European Journal of Radiology Open最新文献_第2页

Machine learning-based multi-class classification of bladder pathologies using fused 3D CT radiomic and 3D auto-encoder deep features 基于机器学习的膀胱病理多分类融合三维CT放射学和三维自编码器深度特征

IF 2.9 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

European Journal of Radiology Open

Pub Date : 2026-01-09 DOI: 10.1016/j.ejro.2026.100728

Hongwei Xiao , Weihao Liu , Huancheng Yang , Zexin Huang , Yangguang Yuan , Tianyu Wang , Hanlin Liu , Kai Wu

Objective

To develop an automated analytical framework that integrates hybrid radiomics and deep learning features from non-contrast CT images for the multi-class classification of bladder pathologies.

Methods

This retrospective study analyzed 902 CT scans (584 normal, 142 calculi, 66 cancers, 110 cystitis). An integrated pipeline was implemented, comprising: 1) automatic bladder segmentation using a 3D-UNet, 2) hybrid feature extraction combining 100 radiomics features and 256 deep features from a 3D convolutional autoencoder, 3) feature selection via variance thresholding and LASSO regression, and 4) final classification using an XGBoost classifier. The dataset was split into training (80 %) and validation (20 %) sets. Performance was evaluated using the area under the receiver operating characteristic curve (AUROC) with a one-vs-rest strategy for multi-class classification. Model stability was assessed via stratified five-fold cross-validation, and interpretability was analyzed with SHapley Additive exPlanations (SHAP).

Results

The framework achieved one-vs-rest AUROCs of 0.94 (95 % CI: 0.89–0.99) for calculi, 0.92 (0.85–0.99) for cancer, 0.90 (0.84–0.95) for normal bladder, and 0.83 (0.75–0.91) for cystitis. The micro-average AUROC for four-class discrimination was 0.94 (0.92–0.96). Binary normal/abnormal classification demonstrated stable performance across cross-validation folds (AUROC range: 0.89–0.92). SHAP analysis revealed that radiomic features dominated decisions for calculi/normal differentiation, while deep features were critical for distinguishing cancer and cystitis.

Conclusion

The proposed hybrid CT analysis framework achieves clinically relevant performance in the automated, multi-class classification of bladder pathologies, excelling particularly in calculi detection. The complementary roles of radiomic and deep features provide an interpretable diagnostic aid, demonstrating potential for integration into clinical workflows to support differential diagnosis.

目的开发一种结合非对比CT图像放射组学和深度学习特征的自动分析框架，用于膀胱病理的多类别分类。方法回顾性分析902例CT扫描（正常584例，结石142例，癌66例，膀胱炎110例）。实现了一个集成的管道，包括：1)使用3D- unet自动膀胱分割，2)结合100个放射组学特征和来自3D卷积自编码器的256个深度特征的混合特征提取，3)通过方差阈值和LASSO回归进行特征选择，4)使用XGBoost分类器进行最终分类。数据集被分成训练集（80 %）和验证集（20 %）。使用接收者工作特征曲线下面积（AUROC）对性能进行评估，并采用一对休息策略进行多类别分类。通过分层五重交叉验证评估模型稳定性，并使用SHapley加性解释（SHAP）分析可解释性。结果该框架的auroc为：结石0.94(95 % CI: 0.89-0.99)，癌症0.92(0.85-0.99)，正常膀胱0.90(0.84-0.95)，膀胱炎0.83（0.75-0.91）。四类鉴别的微平均AUROC为0.94（0.92 ~ 0.96）。二元正常/异常分类在交叉验证折叠中表现稳定（AUROC范围：0.89-0.92）。SHAP分析显示，放射学特征主导了结石/正常分化的决定，而深部特征对区分癌症和膀胱炎至关重要。结论本文提出的混合CT分析框架在膀胱病理的自动、多类别分类中达到了临床相关的性能，尤其在结石检测方面表现突出。放射学和深部特征的互补作用提供了可解释的诊断辅助，展示了整合到临床工作流程以支持鉴别诊断的潜力。

{"title":"Machine learning-based multi-class classification of bladder pathologies using fused 3D CT radiomic and 3D auto-encoder deep features","authors":"Hongwei Xiao , Weihao Liu , Huancheng Yang , Zexin Huang , Yangguang Yuan , Tianyu Wang , Hanlin Liu , Kai Wu","doi":"10.1016/j.ejro.2026.100728","DOIUrl":"10.1016/j.ejro.2026.100728","url":null,"abstract":"<div><h3>Objective</h3><div>To develop an automated analytical framework that integrates hybrid radiomics and deep learning features from non-contrast CT images for the multi-class classification of bladder pathologies.</div></div><div><h3>Methods</h3><div>This retrospective study analyzed 902 CT scans (584 normal, 142 calculi, 66 cancers, 110 cystitis). An integrated pipeline was implemented, comprising: 1) automatic bladder segmentation using a 3D-UNet, 2) hybrid feature extraction combining 100 radiomics features and 256 deep features from a 3D convolutional autoencoder, 3) feature selection via variance thresholding and LASSO regression, and 4) final classification using an XGBoost classifier. The dataset was split into training (80 %) and validation (20 %) sets. Performance was evaluated using the area under the receiver operating characteristic curve (AUROC) with a one-vs-rest strategy for multi-class classification. Model stability was assessed via stratified five-fold cross-validation, and interpretability was analyzed with SHapley Additive exPlanations (SHAP).</div></div><div><h3>Results</h3><div>The framework achieved one-vs-rest AUROCs of 0.94 (95 % CI: 0.89–0.99) for calculi, 0.92 (0.85–0.99) for cancer, 0.90 (0.84–0.95) for normal bladder, and 0.83 (0.75–0.91) for cystitis. The micro-average AUROC for four-class discrimination was 0.94 (0.92–0.96). Binary normal/abnormal classification demonstrated stable performance across cross-validation folds (AUROC range: 0.89–0.92). SHAP analysis revealed that radiomic features dominated decisions for calculi/normal differentiation, while deep features were critical for distinguishing cancer and cystitis.</div></div><div><h3>Conclusion</h3><div>The proposed hybrid CT analysis framework achieves clinically relevant performance in the automated, multi-class classification of bladder pathologies, excelling particularly in calculi detection. The complementary roles of radiomic and deep features provide an interpretable diagnostic aid, demonstrating potential for integration into clinical workflows to support differential diagnosis.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"16 ","pages":"Article 100728"},"PeriodicalIF":2.9,"publicationDate":"2026-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939415","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}

引用次数: 0

Myocardial transit time mapping by CMR: A novel potential parameter of microcirculatory dysfunction in hypertrophic cardiomyopathy with and without atrial fibrillation CMR心肌传递时间映射：伴有和不伴有房颤的肥厚性心肌病微循环功能障碍的新潜在参数

IF 2.9 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

European Journal of Radiology Open

Pub Date : 2026-01-08 DOI: 10.1016/j.ejro.2026.100726

Junlin Yang , Yong Cheng , Jinxiu Yang , Yuncheng Li , Zhen Wang , Shutian An , Jun Wang , Yongqiang Yu , Ren Zhao , Xiaohu Li

Purpose

This study aimed to explore myocardial transit time (MyoTT) by cardiovascular magnetic resonance (CMR) as a potential parameter of coronary microvascular dysfunction (CMD) in hypertrophic cardiomyopathy (HCM) with and without atrial fibrillation (AF).

Materials and Methods

This study enrolled 50 patients with HCM and 50 healthy control subjects to assess cardiac function, native T1, extracellular volume (ECV), myocardial strain, and MyoTT. Factors associated with HCM status were estimated by logistic regression analyses, adjusting for potential confounders. Meanwhile, we evaluated the discriminative performance of the aforementioned parameters for HCM (vs control) and for AF status among HCM patients.

Results

In the HCM and control groups, the HCM group exhibited significant reductions in LV-GLS, RV-EDVI, and RV-ESVI (all P < 0.05) and significant increases in LVMI, LVGPWT, native T1 values, ECV values, absolute MyoTT values and presence of LGE (all P < 0.05). Logistic regression analysis revealed HCM was significantly associated with MyoTT, native T1, ECV and LV-GLS. In HCM with and without AF, the HCM with AF group showed nominally higher native T1 values and absolute MyoTT values before correction. Logistic regression analysis revealed HCM with AF was significantly associated with MyoTT and native T1. The AUCs for ECV, MyoTT, LV-GLS, and native T1 were 0.828, 0.848, 0.708, and 0.862, respectively (P < 0.05). The AUCs for discriminating HCM with AF from HCM without AF for MyoTT and native T1 were 0.740 and 0.681, respectively (P < 0.05).

Conclusion

MyoTT levels are elevated in patients with HCM, particularly when AF is present. This suggests that MyoTT may reflect CMD-related alterations in HCM, especially in those with AF.

目的探讨心肌传递时间（MyoTT）作为肥厚性心肌病（HCM）伴心房颤动（AF）和不伴心房颤动（AF）患者冠状动脉微血管功能障碍（CMD）的潜在参数。材料与方法本研究招募了50例HCM患者和50例健康对照者，评估心功能、原生T1、细胞外体积（ECV）、心肌应变和MyoTT。通过逻辑回归分析估计与HCM状态相关的因素，调整潜在的混杂因素。同时，我们评估了上述参数对HCM（对照）和HCM患者房颤状态的判别性能。结果在HCM组和对照组中，HCM组的LV-GLS、RV-EDVI和RV-ESVI均显著降低（P均 <； 0.05），LVMI、LVGPWT、原生T1值、ECV值、绝对MyoTT值和LGE的存在均显著升高（P均 <； 0.05）。Logistic回归分析显示HCM与MyoTT、原生T1、ECV和LV-GLS显著相关。在伴有和不伴有房颤的HCM中，伴有房颤的HCM组在校正前的T1值和MyoTT绝对值名义上更高。Logistic回归分析显示HCM合并AF与MyoTT和原生T1显著相关。ECV、MyoTT、LV-GLS、原生T1的auc分别为0.828、0.848、0.708、0.862 （P <； 0.05）。MyoTT和原生T1区分有AF的HCM和无AF的HCM的auc分别为0.740和0.681 （P <； 0.05）。结论HCM患者myott水平升高，尤其是房颤患者。这表明MyoTT可能反映了HCM中与cd相关的改变，尤其是房颤患者。

{"title":"Myocardial transit time mapping by CMR: A novel potential parameter of microcirculatory dysfunction in hypertrophic cardiomyopathy with and without atrial fibrillation","authors":"Junlin Yang , Yong Cheng , Jinxiu Yang , Yuncheng Li , Zhen Wang , Shutian An , Jun Wang , Yongqiang Yu , Ren Zhao , Xiaohu Li","doi":"10.1016/j.ejro.2026.100726","DOIUrl":"10.1016/j.ejro.2026.100726","url":null,"abstract":"<div><h3>Purpose</h3><div>This study aimed to explore myocardial transit time (MyoTT) by cardiovascular magnetic resonance (CMR) as a potential parameter of coronary microvascular dysfunction (CMD) in hypertrophic cardiomyopathy (HCM) with and without atrial fibrillation (AF).</div></div><div><h3>Materials and Methods</h3><div>This study enrolled 50 patients with HCM and 50 healthy control subjects to assess cardiac function, native T1, extracellular volume (ECV), myocardial strain, and MyoTT. Factors associated with HCM status were estimated by logistic regression analyses, adjusting for potential confounders. Meanwhile, we evaluated the discriminative performance of the aforementioned parameters for HCM (vs control) and for AF status among HCM patients.</div></div><div><h3>Results</h3><div>In the HCM and control groups, the HCM group exhibited significant reductions in LV-GLS, RV-EDVI, and RV-ESVI (all P < 0.05) and significant increases in LVMI, LVGPWT, native T1 values, ECV values, absolute MyoTT values and presence of LGE (all P < 0.05). Logistic regression analysis revealed HCM was significantly associated with MyoTT, native T1, ECV and LV-GLS. In HCM with and without AF, the HCM with AF group showed nominally higher native T1 values and absolute MyoTT values before correction. Logistic regression analysis revealed HCM with AF was significantly associated with MyoTT and native T1. The AUCs for ECV, MyoTT, LV-GLS, and native T1 were 0.828, 0.848, 0.708, and 0.862, respectively (P < 0.05). The AUCs for discriminating HCM with AF from HCM without AF for MyoTT and native T1 were 0.740 and 0.681, respectively (P < 0.05).</div></div><div><h3>Conclusion</h3><div>MyoTT levels are elevated in patients with HCM, particularly when AF is present. This suggests that MyoTT may reflect CMD-related alterations in HCM, especially in those with AF.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"16 ","pages":"Article 100726"},"PeriodicalIF":2.9,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939568","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}

引用次数: 0

The consequences of percutaneous transhepatic biliary drainage (PTBD) in patients with tumoral obstructive jaundice: A retrospective study and review of literature 肿瘤梗阻性黄疸患者经皮经肝胆道引流（PTBD）的后果：回顾性研究和文献复习

IF 2.9 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

European Journal of Radiology Open

Pub Date : 2026-01-08 DOI: 10.1016/j.ejro.2025.100722

Javad Jalili, Yaser Ahmadi, Mahsa Karbasi, Sarah Vaseghi, Mahdiyeh Baastani Khajeh, Sahar Rezaei, Esmaeil Gharepapagh, Alireza Motamedi

Background

Percutaneous transhepatic biliary drainage (PTBD) is widely used in malignant obstructive jaundice (MOJ), but most series report aggregate complication rates without stratification by tumor type. This limits individualized risk counseling and hampers benchmarking across centers.

Objective

To evaluate short-term outcomes of PTBD in a large single-center cohort, with a focus on tumor-specific complication patterns using patient-level attribution and robust statistical methods.

Methods

We retrospectively analyzed 453 patients with MOJ undergoing PTBD (2017–2022). Complications within 30 days were recorded at the patient level, with downstream sequelae attributed to the index event. Exact tests and Firth penalized logistic regression were applied to mitigate sparse-data bias. Survival status was documented through 90 days.

Results

Technical success was 100 %, and mean bilirubin declined significantly within 48 h. Twenty-two complications occurred in 21 patients (4.6 %), lower than most published series. Catheter displacement (1.5 %) was the most frequent event, while severe bleeding (0.2 %) and biliary peritonitis (0.6 %) were rare and managed without surgery. No cholangitis was observed. Complications were most frequent in cholangiocarcinoma and pancreatic cancer, although differences across tumor types were not statistically significant. Follow-up was complete; no deaths occurred within 90 days, yielding 30-day and 90-day survival rates of 100 %.

Conclusions

PTBD is highly effective and safe in MOJ, with lower complication rates than many prior reports. Tumor-specific analysis revealed variation in complication subtypes but no significant differences in overall incidence. The rigorous methodology and complete follow-up provide a robust framework for individualized risk profiling and multicenter benchmarking.

背景：经皮经肝胆道引流术（PTBD）被广泛应用于恶性梗阻性黄疸（MOJ），但大多数系列报道的综合并发症发生率未按肿瘤类型分层。这限制了个性化的风险咨询，并阻碍了跨中心的基准。目的在一个大型单中心队列中评估PTBD的短期预后，重点研究肿瘤特异性并发症模式，采用患者水平归因和稳健的统计方法。方法回顾性分析2017-2022年453例MOJ行PTBD的患者。在患者层面记录30天内的并发症，下游的后遗症归因于指数事件。采用精确检验和Firth惩罚逻辑回归来减轻稀疏数据偏差。90天内记录生存状况。结果技术成功率为100% %，平均胆红素在48 h内显著下降。21例患者发生22例并发症（4.6 %），低于大多数已发表的系列。导管移位（1.5 %）是最常见的事件，而严重出血（0.2 %）和胆道性腹膜炎（0.6 %）是罕见的，无需手术治疗。未见胆管炎。并发症在胆管癌和胰腺癌中最为常见，尽管肿瘤类型之间的差异没有统计学意义。随访完成；90天内未发生死亡，30天和90天存活率为100% %。结论sptbd治疗MOJ疗效高，安全性好，并发症发生率低于文献报道。肿瘤特异性分析显示并发症亚型存在差异，但总体发生率无显著差异。严格的方法和完整的随访为个性化风险分析和多中心基准测试提供了强有力的框架。

{"title":"The consequences of percutaneous transhepatic biliary drainage (PTBD) in patients with tumoral obstructive jaundice: A retrospective study and review of literature","authors":"Javad Jalili, Yaser Ahmadi, Mahsa Karbasi, Sarah Vaseghi, Mahdiyeh Baastani Khajeh, Sahar Rezaei, Esmaeil Gharepapagh, Alireza Motamedi","doi":"10.1016/j.ejro.2025.100722","DOIUrl":"10.1016/j.ejro.2025.100722","url":null,"abstract":"<div><h3>Background</h3><div>Percutaneous transhepatic biliary drainage (PTBD) is widely used in malignant obstructive jaundice (MOJ), but most series report aggregate complication rates without stratification by tumor type. This limits individualized risk counseling and hampers benchmarking across centers.</div></div><div><h3>Objective</h3><div>To evaluate short-term outcomes of PTBD in a large single-center cohort, with a focus on tumor-specific complication patterns using patient-level attribution and robust statistical methods.</div></div><div><h3>Methods</h3><div>We retrospectively analyzed 453 patients with MOJ undergoing PTBD (2017–2022). Complications within 30 days were recorded at the patient level, with downstream sequelae attributed to the index event. Exact tests and Firth penalized logistic regression were applied to mitigate sparse-data bias. Survival status was documented through 90 days.</div></div><div><h3>Results</h3><div>Technical success was 100 %, and mean bilirubin declined significantly within 48 h. Twenty-two complications occurred in 21 patients (4.6 %), lower than most published series. Catheter displacement (1.5 %) was the most frequent event, while severe bleeding (0.2 %) and biliary peritonitis (0.6 %) were rare and managed without surgery. No cholangitis was observed. Complications were most frequent in cholangiocarcinoma and pancreatic cancer, although differences across tumor types were not statistically significant. Follow-up was complete; no deaths occurred within 90 days, yielding 30-day and 90-day survival rates of 100 %.</div></div><div><h3>Conclusions</h3><div>PTBD is highly effective and safe in MOJ, with lower complication rates than many prior reports. Tumor-specific analysis revealed variation in complication subtypes but no significant differences in overall incidence. The rigorous methodology and complete follow-up provide a robust framework for individualized risk profiling and multicenter benchmarking.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"16 ","pages":"Article 100722"},"PeriodicalIF":2.9,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939569","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}

引用次数: 0

Workflow-embedded AI as a cognitive scaffold: A randomized trial on knowledge retention and diagnostic competency in undergraduate radiology education 嵌入工作流的人工智能作为认知支架：本科放射学教育中知识保留和诊断能力的随机试验

IF 2.9 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

European Journal of Radiology Open

Pub Date : 2026-01-08 DOI: 10.1016/j.ejro.2026.100724

Jing Li , Haiyan Zhao

Background

Traditional didactic methods in medical imaging education, predominantly reliant on static images (non-augmented, traditional PACS workflow that requires manual, unguided search and interpretation), consistently fail to bridge the theory-practice divide, contributing to high diagnostic error rates. While the integration of artificial intelligence (AI) with Picture Archiving and Communication Systems (PACS+AI) offers transformative potential, robust evidence quantifying its impact on longitudinal competency development remains scarce.

Objective

This study aims to quantitatively evaluate the efficacy of a cognitively optimized PACS+AI framework versus conventional PACS in enhancing radiology education across four critical domains: theoretical knowledge, clinical decision-making competencies, AI acceptance, and knowledge retention.

Methods

In a prospective single-blind randomized controlled trial (RCT), 110 medical imaging undergraduates were randomized to PACS+AI (n = 55) or standard PACS (n = 55) groups. Theoretical knowledge was assessed using validated item-bank assessments; clinical decision-making competencies were evaluated through lesion detection, anatomical localization, diagnostic accuracy, and report completeness; AI acceptance was measured using the Technology Acceptance Model (TAM); and knowledge retention was tracked through immediate, 1-month, and 3-month follow-up assessments. The PACS+AI framework provided three core cognitive support functions: automated lesion annotation, structured diagnostic prompting, and workflow-contextualized feedback.

Results

The PACS+AI group demonstrated significantly superior outcomes across all domains: theoretical knowledge retention was substantially higher (79.3 % vs. 19.7 % at 3 months, P < 0.001, d=1.95); clinical decision-making competencies showed progressive improvement with large effect sizes (Δ=12.4–18.1, all P < 0.001, d=1.88–2.48); AI acceptance scores were significantly elevated across all TAM constructs (all P < 0.001, d>1.9); and knowledge retention was maintained longitudinally with amplified effects over time.

Conclusion

The PACS+AI framework significantly enhances radiology education by optimizing cognitive load distribution, resulting in sustained knowledge retention, superior clinical decision-making competencies, and heightened AI acceptance. This integrated teaching model effectively bridges the gap between theory and practice, cultivates professionals adaptable to the artificial intelligence environment, and aligns with the core needs of the new generation of medical education.

医学影像教育中的传统教学方法主要依赖于静态图像（非增强的，传统的PACS工作流程，需要手动，无指导的搜索和解释），始终未能弥合理论与实践的鸿沟，导致高诊断错误率。虽然人工智能（AI）与图像存档和通信系统（PACS+AI）的集成提供了变革潜力，但量化其对纵向能力发展影响的有力证据仍然很少。本研究旨在定量评估认知优化的PACS+AI框架与传统PACS在四个关键领域（理论知识、临床决策能力、人工智能接受和知识保留）加强放射学教育方面的效果。方法采用前瞻性单盲随机对照试验（RCT），将110名医学影像专业本科生随机分为PACS+AI组（n = 55）和标准PACS组（n = 55）。采用有效的题库评估来评估理论知识；通过病变检测、解剖定位、诊断准确性和报告完整性评估临床决策能力；使用技术接受模型（TAM）测量人工智能接受度；通过即时、1个月和3个月的随访评估来跟踪知识保留情况。PACS+AI框架提供了三个核心认知支持功能：自动病变注释、结构化诊断提示和工作流上下文化反馈。结果PACS+AI组在所有领域都表现出显著的优势：理论知识保留率明显更高（3个月时79.3% % vs. 19.7 %,P <； 0.001,d=1.95）；临床决策能力呈进行性改善，且效应量较大（Δ= 12.4-18.1， P均为 <； 0.001,d= 1.88-2.48）；人工智能接受得分在所有TAM结构中显著升高（P均为 <； 0.001,d>1.9）；随着时间的推移，知识保留在纵向上保持着放大效应。结论PACS+AI框架通过优化认知负荷分配，显著增强放射学教育，实现持续的知识保留、卓越的临床决策能力和更高的人工智能接受度。这种一体化的教学模式有效地弥合了理论与实践的差距，培养了适应人工智能环境的专业人才，符合新一代医学教育的核心需求。

{"title":"Workflow-embedded AI as a cognitive scaffold: A randomized trial on knowledge retention and diagnostic competency in undergraduate radiology education","authors":"Jing Li , Haiyan Zhao","doi":"10.1016/j.ejro.2026.100724","DOIUrl":"10.1016/j.ejro.2026.100724","url":null,"abstract":"<div><h3>Background</h3><div>Traditional didactic methods in medical imaging education, predominantly reliant on static images (non-augmented, traditional PACS workflow that requires manual, unguided search and interpretation), consistently fail to bridge the theory-practice divide, contributing to high diagnostic error rates. While the integration of artificial intelligence (AI) with Picture Archiving and Communication Systems (PACS+AI) offers transformative potential, robust evidence quantifying its impact on longitudinal competency development remains scarce.</div></div><div><h3>Objective</h3><div>This study aims to quantitatively evaluate the efficacy of a cognitively optimized PACS+AI framework versus conventional PACS in enhancing radiology education across four critical domains: theoretical knowledge, clinical decision-making competencies, AI acceptance, and knowledge retention.</div></div><div><h3>Methods</h3><div>In a prospective single-blind randomized controlled trial (RCT), 110 medical imaging undergraduates were randomized to PACS+AI (n = 55) or standard PACS (n = 55) groups. Theoretical knowledge was assessed using validated item-bank assessments; clinical decision-making competencies were evaluated through lesion detection, anatomical localization, diagnostic accuracy, and report completeness; AI acceptance was measured using the Technology Acceptance Model (TAM); and knowledge retention was tracked through immediate, 1-month, and 3-month follow-up assessments. The PACS+AI framework provided three core cognitive support functions: automated lesion annotation, structured diagnostic prompting, and workflow-contextualized feedback.</div></div><div><h3>Results</h3><div>The PACS+AI group demonstrated significantly superior outcomes across all domains: theoretical knowledge retention was substantially higher (79.3 % vs. 19.7 % at 3 months, P < 0.001, d=1.95); clinical decision-making competencies showed progressive improvement with large effect sizes (Δ=12.4–18.1, all P < 0.001, d=1.88–2.48); AI acceptance scores were significantly elevated across all TAM constructs (all P < 0.001, d>1.9); and knowledge retention was maintained longitudinally with amplified effects over time.</div></div><div><h3>Conclusion</h3><div>The PACS+AI framework significantly enhances radiology education by optimizing cognitive load distribution, resulting in sustained knowledge retention, superior clinical decision-making competencies, and heightened AI acceptance. This integrated teaching model effectively bridges the gap between theory and practice, cultivates professionals adaptable to the artificial intelligence environment, and aligns with the core needs of the new generation of medical education.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"16 ","pages":"Article 100724"},"PeriodicalIF":2.9,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939272","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}

引用次数: 0

A biology-informed radiomics model for prognostication of hepatocellular carcinoma based on AKR1B10 expression 基于AKR1B10表达的肝细胞癌预后的生物学信息放射组学模型

IF 2.9 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

European Journal of Radiology Open

Pub Date : 2026-01-06 DOI: 10.1016/j.ejro.2026.100725

Hongan Ying , Lili Huang , Weiwen Hong

Background

Current radiomic models for hepatocellular carcinoma (HCC) prognosis rely on direct correlations between imaging features and clinical outcomes, resulting in limited biological interpretability and restricted clinical applicability. This study explores a novel biology-driven radiomic strategy focusing on AKR1B10. AKR1B10 is a functionally established molecular driver of HCC progression, and the study aims to develop an interpretable prediction model bridging imaging phenotypes and underlying tumor biology.

Methods

We analyzed multi-institutional data from The Cancer Genome Atlas (TCGA) and The Cancer Imaging Archive (TCIA). After confirming the prognostic significance of AKR1B10 expression through survival and enrichment analyses, we developed a radiomics model using a cohort of 34 patients with matched computed tomography (CT) images and genomic data. Tumor and peritumoral regions were segmented, and 107 radiomic features were extracted. Feature selection was performed using maximum-relevance-minimum-redundancy (mRMR) and recursive feature elimination (RFE) algorithms, with subsequent model building via logistic regression. The model was evaluated using ROC analysis, calibration curves, and decision curve analysis. Finally, we constructed a prognostic nomogram integrating the radiomics signature with clinical variables.

Results

AKR1B10 overexpression was significantly associated with poor overall survival (HR = 2.187, 95 % CI: 1.385–3.454, P < 0.001) and characteristic activation of oncogenic pathways. The radiomics model demonstrated strong performance in predicting AKR1B10 status (AUC = 0.83, 95 % CI: 0.69–0.97), with significant difference in rad-scores between AKR1B10 high- and low-expression groups (P < 0.001). The integrated nomogram showed excellent predictive accuracy for 3-year survival (AUC = 0.85) and provided clinical net benefit across threshold probabilities.

Conclusions

The biology-informed radiomics model based on AKR1B10 expression demonstrates strong prognostic performance in hepatocellular carcinoma. By directly linking imaging phenotypes to a key molecular driver of HCC, this approach provides a clinically applicable and biologically interpretable tool for pre-operative risk prediction.

当前肝细胞癌（HCC）预后的放射学模型依赖于影像学特征与临床结果之间的直接相关性，导致其生物学可解释性和临床适用性有限。本研究探索了一种新的以AKR1B10为中心的生物学驱动的放射学策略。AKR1B10是HCC进展的功能性分子驱动因子，该研究旨在建立一种可解释的预测模型，将影像学表型与潜在的肿瘤生物学联系起来。方法我们分析了来自癌症基因组图谱（TCGA）和癌症成像档案（TCIA）的多机构数据。在通过生存和富集分析确认了AKR1B10表达的预后意义后，我们利用34名具有匹配计算机断层扫描（CT）图像和基因组数据的患者建立了放射组学模型。对肿瘤和肿瘤周围区域进行分割，提取107个放射学特征。使用最大相关最小冗余（mRMR）和递归特征消除（RFE）算法进行特征选择，随后通过逻辑回归建立模型。采用ROC分析、校正曲线和决策曲线分析对模型进行评价。最后，我们构建了一个结合放射组学特征和临床变量的预后图。结果akr1b10过表达与较差的总生存率（HR = 2.187, 95 % CI: 1.385-3.454, P <； 0.001）和特征性的致癌途径激活显著相关。放射组学模型在预测AKR1B10状态方面表现出色（AUC = 0.83, 95 % CI: 0.69-0.97）， AKR1B10高表达组和低表达组之间的rad评分存在显著差异（P <； 0.001）。综合nomogram对3年生存率（AUC = 0.85）的预测准确性极高，并提供了跨阈值概率的临床净收益。结论基于AKR1B10表达的生物学信息放射组学模型在肝细胞癌中具有良好的预后效果。通过直接将影像学表型与HCC的关键分子驱动因素联系起来，该方法为术前风险预测提供了临床应用和生物学解释的工具。

{"title":"A biology-informed radiomics model for prognostication of hepatocellular carcinoma based on AKR1B10 expression","authors":"Hongan Ying , Lili Huang , Weiwen Hong","doi":"10.1016/j.ejro.2026.100725","DOIUrl":"10.1016/j.ejro.2026.100725","url":null,"abstract":"<div><h3>Background</h3><div>Current radiomic models for hepatocellular carcinoma (HCC) prognosis rely on direct correlations between imaging features and clinical outcomes, resulting in limited biological interpretability and restricted clinical applicability. This study explores a novel biology-driven radiomic strategy focusing on AKR1B10. AKR1B10 is a functionally established molecular driver of HCC progression, and the study aims to develop an interpretable prediction model bridging imaging phenotypes and underlying tumor biology.</div></div><div><h3>Methods</h3><div>We analyzed multi-institutional data from The Cancer Genome Atlas (TCGA) and The Cancer Imaging Archive (TCIA). After confirming the prognostic significance of AKR1B10 expression through survival and enrichment analyses, we developed a radiomics model using a cohort of 34 patients with matched computed tomography (CT) images and genomic data. Tumor and peritumoral regions were segmented, and 107 radiomic features were extracted. Feature selection was performed using maximum-relevance-minimum-redundancy (mRMR) and recursive feature elimination (RFE) algorithms, with subsequent model building via logistic regression. The model was evaluated using ROC analysis, calibration curves, and decision curve analysis. Finally, we constructed a prognostic nomogram integrating the radiomics signature with clinical variables.</div></div><div><h3>Results</h3><div>AKR1B10 overexpression was significantly associated with poor overall survival (HR = 2.187, 95 % CI: 1.385–3.454, P < 0.001) and characteristic activation of oncogenic pathways. The radiomics model demonstrated strong performance in predicting AKR1B10 status (AUC = 0.83, 95 % CI: 0.69–0.97), with significant difference in rad-scores between AKR1B10 high- and low-expression groups (P < 0.001). The integrated nomogram showed excellent predictive accuracy for 3-year survival (AUC = 0.85) and provided clinical net benefit across threshold probabilities.</div></div><div><h3>Conclusions</h3><div>The biology-informed radiomics model based on AKR1B10 expression demonstrates strong prognostic performance in hepatocellular carcinoma. By directly linking imaging phenotypes to a key molecular driver of HCC, this approach provides a clinically applicable and biologically interpretable tool for pre-operative risk prediction.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"16 ","pages":"Article 100725"},"PeriodicalIF":2.9,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939271","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}

引用次数: 0

Contemporary 0.55 T MRI to visualize interstitial lung disease – An exploratory study 当代0.55 T MRI显示间质性肺疾病-一项探索性研究

IF 2.9 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

European Journal of Radiology Open

Pub Date : 2026-01-05 DOI: 10.1016/j.ejro.2025.100720

Nadine Bayerl , Claudius S. Mathy , Christina Bergmann , Tobias Bäuerle , Lisa C. Adams , Armin M. Nagel , Jörg H.W. Distler , Teresa Gerhalter , Michael Uder , Rafael Heiss , Stephan Ellmann

Purpose

To evaluate the feasibility of contemporary 0.55 T MRI for visualizing interstitial lung disease (ILD) compared to high-resolution computed tomography (HRCT) in an exploratory first-experience study.

Materials and methods

Thirty participants (mean age 60 ± 13 years; 13 females) with rheumatologic ILD underwent HRCT and 0.55 T MRI within 31 days. MRI protocols included proton-density-weighted turbo-spin-echo sequences (transverse) and T2-weighted short-tau inversion recovery sequences (coronal). Three blinded radiologists independently assessed overall disease extent, ground-glass opacity (GGO), reticulation, and emphysema using a semi-quantitative scoring system. Differences between modalities were tested using Wilcoxon signed-rank tests, and Bland-Altman analysis evaluated systematic bias.

Results

Overall disease extent showed no statistically significant difference between low-field MRI and HRCT (median 22.5 % vs. 24.5 %), with excellent interobserver agreement (MRI ICC = 0.94; HRCT ICC = 0.97). MRI significantly overestimated GGO (13.1 % vs. 9.7 %) and underestimated reticulation (8.1 % vs. 11.4 %) compared to HRCT. Bland-Altman analysis confirmed no systematic bias for overall disease extent but consistent overestimation of GGO and underestimation of reticulation on MRI.

Conclusions

Contemporary 0.55 T MRI showed no statistically significant difference in overall ILD extent compared to HRCT but tended to overestimate GGO and underestimate reticulation. Despite these limitations, 0.55 T MRI represents a promising candidate for future development as a radiation-free alternative for gross disease burden assessment in ILD, warranting further technical refinement before routine clinical use.

目的在一项探索性的首次体验研究中，与高分辨率计算机断层扫描（HRCT）相比，评估当代0.55 T MRI在观察间质性肺疾病（ILD）方面的可行性。材料和方法30例风湿病ILD患者（平均年龄60岁 ± 13岁；13例女性）在31天内行HRCT和0.55 T MRI检查。MRI方案包括质子密度加权涡轮自旋回波序列（横向）和t2加权短tau反转恢复序列（冠状）。三位盲法放射科医师使用半定量评分系统独立评估总体疾病程度、毛玻璃混浊（GGO）、网状和肺气肿。使用Wilcoxon符号秩检验检验模式之间的差异，Bland-Altman分析评估系统偏倚。结果低场MRI和HRCT的总体疾病程度差异无统计学意义（中位数分别为22.5 %和24.5 %），观察者间一致性极好（MRI ICC = 0.94; HRCT ICC = 0.97）。与HRCT相比，MRI显著高估了GGO（13.1 % vs. 9.7 %）和低估了网状（8.1 % vs. 11.4 %）。Bland-Altman分析证实，总体疾病程度没有系统性偏倚，但MRI上一致高估GGO和低估网状。结论与HRCT相比，当代0.55 T MRI在ILD总体范围上无统计学差异，但倾向于高估GGO而低估网状。尽管存在这些限制，0.55 T MRI代表了未来发展的一个有希望的候选方案，作为ILD总疾病负担评估的无辐射替代方案，在常规临床应用之前需要进一步的技术改进。

{"title":"Contemporary 0.55 T MRI to visualize interstitial lung disease – An exploratory study","authors":"Nadine Bayerl , Claudius S. Mathy , Christina Bergmann , Tobias Bäuerle , Lisa C. Adams , Armin M. Nagel , Jörg H.W. Distler , Teresa Gerhalter , Michael Uder , Rafael Heiss , Stephan Ellmann","doi":"10.1016/j.ejro.2025.100720","DOIUrl":"10.1016/j.ejro.2025.100720","url":null,"abstract":"<div><h3>Purpose</h3><div>To evaluate the feasibility of contemporary 0.55 T MRI for visualizing interstitial lung disease (ILD) compared to high-resolution computed tomography (HRCT) in an exploratory first-experience study.</div></div><div><h3>Materials and methods</h3><div>Thirty participants (mean age 60 ± 13 years; 13 females) with rheumatologic ILD underwent HRCT and 0.55 T MRI within 31 days. MRI protocols included proton-density-weighted turbo-spin-echo sequences (transverse) and T2-weighted short-tau inversion recovery sequences (coronal). Three blinded radiologists independently assessed overall disease extent, ground-glass opacity (GGO), reticulation, and emphysema using a semi-quantitative scoring system. Differences between modalities were tested using Wilcoxon signed-rank tests, and Bland-Altman analysis evaluated systematic bias.</div></div><div><h3>Results</h3><div>Overall disease extent showed no statistically significant difference between low-field MRI and HRCT (median 22.5 % vs. 24.5 %), with excellent interobserver agreement (MRI ICC = 0.94; HRCT ICC = 0.97). MRI significantly overestimated GGO (13.1 % vs. 9.7 %) and underestimated reticulation (8.1 % vs. 11.4 %) compared to HRCT. Bland-Altman analysis confirmed no systematic bias for overall disease extent but consistent overestimation of GGO and underestimation of reticulation on MRI.</div></div><div><h3>Conclusions</h3><div>Contemporary 0.55 T MRI showed no statistically significant difference in overall ILD extent compared to HRCT but tended to overestimate GGO and underestimate reticulation. Despite these limitations, 0.55 T MRI represents a promising candidate for future development as a radiation-free alternative for gross disease burden assessment in ILD, warranting further technical refinement before routine clinical use.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"16 ","pages":"Article 100720"},"PeriodicalIF":2.9,"publicationDate":"2026-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939333","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}

引用次数: 0

Performance of deep-learning reconstruction combined with metal artifact reduction algorithm for dual-energy computed tomography angiography in intracranial aneurysm coil embolization 深度学习重建联合金属伪影还原算法在双能ct血管造影颅内动脉瘤线圈栓塞中的应用

IF 2.9 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

European Journal of Radiology Open

Pub Date : 2025-12-12 DOI: 10.1016/j.ejro.2025.100715

Lina Tao , Yuhan Zhou , Limin Lei, Yajie Wang, Xiaoxu Guo, Yifan Guo, Songwei Yue

Purposes

To evaluate the diagnostic confidence in cerebral aneurysm embolization coil follow-up using the deep learning image reconstruction (DLIR) based virtual monoenergetic images (VMI) combined with metal artifact reduction (MAR) algorithm, with a focus on selecting the most optimal scheme.

Methods

A CTA database of 54 patients was prospectively assembled and reconstructed utilizing adaptive statistical iterative reconstruction-Veo(ASIR-V50 %), DLIR at medium and high levels (DLIR-M and H). VMIs were generated within the 40–140 keV range at 10 keV intervals, both with or without MAR. Objective parameters such as artifact index (AI), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were measured. Subjective evaluation was assessed according to the Likert scale scoring method. The post-embolization therapeutic efficacy was assessed by the aneurysm neck, parent artery, and postprocedural complications.

Results

Firstly, 80 keV to 90 keV provided the best objective and subjective scores for a balance between artifact reduction and vascular display. Secondly, the DLIR-H+MAR combination exhibited the highest CNR at 80 keV to 90 keV, while also receiving the best subjective scores. Moreover, the MAR group showed significantly smaller discrepancies in aneurysm neck length and bilateral parent artery diameters compared to the non-MAR group when compared to DSA (p < 0.001). Importantly, the MAR group demonstrated two cases of aneurysm recurrence, four cases of residual filling, ten cases of parent artery stenosis, and four cases of aneurysmal rupture that were undetected by the non-MAR group.

Conclusion

DLIR-H+MAR at 80 keV to 90 keV proved to be the optimal method for visualizing cerebral arteries and mitigating metal artifacts. Simultaneously, it significantly enhanced the efficacy assessment and complication detection of post-embolization aneurysm.

目的评价基于深度学习图像重建（DLIR）的虚拟单能图像（VMI）联合金属伪影还原（MAR）算法在脑动脉瘤栓塞线圈随访中的诊断置信度，选择最优方案。方法采用自适应统计迭代重建- veo （ASIR-V50 %）、DLIR中高水平（DLIR- m和H）对54例患者的CTA数据库进行前瞻性组装和重构。在40-140 keV范围内，以10 keV的间隔生成VMIs，有或没有mar。测量人工指标（AI）、信噪比（SNR）和噪声对比比（CNR）等客观参数。主观评价采用李克特量表评分法。栓塞后的治疗效果通过动脉瘤颈部、载动脉和术后并发症来评估。结果首先，80 至90 keV为伪影还原和血管显示之间的平衡提供了最佳的客观和主观评分。其次，DLIR-H+MAR组合在80 keV至90 keV之间表现出最高的CNR，同时也获得了最好的主观得分。此外，与DSA相比，MAR组在动脉瘤颈长度和双侧载动脉直径上的差异明显小于非MAR组（p <； 0.001）。重要的是，MAR组有2例动脉瘤复发，4例残余填充，10例载瘤动脉狭窄，4例动脉瘤破裂未被非MAR组发现。结论dlir - h +MAR在80 ~ 90 keV范围内是脑动脉显像和减轻金属伪影的最佳方法。同时，显著提高了栓塞后动脉瘤的疗效评估和并发症的发现。

{"title":"Performance of deep-learning reconstruction combined with metal artifact reduction algorithm for dual-energy computed tomography angiography in intracranial aneurysm coil embolization","authors":"Lina Tao , Yuhan Zhou , Limin Lei, Yajie Wang, Xiaoxu Guo, Yifan Guo, Songwei Yue","doi":"10.1016/j.ejro.2025.100715","DOIUrl":"10.1016/j.ejro.2025.100715","url":null,"abstract":"<div><h3>Purposes</h3><div>To evaluate the diagnostic confidence in cerebral aneurysm embolization coil follow-up using the deep learning image reconstruction (DLIR) based virtual monoenergetic images (VMI) combined with metal artifact reduction (MAR) algorithm, with a focus on selecting the most optimal scheme.</div></div><div><h3>Methods</h3><div>A CTA database of 54 patients was prospectively assembled and reconstructed utilizing adaptive statistical iterative reconstruction-Veo(ASIR-V50 %), DLIR at medium and high levels (DLIR-M and H). VMIs were generated within the 40–140 keV range at 10 keV intervals, both with or without MAR. Objective parameters such as artifact index (AI), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were measured. Subjective evaluation was assessed according to the Likert scale scoring method. The post-embolization therapeutic efficacy was assessed by the aneurysm neck, parent artery, and postprocedural complications.</div></div><div><h3>Results</h3><div>Firstly, 80 keV to 90 keV provided the best objective and subjective scores for a balance between artifact reduction and vascular display. Secondly, the DLIR-H+MAR combination exhibited the highest CNR at 80 keV to 90 keV, while also receiving the best subjective scores. Moreover, the MAR group showed significantly smaller discrepancies in aneurysm neck length and bilateral parent artery diameters compared to the non-MAR group when compared to DSA (<em>p</em> < 0.001). Importantly, the MAR group demonstrated two cases of aneurysm recurrence, four cases of residual filling, ten cases of parent artery stenosis, and four cases of aneurysmal rupture that were undetected by the non-MAR group.</div></div><div><h3>Conclusion</h3><div>DLIR-H+MAR at 80 keV to 90 keV proved to be the optimal method for visualizing cerebral arteries and mitigating metal artifacts. Simultaneously, it significantly enhanced the efficacy assessment and complication detection of post-embolization aneurysm.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"16 ","pages":"Article 100715"},"PeriodicalIF":2.9,"publicationDate":"2025-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145749552","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}

引用次数: 0

Dual-parameter risk stratification based on device landing zone calcification and aortic annular perimeter for paravalvular regurgitation after self-expanding TAVR 基于器械着陆区钙化和主动脉环周长的双参数风险分层对自扩张TAVR后瓣旁反流的影响

IF 2.9 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

European Journal of Radiology Open

Pub Date : 2025-12-10 DOI: 10.1016/j.ejro.2025.100719

Jun Shu , Didi Wen , Jingji Xu , Yu Mao , Hui Ma , Jing Zhang , Yao Zhao , Jian Yang , Minwen Zheng

Purpose

The study aimed to identify independent predictors associated with paravalvular regurgitation (PVR) after self-expanding transcatheter aortic valve replacement (SE-TAVR) and to develop a dual-parameter risk stratification model.

Methods

This retrospective study enrolled 292 severe aortic stenosis patients underwent SE-TAVR. PVR severity was assessed pre-discharge. Multivariate logistic regression identified independent predictors of mild/moderate PVR, optimal cutoff values for significant anatomical parameters were determined using receiver operating characteristic (ROC) curve analysis. Patients were subsequently stratified into three risk groups based on these thresholds.

Results

Mild/moderate PVR occurred in 24.0 % of patients. Independent predictors included aortic annular perimeter (OR:1.067, P = 0.015), device landing zone calcific volume (OR:1.006 per 10 mm³, P = 0.025), and presence of sealing skirt (OR:0.412, P = 0.010). The combination of these predictors had a higher discriminative performance (AUC=0.779) than single predictors (P = 0.036, 0.007, and <0.001, respectively), with significant integrated discrimination improvement (integrated discrimination improvement=5.4–6.7 %, P < 0.001). ROC-derived thresholds (device landing zone calcific volume≥1240 mm³ and aortic annular perimeter≥77 mm) stratified patients into three risk groups with progressively increasing PVR incidence: Group A (neither elevate):8.4 %; Group B (either elevated):23.7 %; and Group C (both elevated):48.7 %. Pairwise comparisons confirming differences between Group A vs. B (P = 0.003) and Group B vs. C (P < 0.001). Sealing skirts significantly reduced PVR in Groups A (P = 0.042) but not in Group B and C (P = 0.082 and 0.342).

Conclusion

The dual-parameter model based on device landing zone calcification and aortic annular perimeter significantly enhances PVR risk stratification after SE-TAVR. The dual-threshold model provides a clinically actionable tool for pre-procedural risk stratification and personalized valve selection.

目的探讨经导管主动脉瓣置换术（SE-TAVR）后瓣旁反流（PVR）的独立预测因素，建立双参数风险分层模型。方法回顾性研究292例重度主动脉瓣狭窄患者行SE-TAVR。出院前评估PVR严重程度。多因素logistic回归确定轻度/中度PVR的独立预测因子，采用受试者工作特征（ROC）曲线分析确定重要解剖参数的最佳截止值。随后根据这些阈值将患者分为三个危险组。结果轻/中度PVR发生率为24.0 %。独立预测因素包括主动脉环周长（OR:1.067, P = 0.015）、器械着陆区钙化体积（OR:1.006 / 10 mm³，P = 0.025）和密封裙的存在（OR:0.412, P = 0.010）。这些预测因子组合比单一预测因子具有更高的判别性能（AUC=0.779）（P = 0.036,0.007和<；0.001），具有显著的综合判别改善（综合判别改善= 5.4-6.7 %,P <； 0.001）。roc衍生阈值（器械着陆区钙化体积≥1240 mm³，主动脉环周长≥77 mm）将PVR发病率逐渐增加的患者分为三个危险组：A组（均未升高）：8.4 %；B组（任一升高）：23.7 %；C组（均升高）：48.7 %。两两比较证实了A组与B组（P = 0.003）和B组与C组（P <； 0.001）之间的差异。封裙显著降低了A组的PVR (P = 0.042)，而B组和C组无显著降低（P = 0.082和0.342）。结论基于器械着陆区钙化和主动脉环周长的双参数模型可显著增强SE-TAVR术后PVR风险分层。双阈值模型为术前风险分层和个性化瓣膜选择提供了临床可操作的工具。

{"title":"Dual-parameter risk stratification based on device landing zone calcification and aortic annular perimeter for paravalvular regurgitation after self-expanding TAVR","authors":"Jun Shu , Didi Wen , Jingji Xu , Yu Mao , Hui Ma , Jing Zhang , Yao Zhao , Jian Yang , Minwen Zheng","doi":"10.1016/j.ejro.2025.100719","DOIUrl":"10.1016/j.ejro.2025.100719","url":null,"abstract":"<div><h3>Purpose</h3><div>The study aimed to identify independent predictors associated with paravalvular regurgitation (PVR) after self-expanding transcatheter aortic valve replacement (SE-TAVR) and to develop a dual-parameter risk stratification model.</div></div><div><h3>Methods</h3><div>This retrospective study enrolled 292 severe aortic stenosis patients underwent SE-TAVR. PVR severity was assessed pre-discharge. Multivariate logistic regression identified independent predictors of mild/moderate PVR, optimal cutoff values for significant anatomical parameters were determined using receiver operating characteristic (ROC) curve analysis. Patients were subsequently stratified into three risk groups based on these thresholds.</div></div><div><h3>Results</h3><div>Mild/moderate PVR occurred in 24.0 % of patients. Independent predictors included aortic annular perimeter (OR:1.067, <em>P</em> = 0.015), device landing zone calcific volume (OR:1.006 per 10 mm³, <em>P</em> = 0.025), and presence of sealing skirt (OR:0.412, <em>P</em> = 0.010). The combination of these predictors had a higher discriminative performance (AUC=0.779) than single predictors (<em>P</em> = 0.036, 0.007, and <0.001, respectively), with significant integrated discrimination improvement (integrated discrimination improvement=5.4–6.7 %, <em>P</em> < 0.001). ROC-derived thresholds (device landing zone calcific volume≥1240 mm³ and aortic annular perimeter≥77 mm) stratified patients into three risk groups with progressively increasing PVR incidence: Group A (neither elevate):8.4 %; Group B (either elevated):23.7 %; and Group C (both elevated):48.7 %. Pairwise comparisons confirming differences between Group A vs. B (<em>P</em> = 0.003) and Group B vs. C (<em>P</em> < 0.001). Sealing skirts significantly reduced PVR in Groups A (<em>P</em> = 0.042) but not in Group B and C (<em>P</em> = 0.082 and 0.342).</div></div><div><h3>Conclusion</h3><div>The dual-parameter model based on device landing zone calcification and aortic annular perimeter significantly enhances PVR risk stratification after SE-TAVR. The dual-threshold model provides a clinically actionable tool for pre-procedural risk stratification and personalized valve selection.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"16 ","pages":"Article 100719"},"PeriodicalIF":2.9,"publicationDate":"2025-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145749553","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}

引用次数: 0

Machine Learning for diagnosis of malignant thyroid nodules based on thyroid ultrasound: Systematic review and meta-analysis of studies with external datasets 基于甲状腺超声的机器学习诊断恶性甲状腺结节：外部数据集研究的系统回顾和荟萃分析

IF 2.9 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

European Journal of Radiology Open

Pub Date : 2025-12-10 DOI: 10.1016/j.ejro.2025.100716

Elisa Gatta , Roberto Gatta , Riccardo Morandi , Samuele Isoli , Sara Corvaglia , Simone Vetrugno , Virginia Maltese , Ilenia Pirola , Claudio Casella , Carlo Cappelli

Introduction

Optimizing the diagnostic approach to thyroid nodules remains a crucial challenge. Ultrasound-based risk stratification systems such as EU-TIRADS have shown reasonable sensitivity and specificity. Therefore, we conducted a systematic review and meta-analysis to assess the diagnostic performance of Artificial Intelligence (AI) models in differentiating benign from malignant thyroid nodules on ultrasound data.

Methods

A comprehensive search of PubMed/MEDLINE, Scopus, and Web of Science was performed up to January 1, 2025. Eligible studies included patients with thyroid nodules undergoing ultrasound, where AI-based models were validated against cytological or histological findings. The AI algorithms were developed using different types of ultrasound-derived data, including B-mode images, radiomics features. Pooled sensitivity, specificity, and area under the curve (AUC) were estimated using a hierarchical summary receiver operating characteristic (HSROC) model.

Results

Twenty-seven studies comprising 146,332 patients and over 600,000 ultrasound images met inclusion criteria. Overall, pooled sensitivity was 87 % (95 % CI: 84–89 %) and specificity 83 % (95 % CI: 79–86 %). The summary operating point indicated a sensitivity of 88 % and specificity of 83 %, with an AUC of 91.9 % (95 % CI: 90.0–93.2 %). Although subgroup analysis suggested higher accuracy when cytology was used as the reference standard compared to histology, the mixed-effects meta-regression did not confirm a statistically significant association (p = 0.238 for sensitivity; p = 0.188 for specificity).

Conclusion

AI-based algorithms show excellent diagnostic performance in distinguishing benign from malignant thyroid nodules, with robust validation across external datasets. These findings support the potential integration of AI into clinical thyroid nodule management, although further multicenter, non-Asian, and histology-based studies are warrantee.

Systematic review registration

PROSPERO (CRD420251108149)

优化甲状腺结节的诊断方法仍然是一个关键的挑战。超声风险分层系统如EU-TIRADS显示出合理的敏感性和特异性。因此，我们进行了一项系统综述和荟萃分析，以评估人工智能（AI）模型在超声数据鉴别甲状腺结节良恶性方面的诊断性能。方法综合检索截至2025年1月1日的PubMed/MEDLINE、Scopus、Web of Science数据库。符合条件的研究包括接受超声检查的甲状腺结节患者，其中基于人工智能的模型与细胞学或组织学结果进行了验证。人工智能算法是使用不同类型的超声衍生数据开发的，包括b模式图像，放射组学特征。使用分级汇总接收者工作特征（HSROC）模型估计合并敏感性、特异性和曲线下面积（AUC）。结果27项研究，146332例患者，60多万张超声图像符合纳入标准。总体而言，合并敏感性为87 %(95 % CI: 84-89 %)，特异性为83 %（95 % CI: 79-86 %）。总结操作点灵敏度为88 %，特异性为83 %，AUC为91.9 %（95 % CI: 90.0 ~ 93.2 %）。虽然亚组分析表明，与组织学相比，细胞学作为参考标准的准确性更高，但混合效应荟萃回归并没有证实统计学上显著的关联（p = 0.238敏感性；p = 0.188特异性）。结论基于人工智能的算法在区分甲状腺结节良恶性方面表现出优异的诊断性能，在外部数据集上具有鲁棒性验证。这些发现支持人工智能在临床甲状腺结节治疗中的潜在整合，尽管进一步的多中心、非亚洲和基于组织学的研究是有保证的。系统评价注册号prospero （CRD420251108149）

{"title":"Machine Learning for diagnosis of malignant thyroid nodules based on thyroid ultrasound: Systematic review and meta-analysis of studies with external datasets","authors":"Elisa Gatta , Roberto Gatta , Riccardo Morandi , Samuele Isoli , Sara Corvaglia , Simone Vetrugno , Virginia Maltese , Ilenia Pirola , Claudio Casella , Carlo Cappelli","doi":"10.1016/j.ejro.2025.100716","DOIUrl":"10.1016/j.ejro.2025.100716","url":null,"abstract":"<div><h3>Introduction</h3><div>Optimizing the diagnostic approach to thyroid nodules remains a crucial challenge. Ultrasound-based risk stratification systems such as EU-TIRADS have shown reasonable sensitivity and specificity. Therefore, we conducted a systematic review and meta-analysis to assess the diagnostic performance of Artificial Intelligence (AI) models in differentiating benign from malignant thyroid nodules on ultrasound data.</div></div><div><h3>Methods</h3><div>A comprehensive search of PubMed/MEDLINE, Scopus, and Web of Science was performed up to January 1, 2025. Eligible studies included patients with thyroid nodules undergoing ultrasound, where AI-based models were validated against cytological or histological findings. The AI algorithms were developed using different types of ultrasound-derived data, including B-mode images, radiomics features. Pooled sensitivity, specificity, and area under the curve (AUC) were estimated using a hierarchical summary receiver operating characteristic (HSROC) model.</div></div><div><h3>Results</h3><div>Twenty-seven studies comprising 146,332 patients and over 600,000 ultrasound images met inclusion criteria. Overall, pooled sensitivity was 87 % (95 % CI: 84–89 %) and specificity 83 % (95 % CI: 79–86 %). The summary operating point indicated a sensitivity of 88 % and specificity of 83 %, with an AUC of 91.9 % (95 % CI: 90.0–93.2 %). Although subgroup analysis suggested higher accuracy when cytology was used as the reference standard compared to histology, the mixed-effects meta-regression did not confirm a statistically significant association (p = 0.238 for sensitivity; p = 0.188 for specificity).</div></div><div><h3>Conclusion</h3><div>AI-based algorithms show excellent diagnostic performance in distinguishing benign from malignant thyroid nodules, with robust validation across external datasets. These findings support the potential integration of AI into clinical thyroid nodule management, although further multicenter, non-Asian, and histology-based studies are warrantee.</div></div><div><h3>Systematic review registration</h3><div>PROSPERO (CRD420251108149)</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"16 ","pages":"Article 100716"},"PeriodicalIF":2.9,"publicationDate":"2025-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145749597","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}

引用次数: 0

Automated assessment of right heart function by artificial intelligence: A systematic review and meta-analysis 用人工智能自动评估右心功能：一项系统综述和荟萃分析

IF 2.9 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

European Journal of Radiology Open

Pub Date : 2025-12-08 DOI: 10.1016/j.ejro.2025.100713

Pooya Eini , Homa serpoush , Mohammad Rezayee , Jason Tremblay

Background

Accurate assessment of right ventricular (RV) size and function is critical for managing cardiac diseases but is challenged by the limitations of traditional echocardiography. Artificial intelligence (AI) models offer potential for improving RV assessment, yet their diagnostic accuracy remains uncertain. This systematic review and meta-analysis evaluates the diagnostic accuracy of AI models for predicting RV size and function, synthesizing performance metrics and assessing evidence quality.

Methods

Adhering to PRISMA guidelines, we searched 5 databases up to June 2025 using MeSH and Emtree terms for "Artificial Intelligence," "Right Ventricular Function," and "Right Ventricular Dysfunction." Two reviewers screened studies, extracted data and assessed quality using PROBAST+AI. Pooled estimates were calculated using STATA 18 with MIDAS and METADATA modules. Heterogeneity was explored via subgroup analyses, meta-regression, and sensitivity analyses. Publication bias was assessed using funnel plot.

Results

From 25 studies, 18 provided data for meta-analysis, yielding a pooled sensitivity of 0.85 (95 % CI: 0.73–0.92), specificity of 0.81 (95 % CI: 0.72–0.88), and AUROC of 0.89 (95 % CI: 0.86–0.92). High heterogeneity (I² = 71.63 % for sensitivity, 73.51 % for specificity) was partially explained by algorithm type and study country. The GRADE assessment indicated moderate certainty of evidence due to heterogeneity and bias in 25 % of studies.

Conclusion

AI models show promising diagnostic accuracy for RV assessment, but high heterogeneity and moderate evidence certainty necessitate cautious interpretation and further research.

背景：准确评估右心室（RV）的大小和功能对心脏疾病的治疗至关重要，但传统超声心动图的局限性对其提出了挑战。人工智能（AI）模型为改进RV评估提供了潜力，但其诊断准确性仍不确定。本系统综述和荟萃分析评估了人工智能模型在预测RV大小和功能、综合性能指标和评估证据质量方面的诊断准确性。方法按照PRISMA指南，使用MeSH和Emtree检索截至2025年6月的5个数据库中的“人工智能”、“右心室功能”和“右心室功能障碍”。两名审稿人筛选研究，提取数据并使用PROBAST+AI评估质量。使用带有MIDAS和METADATA模块的STATA 18计算汇总估计值。通过亚组分析、meta回归和敏感性分析探讨异质性。采用漏斗图评估发表偏倚。结果25项研究中，18项提供了荟萃分析的数据，合并敏感性为0.85(95 % CI: 0.73-0.92)，特异性为0.81(95 % CI: 0.72-0.88)， AUROC为0.89（95 % CI: 0.86-0.92）。高异质性（敏感性I²= 71.63 %，特异性I²= 73.51 %）部分由算法类型和研究国家解释。GRADE评估显示，在25% %的研究中，由于异质性和偏倚，证据具有中等确定性。结论人工智能模型对RV评估具有较好的诊断准确性，但异质性高，证据确定性不高，需要谨慎解释和进一步研究。

{"title":"Automated assessment of right heart function by artificial intelligence: A systematic review and meta-analysis","authors":"Pooya Eini , Homa serpoush , Mohammad Rezayee , Jason Tremblay","doi":"10.1016/j.ejro.2025.100713","DOIUrl":"10.1016/j.ejro.2025.100713","url":null,"abstract":"<div><h3>Background</h3><div>Accurate assessment of right ventricular (RV) size and function is critical for managing cardiac diseases but is challenged by the limitations of traditional echocardiography. Artificial intelligence (AI) models offer potential for improving RV assessment, yet their diagnostic accuracy remains uncertain. This systematic review and meta-analysis evaluates the diagnostic accuracy of AI models for predicting RV size and function, synthesizing performance metrics and assessing evidence quality.</div></div><div><h3>Methods</h3><div>Adhering to PRISMA guidelines, we searched 5 databases up to June 2025 using MeSH and Emtree terms for \"Artificial Intelligence,\" \"Right Ventricular Function,\" and \"Right Ventricular Dysfunction.\" Two reviewers screened studies, extracted data and assessed quality using PROBAST+AI. Pooled estimates were calculated using STATA 18 with MIDAS and METADATA modules. Heterogeneity was explored via subgroup analyses, meta-regression, and sensitivity analyses. Publication bias was assessed using funnel plot.</div></div><div><h3>Results</h3><div>From 25 studies, 18 provided data for meta-analysis, yielding a pooled sensitivity of 0.85 (95 % CI: 0.73–0.92), specificity of 0.81 (95 % CI: 0.72–0.88), and AUROC of 0.89 (95 % CI: 0.86–0.92). High heterogeneity (I² = 71.63 % for sensitivity, 73.51 % for specificity) was partially explained by algorithm type and study country. The GRADE assessment indicated moderate certainty of evidence due to heterogeneity and bias in 25 % of studies.</div></div><div><h3>Conclusion</h3><div>AI models show promising diagnostic accuracy for RV assessment, but high heterogeneity and moderate evidence certainty necessitate cautious interpretation and further research.</div></div>","PeriodicalId":38076,"journal":{"name":"European Journal of Radiology Open","volume":"16 ","pages":"Article 100713"},"PeriodicalIF":2.9,"publicationDate":"2025-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145692986","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}

引用次数: 0