450","url":null,"abstract":"<p><p>Background The US Liver Imaging Reporting and Data System (LI-RADS) includes an assessment category (US-1, negative; US-2, subthreshold; and US-3, positive) and a visualization score reflecting image quality (VIS-A, no or minimal limitations; VIS-B, moderate limitations; and VIS-C, severe limitations). The US-3 and VIS-C impact patient treatment. Purpose To establish the distributions of categories and visualization scores, estimate the proportions of hepatocellular carcinoma (HCC) and overall malignancy in the US-3 category, and identify variables associated with the VIS-C score by conducting a meta-analysis. Materials and Methods A systematic search of articles published between January 1, 2017, and September 17, 2023, identified studies reporting distributions of US LI-RADS version 2017 categories or visualization scores. Characteristics of the study design, patient cohorts, and outcomes of interest (distributions of US categories and visualization scores, percentages of probable or definite HCC and malignancy in US-3 category, and variables associated with VIS-C) were extracted. For the meta-analysis, estimates were established with random-effects models. Results Fifteen studies comprising 39 166 US examinations were included. Of all examinations, 89.7% (95% CI: 86.8, 91.8) were categorized US-1; 4.4% (95% CI: 2.8, 6.2), US-2; and 5.9% (95% CI: 4.1, 8.0), US-3. Of the US-3 examinations, 25.9% (95% CI: 17.1, 34.7) had probable or definite HCC and 26.4% (95% CI: 18.4, 34.5) had overall malignancy. Among all examinations, 59.7% (95% CI: 46.9, 67.8) were assigned VIS-A; 32.5% (95% CI: 21.9, 41.6), VIS-B; and 7.8% (95% CI: 2.8, 14.3), VIS-C. Obesity (odds ratio [OR], 2.37; 95% CI: 1.57, 3.59), nonalcoholic fatty liver disease (NAFLD) (OR, 2.24; 95% CI: 1.64, 3.06), and Child-Pugh B or C (OR, 2.41; 95% CI: 1.43, 4.06) were associated with VIS-C score. Conclusion Overall, 90% of surveillance US results were negative (US-1), and 92% were of adequate quality (VIS-A or VIS-B); 26% of patients with US-3 results had HCC. VIS-C was associated with obesity, NAFLD, and cirrhosis. Systemic review registry no. CRD42022384925 © RSNA, 2025 <i>Supplemental material is available for this article</i>.</p>","PeriodicalId":20896,"journal":{"name":"Radiology","volume":"314 3","pages":"e240450"},"PeriodicalIF":12.1,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143606271","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From SUV Ratio to Fill States: Advancing PET Quantification of Alzheimer Disease through Spatial Extent.","authors":"Mijin Yun, Daesung Kim","doi":"10.1148/radiol.250545","DOIUrl":"https://doi.org/10.1148/radiol.250545","url":null,"abstract":"","PeriodicalId":20896,"journal":{"name":"Radiology","volume":"314 3","pages":"e250545"},"PeriodicalIF":12.1,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143701334","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Imaging-based Hepatocellular Carcinoma Diagnosis: An Extension of LI-RADS to Patients with Noncirrhotic Hepatitis C.","authors":"Helmut Schöllnast","doi":"10.1148/radiol.250627","DOIUrl":"https://doi.org/10.1148/radiol.250627","url":null,"abstract":"","PeriodicalId":20896,"journal":{"name":"Radiology","volume":"314 3","pages":"e250627"},"PeriodicalIF":12.1,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143701348","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Patching within Spinal Longitudinal Extradural Collections for Ventral Dural Tears.","authors":"Horst Urbach","doi":"10.1148/radiol.250411","DOIUrl":"https://doi.org/10.1148/radiol.250411","url":null,"abstract":"","PeriodicalId":20896,"journal":{"name":"Radiology","volume":"314 3","pages":"e250411"},"PeriodicalIF":12.1,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143701350","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Revisiting the Diagnostic Criteria for Interstitial Pneumonia with Autoimmune Features.","authors":"Jeanne B Ackman","doi":"10.1148/radiol.250430","DOIUrl":"https://doi.org/10.1148/radiol.250430","url":null,"abstract":"","PeriodicalId":20896,"journal":{"name":"Radiology","volume":"314 3","pages":"e250430"},"PeriodicalIF":12.1,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143701351","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"¹⁸F-Fluoroestradiol PET/CT for Staging Low-Grade Estrogen Receptor-Positive Breast Cancer.","authors":"Amy M Fowler","doi":"10.1148/radiol.250135","DOIUrl":"10.1148/radiol.250135","url":null,"abstract":"","PeriodicalId":20896,"journal":{"name":"Radiology","volume":"314 3","pages":"e250135"},"PeriodicalIF":12.1,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11950881/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143543231","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Comparison of Contrast-enhanced Mammography and Low-Energy Imaging with or without Supplemental Whole-Breast US in Breast Cancer Detection.","authors":"Joao V Horvat, Tali Amir, Gordon P Watt, Christopher E Comstock, Noam Nissan, Maxine S Jochelson, Janice S Sung","doi":"10.1148/radiol.242006","DOIUrl":"10.1148/radiol.242006","url":null,"abstract":"<p><p>Background Contrast-enhanced mammography (CEM) is an emerging modality that generates low-energy (LE) images that are visually equivalent to full-field digital mammography (FFDM) and recombined images that show lesion vascularity such as MRI. Supplemental whole-breast US increases cancer detection rates when performed with FFDM but not with MRI. Purpose To compare the performance of CEM, LE images, and LE images supplemented with whole-breast US in breast cancer detection during screening. Materials and Methods This prospective study recruited female participants from December 2014 to February 2019 who were scheduled for screening mammography and whole-breast US. CEM (including LE images and recombined images) and whole-breast US images were interpreted by separate breast radiologists blinded to the findings on images from the other modality. Statistical differences in sensitivity and specificity, positive predictive value (PPV), negative predictive value, and abnormal interpretation rate were assessed. Biopsy recommendation rate and PPVs of biopsies performed (PPV₃) were calculated at the lesion level. Results Across 468 participants (median age, 54 years [IQR, 48-59 years]; all female participants), nine screen-detected cancers were diagnosed in eight participants: one cancer was depicted at LE imaging alone (cancer detection rate, 2.1 of 1000), four were depicted at LE imaging with whole-breast US (cancer detection rate, 8.5 of 1000), and eight were depicted at CEM (cancer detection rate, 17.1 of 1000; <i>P</i> < .05). The abnormal interpretation rate was 10.3% (48 of 468) for LE images, 13.7% (64 of 468) for LE images with whole-breast US, and 18.6% (87 of 468) for CEM (<i>P</i> < .001). The biopsy recommendation rate was 15.0 of 1000 for LE images, 38.4 of 1000 for LE images with whole-breast US, and 42.7 of 1000 for CEM. Seven biopsies were recommended based on LE images (PPV₃ of one of seven [14.3%]), 18 biopsies based on LE images with whole-breast US (with a PPV₃ of five of 18 [27.8%]), and 20 biopsies based on CEM (PPV₃ of 9 of 20 [45.0%]). Conclusion Breast cancer detection improved with CEM compared with LE images alone or LE images with whole-breast US. ClinicalTrials.gov Identifier: NCT02310698 © RSNA, 2025 <i>Supplemental material is available for this article.</i></p>","PeriodicalId":20896,"journal":{"name":"Radiology","volume":"314 3","pages":"e242006"},"PeriodicalIF":12.1,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11950873/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143606264","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Dysplasia Epiphysealis Hemimelica of Humerus and Glenoid.","authors":"Chia-Hui Chen, Chien-Kuo Wang","doi":"10.1148/radiol.241712","DOIUrl":"https://doi.org/10.1148/radiol.241712","url":null,"abstract":"","PeriodicalId":20896,"journal":{"name":"Radiology","volume":"314 3","pages":"e241712"},"PeriodicalIF":12.1,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143701328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}