Academic Radiology最新文献_第6页

Comparative Evaluation of Advanced Deep Learning, Image-to-Text Models, and Radiomics for Predicting Tumor Budding and Tumor-Stroma Ratio from Breast Ultrasound in Invasive Ductal Carcinoma 高级深度学习、图像-文本模型和放射组学在浸润性导管癌乳腺超声预测肿瘤出芽和肿瘤-间质比率方面的比较评价。

IF 3.9 2区医学 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

Academic Radiology

Pub Date : 2026-01-01 DOI: 10.1016/j.acra.2025.10.020

Esat Kaba , Murat Tören , Berkutay Asan , Yusuf Çubukçu , Göktürk Öztürk , Oğuzhan Okcu , Çiğdem Öztürk , Safiye Sümeyye Çubukçu , Recep Sencer Cinoğlu , Ersan Özer , Fatma Beyazal Çeliker , Nur Hürsoy

Rationale and Objectives

This study aimed to predict tumor budding (TB) and tumor-stromal ratio (TSR), which are important parameters of the tumor microenvironment in invasive ductal carcinoma, from preoperative ultrasound images. To this end, image classification-based deep learning (DL), image-to-text-based DL, and radiomics-based machine learning (ML) approaches were compared.

Materials and Methods

We included 153 patients diagnosed with histopathologically invasive ductal carcinoma. TB and TSR were classified into two groups, “low” and “high,” and separate models were developed for each dataset. Three different methodological approaches were applied: (1) advanced image classification DL models (YOLOv11x-cls, DINOv2, Vision Transformer [ViT]), (2) the Bootstrapping Language-Image Pre-training (BLIP-2) model that converts images to text, and (3) ML algorithms with radiomic features (KNN, SVM, XGBoost). All models were trained on the training set, and their performance was then evaluated on the validation and test sets.

Results

In TB prediction, the XGBoost model demonstrated the most superior performance (AUC: 0.87, accuracy: 0.87 on the validation set; AUC: 0.76, accuracy: 0.78 on the test set). In contrast, image classification–based DL models yielded lower AUC values ranging from 0.55 to 0.71 on the validation set, while the BLIP-2 model achieved an AUC value of 0.67. In the TSR prediction, XGBoost showed the highest discriminatory ability (AUC: 0.92, accuracy: 0.92 in the validation set; AUC: 0.84, accuracy: 0.85 in the test set). In contrast, image classification–based DL models exhibited AUC values ranging from 0.54 to 0.75 in the validation set, while the BLIP-2 model exhibited an AUC of 0.65.

Conclusion

The findings obtained indicate that radiomics-based ML models show promise in non-invasive TB and TSR prediction using ultrasound images in breast cancer. The clinical integration of these approaches could significantly contribute to the development of personalized treatment strategies for invasive ductal carcinoma and enhance patient management.

依据与目的：本研究旨在通过术前超声图像预测浸润性导管癌肿瘤微环境的重要参数——肿瘤出芽（TB）和瘤间质比（TSR）。为此，比较了基于图像分类的深度学习（DL）、基于图像到文本的深度学习（DL）和基于放射组学的机器学习（ML）方法。材料和方法：我们纳入153例经组织病理学诊断为浸润性导管癌的患者。TB和TSR被分为“低”和“高”两组，并为每个数据集开发了单独的模型。采用了三种不同的方法：(1)高级图像分类深度学习模型（YOLOv11x-cls, DINOv2, Vision Transformer [ViT]），(2)将图像转换为文本的Bootstrapping Language-Image Pre-training （BLIP-2）模型，以及(3)具有放射学特征的ML算法（KNN， SVM, XGBoost）。所有模型都在训练集上进行训练，然后在验证集和测试集上对其性能进行评估。结果：在TB预测中，XGBoost模型在验证集上的AUC为0.87，准确率为0.87；在测试集上的AUC为0.76，准确率为0.78。相比之下，基于图像分类的深度学习模型在验证集上的AUC值较低，在0.55 ~ 0.71之间，而blp -2模型的AUC值为0.67。在TSR预测中，XGBoost表现出最高的区分能力（验证集AUC: 0.92，准确率：0.92；测试集AUC: 0.84，准确率：0.85）。相比之下，基于图像分类的深度学习模型在验证集中的AUC值为0.54 ~ 0.75，而BLIP-2模型的AUC值为0.65。结论：基于放射组学的ML模型在非侵袭性结核和乳腺癌超声图像TSR预测中具有应用前景。这些方法的临床整合可以显著促进浸润性导管癌个性化治疗策略的发展，并提高患者的管理。

{"title":"Comparative Evaluation of Advanced Deep Learning, Image-to-Text Models, and Radiomics for Predicting Tumor Budding and Tumor-Stroma Ratio from Breast Ultrasound in Invasive Ductal Carcinoma","authors":"Esat Kaba , Murat Tören , Berkutay Asan , Yusuf Çubukçu , Göktürk Öztürk , Oğuzhan Okcu , Çiğdem Öztürk , Safiye Sümeyye Çubukçu , Recep Sencer Cinoğlu , Ersan Özer , Fatma Beyazal Çeliker , Nur Hürsoy","doi":"10.1016/j.acra.2025.10.020","DOIUrl":"10.1016/j.acra.2025.10.020","url":null,"abstract":"<div><h3>Rationale and Objectives</h3><div>This study aimed to predict tumor budding (TB) and tumor-stromal ratio (TSR), which are important parameters of the tumor microenvironment in invasive ductal carcinoma, from preoperative ultrasound images. To this end, image classification-based deep learning (DL), image-to-text-based DL, and radiomics-based machine learning (ML) approaches were compared.</div></div><div><h3>Materials and Methods</h3><div>We included 153 patients diagnosed with histopathologically invasive ductal carcinoma. TB and TSR were classified into two groups, “low” and “high,” and separate models were developed for each dataset. Three different methodological approaches were applied: (1) advanced image classification DL models (YOLOv11x-cls, DINOv2, Vision Transformer [ViT]), (2) the Bootstrapping Language-Image Pre-training (BLIP-2) model that converts images to text, and (3) ML algorithms with radiomic features (KNN, SVM, XGBoost). All models were trained on the training set, and their performance was then evaluated on the validation and test sets.</div></div><div><h3>Results</h3><div>In TB prediction, the XGBoost model demonstrated the most superior performance (AUC: 0.87, accuracy: 0.87 on the validation set; AUC: 0.76, accuracy: 0.78 on the test set). In contrast, image classification–based DL models yielded lower AUC values ranging from 0.55 to 0.71 on the validation set, while the BLIP-2 model achieved an AUC value of 0.67. In the TSR prediction, XGBoost showed the highest discriminatory ability (AUC: 0.92, accuracy: 0.92 in the validation set; AUC: 0.84, accuracy: 0.85 in the test set). In contrast, image classification–based DL models exhibited AUC values ranging from 0.54 to 0.75 in the validation set, while the BLIP-2 model exhibited an AUC of 0.65.</div></div><div><h3>Conclusion</h3><div>The findings obtained indicate that radiomics-based ML models show promise in non-invasive TB and TSR prediction using ultrasound images in breast cancer. The clinical integration of these approaches could significantly contribute to the development of personalized treatment strategies for invasive ductal carcinoma and enhance patient management.</div></div>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":"33 1","pages":"Pages 61-76"},"PeriodicalIF":3.9,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145427257","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

The METAL Score for Early Stratification of Radiological Response and Survival in Hepatocellular Carcinoma Treated with TACE and Immunotherapy Combinations 肝细胞癌TACE联合免疫治疗放射反应和生存早期分层的METAL评分。

IF 3.9 2区医学 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

Academic Radiology

Pub Date : 2026-01-01 DOI: 10.1016/j.acra.2025.10.027

Qian Chen , Long-Wang Lin , Zhi-Cheng Jin , Bin Xiong , Hong-Tao Hu , Jian-Jian Chen , Guan-Hui Zhou , Hai-Feng Zhou , Rong Ding , Xiao-Li Zhu , Ming Huang , Hai-Bin Shi , Zhong-Wei Zhao , Jian-Song Ji , Wei-Zhu Yang , Guo-Hui Xu , Ai-Bing Xu , Zhou-Chao Hu , Wei-Dong Wang , Chang Zhao , Gao-Jun Teng

Rationale and Objectives

To develop and validate a prognostic stratification tool for patients with unresectable hepatocellular carcinoma (uHCC) receiving transarterial chemoembolization (TACE) combined with targeted and immunotherapy (triple therapy).

Materials and Methods

This multicenter retrospective study included uHCC patients treated with first-line triple therapy at 13 tertiary hospitals from December 2018 to August 2023. A prognostic score was developed (n = 378) based on multivariate Cox regression and validated in an external cohort (n = 148). Radiological response and overall survival (OS) were assessed across different risk strata. Score performance was compared to existing prognostic models and further evaluated in a separate cohort receiving systemic therapy without TACE (n = 232).

Results

A three-variable score (METAL) integrating metastatic burden, alpha-fetoprotein, and albumin-bilirubin grade was developed. Median OS for low-, intermediate-, and high-risk groups was 30.8, 25.2, and 11.6 months in the derivation cohort, and 34.6, 23.0, and 14.8 months in the external validation cohort, respectively (all p<0.001). In both cohorts, confirmed objective response rates (ORR) also declined progressively across risk strata (all p<0.001). The current score was superior to existing TACE- or immunotherapy-related models, with higher discrimination and lower prediction error (C-index, 0.70 and 0.70; integrated Brier score, 0.140 and 0.122 in the derivation and validation cohorts, respectively). The METAL-low score successfully identified a subgroup achieving significantly longer OS (32.6 vs. 23.8 months; p = 0.034) and higher ORR (77.2% vs. 48.1%; p<0.001) with triple therapy compared to systemic therapy alone.

Conclusion

The proposed METAL score enables early identification of ideal uHCC candidates most likely to benefit from triple therapy.

理由和目的：开发和验证一种预后分层工具，用于接受经动脉化疗栓塞（TACE）联合靶向和免疫治疗（三联治疗）的不可切除肝细胞癌（uHCC）患者。材料与方法：本多中心回顾性研究纳入了2018年12月至2023年8月在13家三级医院接受一线三联治疗的uHCC患者。基于多变量Cox回归建立预后评分（n=378），并在外部队列（n=148）中进行验证。评估不同风险层的放射反应和总生存期（OS）。比较现有预后模型的评分表现，并在接受全身治疗而不进行TACE的单独队列中进一步评估（n=232）。结果：一个三变量评分（METAL）整合转移负担，甲胎蛋白和白蛋白胆红素分级。低、中、高风险组的中位OS在衍生组中分别为30.8、25.2和11.6个月，在外部验证组中分别为34.6、23.0和14.8个月（结论：提出的METAL评分能够早期识别最有可能从三联治疗中获益的理想的uHCC候选人。）

{"title":"The METAL Score for Early Stratification of Radiological Response and Survival in Hepatocellular Carcinoma Treated with TACE and Immunotherapy Combinations","authors":"Qian Chen , Long-Wang Lin , Zhi-Cheng Jin , Bin Xiong , Hong-Tao Hu , Jian-Jian Chen , Guan-Hui Zhou , Hai-Feng Zhou , Rong Ding , Xiao-Li Zhu , Ming Huang , Hai-Bin Shi , Zhong-Wei Zhao , Jian-Song Ji , Wei-Zhu Yang , Guo-Hui Xu , Ai-Bing Xu , Zhou-Chao Hu , Wei-Dong Wang , Chang Zhao , Gao-Jun Teng","doi":"10.1016/j.acra.2025.10.027","DOIUrl":"10.1016/j.acra.2025.10.027","url":null,"abstract":"<div><h3>Rationale and Objectives</h3><div>To develop and validate a prognostic stratification tool for patients with unresectable hepatocellular carcinoma (uHCC) receiving transarterial chemoembolization (TACE) combined with targeted and immunotherapy (triple therapy).</div></div><div><h3>Materials and Methods</h3><div>This multicenter retrospective study included uHCC patients treated with first-line triple therapy at 13 tertiary hospitals from December 2018 to August 2023. A prognostic score was developed (<em>n  </em>=148). Radiological response and overall survival (OS) were assessed across different risk strata. Score performance was compared to existing prognostic models and further evaluated in a separate cohort receiving systemic therapy without TACE (<em>n  </em>=0.034) and higher ORR (77.2% vs. 48.1%; <em>p</em><0.001) with triple therapy compared to systemic therapy alone.</div></div><div><h3>Conclusion</h3><div>The proposed METAL score enables early identification of ideal uHCC candidates most likely to benefit from triple therapy.</div></div>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":"33 1","pages":"Pages 47-58"},"PeriodicalIF":3.9,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145446411","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Response to Methodological Comments on Nomogram Models for ALNM in TNBC TNBC中ALNM的Nomogram模型方法学评论。

IF 3.9 2区医学 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

Academic Radiology

Pub Date : 2026-01-01 DOI: 10.1016/j.acra.2025.10.036

Yantong Jin, Yang Wang, Xingyuan Liu, Bo Gao

引用次数: 0

A Retrospective Evaluation of Organ Specific Radiation Doses from Lung Cancer Screening CT: Differences by Race and Sex 肺癌筛查CT对器官特异性辐射剂量的回顾性评价：种族和性别差异。

IF 3.9 2区医学 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

Academic Radiology

Pub Date : 2026-01-01 DOI: 10.1016/j.acra.2025.07.014

Liam McGuirk , William O’Connell , Jesse Chusid , Suhail Raoof , Gerard A. Silvestri , Brett C. Bade , Stuart L. Cohen

Rationale and Objectives

Lung cancer is the leading cause of cancer death in the US. Lung cancer screening (LCS) with low-dose computed tomography (LDCT) has been shown to decrease lung cancer mortality but may be associated with increased malignancy risk secondary to radiation exposure. The purpose of this study is to provide organ dose estimates of radiation from LCS LDCTs by race and sex in a real-world population.

Materials and Methods

This retrospective cohort study included patients who had a LDCT for LCS in a large health system between January 2020 and December 2022. LDCTs and patient characteristics were obtained from radiation dose monitoring software (RDMS); self-reported patient race was obtained from the radiology information system. Organ doses were calculated using best fit Christy phantoms from RDMS. Patient body habitus was approximated with water equivalent diameter (WED). Regression compared organ dose from top three organs by sex and race controlling for body habitus (represented by WED).

Results

A total of 13,431 imaging studies from 9771 patients met the criteria for inclusion. Organs receiving the most radiation exposure were thymus (2.9 [standard deviation (SD) 1.6] mGy), lungs (2.8 [1.6] mGy), breasts (2.8 [1.6] mGy), and heart (2.5 [1.4] mGy). Multivariate regression analysis revealed significant positive associations between body size and lung, thymus, and breast radiation dose; African American race and radiation exposure to lungs, thymus, and breast; and male sex and radiation exposure in lung and thymus dose.

Conclusion

Organ dose estimates of radiation from LDCTs are provided from a real-world population, which identify differences by sex and race when accounting for body habitus. This information can be used to better understand radiation risk from LCS.

基本原理和目的：肺癌是美国癌症死亡的主要原因。低剂量计算机断层扫描（LDCT）肺癌筛查（LCS）已被证明可以降低肺癌死亡率，但可能与辐射暴露后继发恶性肿瘤风险增加有关。本研究的目的是在现实世界的人群中按种族和性别提供LCS ldct辐射的器官剂量估计。材料和方法：本回顾性队列研究纳入了2020年1月至2022年12月在大型卫生系统中接受LDCT治疗LCS的患者。通过辐射剂量监测软件（RDMS）获取ldct和患者特征；从放射学信息系统中获得患者自我报告的种族。器官剂量的计算使用最适合克里斯蒂幻影从RDMS。用水当量直径（water equivalent diameter， WED）近似计算患者体质。回归比较了按性别和种族控制体质（以WED为代表）的前三名器官的器官剂量。结果：9771例患者的13431份影像学研究符合纳入标准。受辐射照射最多的器官是胸腺（2.9[标准差（SD） 1.6] mGy）、肺（2.8 [1.6]mGy）、乳房（2.8 [1.6]mGy）和心脏（2.5 [1.4]mGy）。多因素回归分析显示，体型与肺、胸腺和乳房辐射剂量呈正相关；非裔美国人种族与肺部、胸腺和乳房的辐射暴露；男性肺部和胸腺的辐射剂量。结论：ldct辐射的器官剂量估计值来自真实世界的人群，在考虑身体体质时，可以识别性别和种族的差异。这些信息可以用来更好地了解LCS的辐射风险。

{"title":"A Retrospective Evaluation of Organ Specific Radiation Doses from Lung Cancer Screening CT: Differences by Race and Sex","authors":"Liam McGuirk , William O’Connell , Jesse Chusid , Suhail Raoof , Gerard A. Silvestri , Brett C. Bade , Stuart L. Cohen","doi":"10.1016/j.acra.2025.07.014","DOIUrl":"10.1016/j.acra.2025.07.014","url":null,"abstract":"<div><h3>Rationale and Objectives</h3><div>Lung cancer is the leading cause of cancer death in the US. Lung cancer screening (LCS) with low-dose computed tomography (LDCT) has been shown to decrease lung cancer mortality but may be associated with increased malignancy risk secondary to radiation exposure. The purpose of this study is to provide organ dose estimates of radiation from LCS LDCTs by race and sex in a real-world population.</div></div><div><h3>Materials and Methods</h3><div>This retrospective cohort study included patients who had a LDCT for LCS in a large health system between January 2020 and December 2022. LDCTs and patient characteristics were obtained from radiation dose monitoring software (RDMS); self-reported patient race was obtained from the radiology information system. Organ doses were calculated using best fit Christy phantoms from RDMS. Patient body habitus was approximated with water equivalent diameter (WED). Regression compared organ dose from top three organs by sex and race controlling for body habitus (represented by WED).</div></div><div><h3>Results</h3><div>A total of 13,431 imaging studies from 9771 patients met the criteria for inclusion. Organs receiving the most radiation exposure were thymus (2.9 [standard deviation (SD) 1.6] mGy), lungs (2.8 [1.6] mGy), breasts (2.8 [1.6] mGy), and heart (2.5 [1.4] mGy). Multivariate regression analysis revealed significant positive associations between body size and lung, thymus, and breast radiation dose; African American race and radiation exposure to lungs, thymus, and breast; and male sex and radiation exposure in lung and thymus dose.</div></div><div><h3>Conclusion</h3><div>Organ dose estimates of radiation from LDCTs are provided from a real-world population, which identify differences by sex and race when accounting for body habitus. This information can be used to better understand radiation risk from LCS.</div></div>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":"33 1","pages":"Pages 216-223"},"PeriodicalIF":3.9,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144719082","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

MRI-Based Deep Learning Algorithms vs. Radiologists for Lymph Node Metastasis in Colorectal Cancer: A Systematic Review and Meta-analysis 基于mri的深度学习算法与放射科医生在结直肠癌淋巴结转移方面的对比：系统回顾和荟萃分析。

IF 3.9 2区医学 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

Academic Radiology

Pub Date : 2026-01-01 DOI: 10.1016/j.acra.2025.09.048

Fengguo Wang, Wenjun Deng, Zhilin Zhong

Purpose

This systematic review and meta-analysis aimed to compare the diagnostic performance of MRI-based deep learning (DL) algorithms versus radiologists in detecting lymph node metastasis (LNM) in colorectal cancer (CRC).

Methods

A comprehensive literature search was conducted in PubMed, Embase, and Web of Science up to June 30, 2025 for studies evaluating MRI-based DL algorithms for LNM diagnosis, using histopathology as the reference standard. Pooled sensitivity, specificity, and area under the curve (AUC) were calculated using a bivariate random-effects model. Risk of bias and applicability were assessed using the PROBAST+AI tool. Certainty of evidence was rated with the GRADE approach.

Results

A total of 10 studies met inclusion criteria. Internal validation cohorts (9 studies, n = 1850) showed pooled sensitivity of 0.89 (95% CI: 0.80–0.94), specificity of 0.85 (95% CI: 0.77–0.91), and AUC of 0.93 (95% CI: 0.91–0.95). Radiologists achieved lower pooled sensitivity of 0.65 (95% CI: 0.60–0.71) and specificity of 0.74 (95% CI: 0.71–0.77), with an AUC of 0.76 (95% CI: 0.73–0.80). DL algorithms in internal validation cohorts consistently outperformed junior radiologists in all metrics, and demonstrated higher sensitivity and AUC than senior radiologists(all P<0.05).

Conclusion

MRI-based DL algorithms show promising diagnostic performance in detecting LNM in CRC, with performance generally higher than those reported for radiologists in internal validation cohorts, particularly junior-experienced readers. However, most included studies were retrospective and originated from China, limiting generalizability. Prospective, multicenter studies are warranted to validate these findings across diverse populations.

目的：本系统综述和荟萃分析旨在比较基于mri的深度学习（DL）算法与放射科医生在检测结直肠癌（CRC）淋巴结转移（LNM）方面的诊断性能。方法：以组织病理学为参考标准，综合检索截至2025年6月30日的PubMed、Embase和Web of Science中评估基于mri的DL算法诊断LNM的研究。采用双变量随机效应模型计算合并敏感性、特异性和曲线下面积（AUC）。使用PROBAST+AI工具评估偏倚风险和适用性。证据的确定性用GRADE方法评定。结果：共有10项研究符合纳入标准。内部验证队列（9项研究，n=1850）显示，合并敏感性为0.89 (95% CI: 0.80-0.94)，特异性为0.85 (95% CI: 0.77-0.91)， AUC为0.93 （95% CI: 0.91-0.95）。放射科医生获得了较低的合并敏感性0.65 （95% CI: 0.60-0.71）和特异性0.74 (95% CI: 0.71-0.77)， AUC为0.76 （95% CI: 0.73-0.80）。内部验证队列中的DL算法在所有指标上始终优于初级放射科医生，并且显示出比高级放射科医生更高的灵敏度和AUC(所有结论：基于mri的DL算法在检测CRC中的LNM方面显示出有希望的诊断性能，其性能通常高于内部验证队列中的放射科医生，特别是初级经验丰富的读者。然而，大多数纳入的研究是回顾性的，并且来自中国，限制了普遍性。前瞻性的、多中心的研究有必要在不同的人群中验证这些发现。

{"title":"MRI-Based Deep Learning Algorithms vs. Radiologists for Lymph Node Metastasis in Colorectal Cancer: A Systematic Review and Meta-analysis","authors":"Fengguo Wang, Wenjun Deng, Zhilin Zhong","doi":"10.1016/j.acra.2025.09.048","DOIUrl":"10.1016/j.acra.2025.09.048","url":null,"abstract":"<div><h3>Purpose</h3><div>This systematic review and meta-analysis aimed to compare the diagnostic performance of MRI-based deep learning (DL) algorithms versus radiologists in detecting lymph node metastasis (LNM) in colorectal cancer (CRC).</div></div><div><h3>Methods</h3><div>A comprehensive literature search was conducted in PubMed, Embase, and Web of Science up to June 30, 2025 for studies evaluating MRI-based DL algorithms for LNM diagnosis, using histopathology as the reference standard. Pooled sensitivity, specificity, and area under the curve (AUC) were calculated using a bivariate random-effects model. Risk of bias and applicability were assessed using the PROBAST+AI tool. Certainty of evidence was rated with the GRADE approach.</div></div><div><h3>Results</h3><div>A total of 10 studies met inclusion criteria. Internal validation cohorts (9 studies, <em>n <!-->1850) showed pooled sensitivity of 0.89 (95% CI: 0.80–0.94), specificity of 0.85 (95% CI: 0.77–0.91), and AUC of 0.93 (95% CI: 0.91–0.95). Radiologists achieved lower pooled sensitivity of 0.65 (95% CI: 0.60–0.71) and specificity of 0.74 (95% CI: 0.71–0.77), with an AUC of 0.76 (95% CI: 0.73–0.80). DL algorithms in internal validation cohorts consistently outperformed junior radiologists in all metrics, and demonstrated higher sensitivity and AUC than senior radiologists(all <em>P</em><0.05).</div></div><div><h3>Conclusion</h3><div>MRI-based DL algorithms show promising diagnostic performance in detecting LNM in CRC, with performance generally higher than those reported for radiologists in internal validation cohorts, particularly junior-experienced readers. However, most included studies were retrospective and originated from China, limiting generalizability. Prospective, multicenter studies are warranted to validate these findings across diverse populations.</div></div>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":"33 1","pages":"Pages 22-34"},"PeriodicalIF":3.9,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145314131","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

MRI-based Habitat Imaging Enhances the Preoperative Prediction of Microvascular Invasion in Hepatocellular Carcinoma 基于mri的栖息地成像增强了肝细胞癌微血管侵袭的术前预测。

IF 3.9 2区医学 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

Academic Radiology

Pub Date : 2026-01-01 DOI: 10.1016/j.acra.2025.10.029

Xiaogang Li , Anbin Hu

引用次数: 0

Corrigendum to ‘An MRI-based Radiomics Approach to Improve Breast Cancer Histological Grading’ [Acad Radiol 30 (2023) 1794-1804] “基于mri的放射组学方法改善乳腺癌组织学分级”的勘误表[Acad Radiol 30(2023) 1794-1804]。

IF 3.9 2区医学 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

Academic Radiology

Pub Date : 2026-01-01 DOI: 10.1016/j.acra.2025.08.027

Meng Jiang MD , Chang-Li Li MD , Xiao-Mao Luo MD , Zhi-Rui Chuan MD , Rui-Xue Chen MD , Chao-Ying Jin MD

引用次数: 0

Practical Strategy to Mitigate Overdiagnosis in Asian Low-dose Computed Tomography Lung Cancer Screening 减轻亚洲低剂量ct肺癌筛查中过度诊断的实用策略。

IF 3.9 2区医学 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

Academic Radiology

Pub Date : 2026-01-01 DOI: 10.1016/j.acra.2025.09.044

Yun-Ju Wu , Yi-Chi Hung , En-Kuei Tang , Fu-Zong Wu

Rationale and Objectives

Low-dose computed tomography (LDCT) screening reduces lung cancer mortality but may lead to overdiagnosis and overtreatment, especially among the Asian LDCT screening population presenting with subsolid nodules (SSNs). Thus, we explored whether implementing chest X-ray (CXR) triage as a preliminary screening tool can improve patient selection and reduce unnecessary interventions.

Materials and Methods

This retrospective cohort study compared patients who underwent direct LDCT screening with those who underwent LDCT and CXR triage. Medical records of 506 patients with SSNs ≤3 cm from January 2006 to December 2021 were reviewed. Primary outcomes included lung cancer prognosis, relapse, and histopathologic spectrum of lung adenocarcinoma to assess potential overdiagnosis. The CXR triage group consisted of patients who had both LDCT and CXR within 3 months, with CXR-visible nodules confirmed by surgical pathology. A 1:1 propensity score matching (PSM) was performed using age, sex, and nodule size, resulting in 97 matched pairs for balanced outcome comparison.

Results

After PSM, 194 patients were analyzed. In the direct LDCT group, 37.1% of resected lesions were adenocarcinoma in situ (AIS) or atypical adenomatous hyperplasia (AAH), indicating a higher likelihood of overdiagnosis. Conversely, the CXR triage group had no patients with AIS/AAH, a significant difference. No significant differences were observed in disease-free survival or relapse rates between groups during follow-up, suggesting that CXR triage may reduce overtreatment without negatively impacting short-term prognosis. Multivariate Cox regression revealed that age, sex, nodule size, smoking status, and screening method were not significant predictors of mortality. Kaplan–Meier survival analysis over 15 years showed similarly high survival rates in both groups, with no significant difference.

Conclusion

CXR triage is a preliminary step in LDCT screening for lung cancer, particularly in non-smoking Asian high-risk populations, to reduce unnecessary surgical procedures while preserving diagnostic effectiveness. Further prospective multicenter studies are needed to validate its long-term safety, cost-effectiveness, and acceptability in routine clinical practice.

理由和目的：低剂量计算机断层扫描（LDCT）筛查可降低肺癌死亡率，但可能导致过度诊断和过度治疗，特别是在亚洲LDCT筛查人群中出现亚实性结节（ssn）。因此，我们探讨了将胸部x线（CXR）分诊作为初步筛查工具是否可以改善患者选择并减少不必要的干预。材料和方法：本回顾性队列研究比较了接受直接LDCT筛查的患者与接受LDCT和CXR分诊的患者。回顾了2006年1月至2021年12月506例ssn≤3cm患者的病历。主要结局包括肺癌预后、复发和肺腺癌的组织病理学谱，以评估潜在的过度诊断。CXR分诊组包括在3个月内进行LDCT和CXR检查并经手术病理证实有CXR可见结节的患者。使用年龄、性别和结节大小进行1:1的倾向评分匹配（PSM），得到97对匹配的结果进行平衡比较。结果：经PSM治疗194例患者。在直接LDCT组中，37.1%的切除病灶为原位腺癌（AIS）或非典型腺瘤性增生（AAH），这表明过度诊断的可能性更高。相反，CXR分诊组没有AIS/AAH患者，差异有统计学意义。随访期间，两组无病生存率或复发率无显著差异，提示CXR分诊可减少过度治疗，而不会对短期预后产生负面影响。多因素Cox回归显示，年龄、性别、结节大小、吸烟状况和筛查方法不是死亡率的显著预测因素。超过15年的Kaplan-Meier生存分析显示，两组患者的生存率相似，没有显著差异。结论：CXR分类是LDCT筛查肺癌的初步步骤，特别是在非吸烟的亚洲高危人群中，可以减少不必要的外科手术，同时保持诊断的有效性。需要进一步的前瞻性多中心研究来验证其长期安全性、成本效益和在常规临床实践中的可接受性。

{"title":"Practical Strategy to Mitigate Overdiagnosis in Asian Low-dose Computed Tomography Lung Cancer Screening","authors":"Yun-Ju Wu , Yi-Chi Hung , En-Kuei Tang , Fu-Zong Wu","doi":"10.1016/j.acra.2025.09.044","DOIUrl":"10.1016/j.acra.2025.09.044","url":null,"abstract":"<div><h3>Rationale and Objectives</h3><div>Low-dose computed tomography (LDCT) screening reduces lung cancer mortality but may lead to overdiagnosis and overtreatment, especially among the Asian LDCT screening population presenting with subsolid nodules (SSNs). Thus, we explored whether implementing chest X-ray (CXR) triage as a preliminary screening tool can improve patient selection and reduce unnecessary interventions.</div></div><div><h3>Materials and Methods</h3><div>This retrospective cohort study compared patients who underwent direct LDCT screening with those who underwent LDCT and CXR triage. Medical records of 506 patients with SSNs ≤3cm from January 2006 to December 2021 were reviewed. Primary outcomes included lung cancer prognosis, relapse, and histopathologic spectrum of lung adenocarcinoma to assess potential overdiagnosis. The CXR triage group consisted of patients who had both LDCT and CXR within 3 months, with CXR-visible nodules confirmed by surgical pathology. A 1:1 propensity score matching (PSM) was performed using age, sex, and nodule size, resulting in 97 matched pairs for balanced outcome comparison.</div></div><div><h3>Results</h3><div>After PSM, 194 patients were analyzed. In the direct LDCT group, 37.1% of resected lesions were adenocarcinoma in situ (AIS) or atypical adenomatous hyperplasia (AAH), indicating a higher likelihood of overdiagnosis. Conversely, the CXR triage group had no patients with AIS/AAH, a significant difference. No significant differences were observed in disease-free survival or relapse rates between groups during follow-up, suggesting that CXR triage may reduce overtreatment without negatively impacting short-term prognosis. Multivariate Cox regression revealed that age, sex, nodule size, smoking status, and screening method were not significant predictors of mortality. Kaplan–Meier survival analysis over 15 years showed similarly high survival rates in both groups, with no significant difference.</div></div><div><h3>Conclusion</h3><div>CXR triage is a preliminary step in LDCT screening for lung cancer, particularly in non-smoking Asian high-risk populations, to reduce unnecessary surgical procedures while preserving diagnostic effectiveness. Further prospective multicenter studies are needed to validate its long-term safety, cost-effectiveness, and acceptability in routine clinical practice.</div></div>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":"33 1","pages":"Pages 255-265"},"PeriodicalIF":3.9,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145309998","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

CT-Based Radiomics and Deep Learning for Preoperative Thyroid Nodule Classification: A Systematic Review, Meta-analysis, and Radiologist Comparison 基于ct的放射组学和深度学习用于术前甲状腺结节分类：系统回顾，荟萃分析和放射科医生比较。

IF 3.9 2区医学 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

Academic Radiology

Pub Date : 2026-01-01 DOI: 10.1016/j.acra.2025.09.045

Nima Broomand Lomer , Amir Mahmoud Ahmadzadeh , Mohammad Amin Ashoobi , Sepideh Abdi , Amirhosein Ghasemi , Ali Gholamrezanezhad

Introduction

Computed tomography (CT) can evaluate thyroid cancer invasion into adjacent structures and is useful in identifying incidental thyroid nodules. Computer-aided diagnostic approaches may provide valuable clinical advantages in this domain. Here, we aim to evaluate the diagnostic performance of radiomics and deep-learning methods using CT imaging for preoperative nodule classification.

Methods

A comprehensive search of PubMed, Embase, Scopus, and Web of Science was conducted from inception to June 2, 2025. Study quality was assessed using QUADAS-2 and METRICS. A bivariate meta-analysis estimated pooled sensitivity, specificity, positive and negative likelihood ratios (PLR and NLR), diagnostic odds ratio (DOR), and area under the curve (AUC). Two supplementary analyses compared AI model performance with radiologists and assessed diagnostic utility across CT imaging phases (plain, venous, arterial). Subgroup and sensitivity analyses explored sources of heterogeneity. Publication bias was evaluated using Deek’s funnel plot.

Results

The meta-analysis included 12 radiomics studies (sensitivity: 0.85, specificity: 0.83, PLR: 4.60, NLR: 0.19, DOR: 30.29, AUC: 0.894) and five deep-learning studies (sensitivity: 0.87, specificity: 0.93, PLR: 14.04, NLR: 0.15, DOR: 95.76, AUC: 0.911). Radiomics models showed low heterogeneity, while deep-learning models showed substantial heterogeneity, potentially due to differences in validation, segmentation, METRICS quality, and reference standards. Overall, these models outperformed radiologists, and models using plain CT images outperformed those based on arterial or venous phases.

Conclusion

Radiomics and deep-learning models have demonstrated promising performance in classifying thyroid nodules and may improve radiologists’ accuracy in indeterminate cases, while reducing unnecessary biopsies.

计算机断层扫描（CT）可以评估甲状腺癌对邻近结构的侵犯，并有助于识别偶发甲状腺结节。计算机辅助诊断方法可能在这一领域提供有价值的临床优势。在这里，我们的目的是评估放射组学和深度学习方法在术前结节分类中的诊断性能。方法：综合检索PubMed、Embase、Scopus和Web of Science自成立至2025年6月2日。采用QUADAS-2和METRICS评估研究质量。双变量荟萃分析估计了合并的敏感性、特异性、阳性和阴性似然比（PLR和NLR）、诊断优势比（DOR）和曲线下面积（AUC）。两项补充分析比较了人工智能模型与放射科医生的表现，并评估了CT成像各阶段（普通、静脉、动脉）的诊断效用。亚组分析和敏感性分析探讨了异质性的来源。采用Deek漏斗图评价发表偏倚。结果：meta分析包括12项放射组学研究（敏感性：0.85，特异性：0.83,PLR: 4.60, NLR: 0.19, DOR: 30.29, AUC: 0.894）和5项深度学习研究（敏感性：0.87，特异性：0.93,PLR: 14.04, NLR: 0.15, DOR: 95.76, AUC: 0.911）。放射组学模型显示出较低的异质性，而深度学习模型显示出较大的异质性，这可能是由于验证、分割、METRICS质量和参考标准的差异。总的来说，这些模型优于放射科医生，使用普通CT图像的模型优于基于动脉或静脉相的模型。结论：放射组学和深度学习模型在甲状腺结节分类方面表现良好，可以提高放射科医生在不确定病例中的准确性，同时减少不必要的活检。

{"title":"CT-Based Radiomics and Deep Learning for Preoperative Thyroid Nodule Classification: A Systematic Review, Meta-analysis, and Radiologist Comparison","authors":"Nima Broomand Lomer , Amir Mahmoud Ahmadzadeh , Mohammad Amin Ashoobi , Sepideh Abdi , Amirhosein Ghasemi , Ali Gholamrezanezhad","doi":"10.1016/j.acra.2025.09.045","DOIUrl":"10.1016/j.acra.2025.09.045","url":null,"abstract":"<div><h3>Introduction</h3><div>Computed tomography (CT) can evaluate thyroid cancer invasion into adjacent structures and is useful in identifying incidental thyroid nodules. Computer-aided diagnostic approaches may provide valuable clinical advantages in this domain. Here, we aim to evaluate the diagnostic performance of radiomics and deep-learning methods using CT imaging for preoperative nodule classification.</div></div><div><h3>Methods</h3><div>A comprehensive search of PubMed, Embase, Scopus, and Web of Science was conducted from inception to June 2, 2025. Study quality was assessed using QUADAS-2 and METRICS. A bivariate meta-analysis estimated pooled sensitivity, specificity, positive and negative likelihood ratios (PLR and NLR), diagnostic odds ratio (DOR), and area under the curve (AUC). Two supplementary analyses compared AI model performance with radiologists and assessed diagnostic utility across CT imaging phases (plain, venous, arterial). Subgroup and sensitivity analyses explored sources of heterogeneity. Publication bias was evaluated using Deek’s funnel plot.</div></div><div><h3>Results</h3><div>The meta-analysis included 12 radiomics studies (sensitivity: 0.85, specificity: 0.83, PLR: 4.60, NLR: 0.19, DOR: 30.29, AUC: 0.894) and five deep-learning studies (sensitivity: 0.87, specificity: 0.93, PLR: 14.04, NLR: 0.15, DOR: 95.76, AUC: 0.911). Radiomics models showed low heterogeneity, while deep-learning models showed substantial heterogeneity, potentially due to differences in validation, segmentation, METRICS quality, and reference standards. Overall, these models outperformed radiologists, and models using plain CT images outperformed those based on arterial or venous phases.</div></div><div><h3>Conclusion</h3><div>Radiomics and deep-learning models have demonstrated promising performance in classifying thyroid nodules and may improve radiologists’ accuracy in indeterminate cases, while reducing unnecessary biopsies.</div></div>","PeriodicalId":50928,"journal":{"name":"Academic Radiology","volume":"33 1","pages":"Pages 98-116"},"PeriodicalIF":3.9,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145314130","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Balancing Advocacy and Education: A Complementary Perspective on Resident Unionization in Radiology 平衡宣传与教育：放射科住院医师工会化的互补视角。

IF 3.9 2区医学 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING

Academic Radiology

Pub Date : 2026-01-01 DOI: 10.1016/j.acra.2025.09.047

Aura María Ramírez Cabrera MD, Radiology Resident , María José Veloza Vega MD, Radiology

引用次数: 0