Radiology-Artificial Intelligence最新文献

{"title":"Metrics for Artificial Intelligence in Medicine: A Reference Resource.","authors":"Ricardo A Gonzales, Marcelo Straus Takahashi, Tara Retson, Imon Banerjee, Seong Ho Park, Charles E Kahn","doi":"10.1148/ryai.260070","DOIUrl":"10.1148/ryai.260070","url":null,"abstract":"<p><p>The effective integration of artificial intelligence (AI) systems into clinical medicine depends on comprehensive and transparent performance evaluation; however, the lack of standardized and widely accepted metrics poses challenges for reproducibility and model adoption. A comprehensive, machine-interpretable framework is presented to formalize the nomenclature and descriptions of 207 graphical, matrix, and scalar metrics used to measure AI model performance. The metrics taxonomy, developed as part of the Radiology Ontology of AI Datasets, Models and Projects (ROADMAP), provides a logically structured representation that captures the semantics of AI evaluation metrics, supports reasoning over metric classes, and enables automated completeness checks for AI model reporting. For each metric, the taxonomy incorporates a definition and citations to authoritative reference sources; where applicable, the taxonomy also includes synonyms, abbreviations, alternate language forms, mathematical formulae, and numerical bounds. The taxonomy supports evaluation of models operating on structured data, medical images, audio signals, and/or unstructured text. Logical axioms link each metric to one or more of 18 AI model performance criteria, including classification, calibration, image segmentation, and text analysis. By harmonizing terminology and enabling structured queries, ROADMAP's taxonomy of AI performance metrics facilitates model comparison, bias detection, and selection of appropriate evaluation methods across diverse datasets and clinical tasks. <i>Supplemental material is available for this article.</i> © RSNA, 2026 See also special report on ROADMAP ontology.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e260070"},"PeriodicalIF":20.1,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147436225","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Bridging Industry and Academia: Proceedings from the 2025 Academy Roundtable on AI Implementation in Medical Imaging.","authors":"Alexander J Towbin, Woojin Kim, Nina Kottler","doi":"10.1148/ryai.250671","DOIUrl":"10.1148/ryai.250671","url":null,"abstract":"<p><p>Despite rapid advancements in artificial intelligence (AI) for medical imaging, widespread clinical adoption remains limited. In March 2025, the Academy for Radiology & Biomedical Imaging Research convened a cross-sector roundtable to examine operational and structural challenges in AI development and implementation. Researchers, department leaders, government representatives, and industry executives participated in a structured two-stage discussion using the AI lifecycle and a simplified failure modes and effects analysis (sFMEA) framework. In the first stage, attendees examined each phase of the AI lifecycle to identify domains where implementation barriers arise. In the second stage, mixed stakeholder groups applied a qualitative sFMEA approach to analyze process vulnerabilities within those domains and discuss mitigation approaches. This article summarizes the session design, synthesizes key domains, and presents illustrative mitigation approaches across five areas: governance, use cases, implementation, cost, and regulation. The discussion identified recurring challenges related to fragmented priorities, infrastructure constraints, and regulatory complexity, as well as the need for clearer governance structures and more consistent evaluation processes to improve coordination across stakeholders. <b>Keywords:</b> Health Policy and Practice, Impact of AI on Education <i>Supplemental material is available for this article.</i> © RSNA, 2026.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250671"},"PeriodicalIF":20.1,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147843065","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Clinic-aligned Dual Distillation of Video and Image Foundation Models for Automated Breast Cancer US Diagnosis.","authors":"Chengqian Zhao, Yijie Dong, Xuan Xie, Guoqing Wu, Feiyu Yin, Wenwen Zeng, Pengfei Song, Yonghuang Wu, Lina Fu, Bingqian Nie, Hong Ding, Jianqiao Zhou, Jinhua Yu","doi":"10.1148/ryai.250600","DOIUrl":"10.1148/ryai.250600","url":null,"abstract":"<p><p>Purpose To evaluate the clinical applicability of the US dual-distillation model (USDist) through comparative analysis with state-of-the-art models, ablation analysis of dual-distillation components, and assessment on portable US devices. Materials and Methods This retrospective multicenter study evaluated USDist using US video datasets collected from 16 medical centers (August 2016-December 2024) and two independent public datasets. The model integrates spatiotemporal dual-distillation and dynamic-static feature fusion to transfer feature representations from video and image foundation models. Diagnostic performance was assessed using receiver operating characteristic (ROC) analysis with DeLong testing and Holm correction, along with evaluation of computational efficiency and qualitative feature visualization. Results A total of 5033 patients were analyzed (mean age ± SD, 49 years ± 12; 5031 female). USDist achieved a mean area under the ROC curve (AUC) of 0.95 (95% CI: 0.93, 0.97) for breast cancer diagnosis across datasets. In the main cohort, USDist outperformed foundation models while using 98.3% fewer parameters. Across multicenter datasets, diagnostic performance was similar to that of foundation models (all <i>P</i> < .05). On a portable US device, USDist maintained an AUC of 0.92 (95% CI: 0.86, 0.95) with 4.1% of the computational cost of full-parameter fine-tuning. Conclusion USDist demonstrated high diagnostic performance for automated breast cancer diagnosis, with substantial parameter reduction compared with foundation models, and maintained performance across multicenter and portable US settings. <b>Keywords:</b> Ultrasound, Breast Cancer, Dual Distillation Model <i>Supplemental material is available for this article.</i> © RSNA, 2026 See also commentary by Whitman and Cohen in this issue.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250600"},"PeriodicalIF":20.1,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147634336","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Commercial AI Model Diagnostic Accuracy for Intracranial Large- and Medium-Vessel Occlusion at Emergency CT Angiography.","authors":"Henrik Andersson, Björn Hansen, Johan Wassélius","doi":"10.1148/ryai.250749","DOIUrl":"10.1148/ryai.250749","url":null,"abstract":"<p><p>The diagnostic accuracy of AIDOC-VO, the first commercial artificial intelligence (AI) tool for intracranial large- and medium-vessel occlusion (LVO and MeVO) detection at head and neck CT angiography (CTA), was evaluated in a multicenter emergency setting. A prospective diagnostic-accuracy study of 3031 adult CTA examinations (mean age ± SD, 67.3 years ± 16.4; 1549 females) acquired March-July 2024 across a 10-hospital region was performed. The AI model was compared with clinical radiology reporting. Examinations flagged positive or doubt by either the AI model or report underwent blinded rereading for reference-standard establishment. Of 3031 CTA examinations, valid AI model output was yielded for 2804 (92.5%), of which 224 of 2804 (8.0%) had vessel occlusion (VO) on reference-standard reading. For VO detection within intended use (218 of 224), sensitivity was 81.7% (178 of 218) (clinical report: 81.2% [177 of 218]; <i>P</i> = .91), and specificity was 99.6% (2569 of 2580) (clinical report: 99.3% [2561 of 2580]; <i>P</i> = .12). LVO sensitivity was 92.8% (64 of 69) (clinical report: 87.0% [60 of 69]; <i>P</i> = .42) and MeVO sensitivity was 76.1% (121 of 159) (clinical report: 79.2% [126 of 159]; <i>P</i> = .55). The AI model identified VOs missed by radiologists in 42 examinations, for an enhanced detection rate of 18.8% (42 of 224; 15 per 1000 CT angiograms), and generated 11 false alerts (3.9 per 1000 CT angiograms). Performance did not differ significantly from clinical radiology reporting. <b>Keywords:</b> CT, CT-Angiography, CNS, Ischemia/Infarction, Stroke, Diagnosis, Classification, Application Domain, Arteries, Artificial Intelligence, Large Vessel Occlusion, Medium Vessel Occlusion <i>Supplemental material is available for this article.</i> © The Author(s) 2026. Published by the Radiological Society of North America under a CC BY 4.0 license.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250749"},"PeriodicalIF":20.1,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147692616","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development of an Integrated Deep Learning Approach for Detecting Fetal Brain Abnormalities at Routine Second Trimester US Scan: A Multicenter Study.","authors":"Ruben Ramirez Zegarra, Alessandra Familiari, Andrea Dall'Asta, Chiara Di Ilio, Beatrice Valentini, Tiziana Fanelli, Paolo Volpe, Monica Minopoli, Basky Thilaganathan, Edwin Quarello, Ricciarda Raffaelli, Julia Binder, Veronica Falcone, Gianpaolo Grisolia, Giuseppe Rizzo, Gianluca Gragnaniello, Huong E Tran, Luca Boldrini, Tullio Ghi","doi":"10.1148/ryai.250737","DOIUrl":"10.1148/ryai.250737","url":null,"abstract":"<p><p>Purpose To develop and validate an anatomy-aware, two-stage, end-to-end deep learning pipeline for fetal brain abnormality automated detection on standardized second-trimester brain US images. Materials and Methods This retrospective multicenter study included 319 fetal brain images (218 normal, 101 abnormal) between 19 weeks ± 0 and 23 weeks ± 6 of gestation from nine international fetal medicine centers, each with paired standard transventricular and transcerebellar axial plane images acquired at second-trimester US between January 2010 and December 2022. Abnormalities were confirmed by neonatal imaging or autopsy. Images were annotated for six key brain regions by two experienced fetal medicine specialists. An anatomy-aware, two-stage deep learning pipeline was developed, consisting of a You Only Look Once version 5-based object detector followed by a classification network using a Mini-ResNet feature extractor within a HexaNet architecture. The pipeline classified each image as normal or abnormal. Object detection performance was evaluated using mean average precision at an intersection-over-union threshold of 0.5 (mAP@0.5). Classification performance was assessed using the area under the receiver operating characteristic curve, sensitivity, specificity, and F1 score. Results The object detection model achieved a mAP@0.5 of 0.93 (95% CI: 0.90, 0.96) on the test dataset. The classification model achieved an area under the receiver operating characteristic curve of 0.96 (95% CI: 0.90, 1.00), a sensitivity of 87% (95% CI: 67, 100 [13 of 15]), a specificity of 91% (95% CI: 79, 100 [29 of 32]), and an F1 score of 0.84 (95% CI: 0.67, 0.96) for distinguishing normal from abnormal fetal brain images. Conclusion The developed model achieved high diagnostic performance for the detection of brain anomalies at routine fetal second-trimester US. <b>Keywords:</b> Artificial Intelligence, Machine Learning, Fetal Neurology, Neurosonography, Fetal Brain Malformation, YOLOv5, Neural Networks <i>Supplemental material is available for this article.</i> © RSNA, 2026 See also commentary by Rafful in this issue.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250737"},"PeriodicalIF":20.1,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147634352","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Bone Metastasis Detection at CT with Deep Learning Models Trained Using Multicenter, Multimodal Reference Standards: Development and Evaluation.","authors":"Jung-Oh Lee, Dong Hyun Kim, Hee-Dong Chae, Eugene Lee, Ji Hee Kang, Ji Hyun Lee, Hyo Jin Kim, Jiwoon Seo, Jee Won Chai","doi":"10.1148/ryai.250283","DOIUrl":"10.1148/ryai.250283","url":null,"abstract":"<p><p>Purpose To develop and validate deep learning models for detecting bone metastases on abdominal and thoracic CT scans, considering lesion visibility, and to compare model performance against human readers. Materials and Methods This retrospective multicenter study included CT scans in patients with bone metastases at four medical centers (August 2013-October 2021). MRI and PET/CT served as reference standards to categorize lesions as visible, indeterminate, or invisible based on CT visibility. Two nnU-Net models were trained: model 1 with only CT-visible metastases and model 2 with both visible and indeterminate metastases. Lesion-level performance was evaluated using precision and recall. Scan-level performance was evaluated using area under the receiver operating characteristic curve. Model performance was compared with that of three musculoskeletal radiologists and three radiologists in training. Results A total of 502 CT scans in 332 patients (mean age ± SD, 64.2 years ± 11.2; 171 male) with 4999 bone metastases were included. Although lesion-level precision was similar between both models (model 2: 80.1%; model 1: 78.8%; <i>P</i> = .41), model 2 achieved higher recall overall (41.8% vs 33.9%; <i>P</i> < .001) and among visible lesions (53.6% vs 44.7%; <i>P</i> < .001). Both models' precision exceeded that of radiologists in training (66.6%; <i>P</i> < .003) and musculoskeletal radiologists (66.5%; <i>P</i> < .004). Only model 2 achieved recall comparable with that of both radiologists in training (39.4%; <i>P</i> = .37) and musculoskeletal radiologists (43.8%; <i>P</i> = .47), as well as a comparable scan-level area under the receiver operating characteristic curve (0.80 [95% CI: 0.67, 0.90]; <i>P</i> > .05). Conclusion The deep learning model trained with multimodal reference standards achieved expert-level bone metastasis detection performance at body CT. <b>Keywords:</b> CT, Abdomen/GI, Skeletal/Axial, Skeletal-Appendicular, Thorax, Metastases, Technology Assessment, Comparative Studies, Segmentation, Supervised Learning, Convolutional Neural Network (CNN), Bone Metastasis, Body CT, Deep Learning, Multimodal Reference Standards <i>Supplemental material is available for this article.</i> © RSNA, 2026 See also commentary by Khosravi in this issue.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250283"},"PeriodicalIF":20.1,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147515310","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Cognitively Biased Prompt Effects on Large Language Model Accuracy for Radiology Board-style Examination Questions.","authors":"Nicholas T Dietrich, Dhruv Patel, Joseph Bellissimo, Christopher T Loh, Pascal N Tyrrell","doi":"10.1148/ryai.250585","DOIUrl":"10.1148/ryai.250585","url":null,"abstract":"<p><p>Large language models (LLMs) are increasingly explored for radiology-related applications, yet their vulnerability to cognitive biases remains undercharacterized. The aim of this study was to investigate whether targeted prompts exploiting cognitive biases degrade LLM accuracy on radiology board-style questions. Ten contemporary LLMs were evaluated on 200 text-based and 200 multimodal American Board of Radiology examination-style questions under baseline and three cognitive bias prompts: authority bias prompts (ABPs), complexity bias prompts (CBPs), and anchoring bias prompts (AnBPs). Two mitigation approaches-a prompt bias audit and a one-shot mitigation strategy-were also evaluated. Under baseline prompts, models achieved a mean accuracy ± SD of 84.8% ± 5.5 (154-186 of 200) for text-based and 59.5% ± 7.7 (101-143 of 200) for multimodal questions. All models showed reduced accuracy to cognitively biased prompts, with ABP, CBP, and AnBP yielding absolute declines of 21.1%, 10.1%, and 4.4%, respectively, for text questions (<i>P</i> < .001 for each), and 44.9%, 44.4%, and 39.6%, respectively, for multimodal questions (<i>P</i> < .001 for each). The prompt bias audit increased accuracy by 5.6% for text-based and 15.8% for multimodal questions, whereas the one-shot mitigation yielded gains of 4.0% for text questions and 24.9% for multimodal questions. These findings demonstrate that LLMs are susceptible to cognitively biased inputs. <b>Keywords:</b> Technology Assessment, Use of AI in Education, Social Implications <i>Supplemental material is available for this article.</i> © RSNA, 2026 See also commentary by Tayebi Arasteh and Truhn in this issue.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250585"},"PeriodicalIF":20.1,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147692551","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-Institutional Annotated Multiparametric MRI Dataset of Pediatric High-Grade Gliomas.","authors":"Anahita Fathi Kazerooni, Zhifan Jiang, Deep Gandhi, Nastaran Khalili, Xinyang Liu, Wenxin Tu, Jeffrey B Ware, Ariana M Familiar, Bhavyasri Vunnava, Anna Zapaishchykova, Aaron S McAllister, Mariana Sanchez-Montano, Nakul Sheth, Khanak K Nandolia, Hollie Anne Lai, Julija Pavaine, Sanjay P Prabhu, Debanjan Haldar, Sanaz Varshochi, Sina Bagheri, Hannah Anderson, Shuvanjan Haldar, Neda Khalili, Anurag Gottipati, Ibraheem Salman Shaikh, Ethan Castellino, Avani Mangoli, Harsh Gohil, Nazanin Maleki, Justin Low, Trent Hummel, Roger J Packer, Andrea Franson, Phillip B Storm, Spyridon Bakas, Evan Calabrese, Mariam Aboian, Peter de Blank, Benjamin H Kann, Brian Rood, Adam C Resnick, Ali Nabavizadeh, Arastoo Vossough, Marius George Linguraru","doi":"10.1148/ryai.250902","DOIUrl":"10.1148/ryai.250902","url":null,"abstract":"","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250902"},"PeriodicalIF":20.1,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13227329/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147783292","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Beyond the Age Gap: Longitudinal Aging Velocity as a Dynamic Biomarker in Chest Radiography.","authors":"Paul S Babyn","doi":"10.1148/ryai.260358","DOIUrl":"10.1148/ryai.260358","url":null,"abstract":"","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":"8 3","pages":"e260358"},"PeriodicalIF":20.1,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147783297","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Longitudinal DCE MRI Vascular Textures: Radiologic and Biologic Insights for pCR Prediction in HER2-Negative Breast Cancer.","authors":"Xinzhi Teng, Junjie Ma, Jiang Zhang, Miaoqing Zhao, Xiangjiao Meng, Yong Yin, Haonan Xiao, Qingpei Lai, Xinyu Zhang, Yufeng Jiang, Jing Cai","doi":"10.1148/ryai.250734","DOIUrl":"10.1148/ryai.250734","url":null,"abstract":"<p><p>Purpose To develop a pathologic complete response (pCR) prediction model for human epidermal growth factor receptor 2 (HER2)-negative breast cancer by analyzing longitudinal changes in dynamic contrast-enhanced MRI (DCE MRI)-derived vascular textures. Materials and Methods Retrospective baseline and midtreatment DCE MRI data from I-SPY2 (May 2010-November 2016) and ACRIN 6698 (August 2012-January 2015) trials were used for development and internal tests (ClinicalTrials.gov no. NCT01042379). An independent hospital cohort (December 2023-December 2024) served as the external test. Image Biomarker Standardization Initiative-standardized vascular textures were extracted from the functional tumor volume (FTV). The DCE MRI vascularization-based response tracking (DCE-VASC-TRACK) model incorporated repeatable vascular texture changes associated with pCR at surgery, alongside hormone receptor status, age, baseline FTV, and midtreatment FTV change. Performance was evaluated using the area under the receiver operating curve (AUC). Biologic associations were explored using gene set enrichment analysis. Results The study included 417 (development), 162 (internal test), and 167 (external test) women (mean ± SD ages: 49 years ± 10, 48 years ± 10, 48 years ± 10, respectively). Changes in two features-complexity and run-length variance-were significantly associated with pCR (adjusted odds ratios per SD increase: 2.13 [95% CI: 1.75, 2.63] and 2.34 [95% CI: 1.87, 2.92]; <i>P</i> < .001). In the external test cohort, DCE-VASC-TRACK outperformed the FTV-based model (AUC, 0.86 [95% CI: 0.79, 0.92] vs 0.72 [95% CI: 0.62, 0.79]; <i>P</i> < .001). Vascular textures showed enrichment in angiogenesis, protein secretion, and transforming growth factor-β signaling pathways compared with clinical factors. Conclusion Incorporating DCE MRI vascular texture dynamics at midtreatment significantly improved pCR prediction compared with clinical and functional tumor volume features alone. <b>Keywords:</b> Breast Cancer, Molecular Imaging, Dynamic Contrast-enhanced MRI, Radiogenomics <i>Supplemental material is available for this article.</i> © The Author(s) 2026. Published by the Radiological Society of North America under a CC BY 4.0 license. Clinical trial registration no. NCT01042379 See also commentary by Schnitzler in this issue.</p>","PeriodicalId":29787,"journal":{"name":"Radiology-Artificial Intelligence","volume":" ","pages":"e250734"},"PeriodicalIF":20.1,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147515356","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}