Background Photon-counting CT (PCCT) offers improved spatial resolution, contrast to noise ratio, and dose efficiency, but its clinical utility remains incompletely defined for breast cancer. Purpose To evaluate the feasibility of PCCT for pretreatment breast cancer assessment through comparisons with MRI, full-field digital mammography (FFDM), and fluorine 18 (18F) fluorodeoxyglucose (FDG) PET/CT. Materials and Methods In this prospective study (March-May 2025), female participants with breast lesions categorized as Breast Imaging Reporting and Data System 4C or higher at US or FFDM underwent breast MRI and multiphasic contrast-enhanced PCCT. 18F-FDG PET/CT was performed in a subset with locally advanced disease. Four radiologists independently evaluated lesion morphologic characteristics, additional findings, and clinical TNM stage. Agreement was analyzed using intraclass correlation coefficients (ICCs) and κ statistics. The diagnostic performance for additional lesions and nodal metastasis was compared with the reference standard (pathologic examination). Results Among 126 participants (mean age, 58.1 years ± 12.3 [SD]), interreader agreement across PCCT, MRI, and FFDM was good to excellent. PCCT agreed with MRI for lesion characterization (κ = 0.57-0.96) and clinical T categorization (κ = 0.86-0.88), with highest agreement with pathologic size (ICC, 0.70-0.81). For 46 pathologically confirmed additional lesions, PCCT was more sensitive than FFDM (difference, 44% [95% CI: 19, 66]) and similar to MRI (difference, 7% [95% CI: -5, 21]). Additionally, 44% (95% CI: 27, 52) of microcalcifications were missed at PCCT versus FFDM. For pathologically confirmed nodal metastasis, PCCT was more sensitive (difference, 10% [95% CI: 1, 20]) and accurate (difference, 6% [95% CI: 1, 11]) than MRI. For clinical N category, PCCT agreed with PET/CT (κ = 0.82 [95% CI: 0.62, 0.96]; n = 19). Two distant metastases identified at PCCT were consistent with 18F-FDG PET/CT and pathologic findings. Conclusion PCCT demonstrated similar performance to MRI for lesion characterization and detection of additional lesions, with better performance for nodal metastasis evaluation; however, detection of microcalcifications was limited. © RSNA, 2026 Supplemental material is available for this article.
Background Chest CT is a primary method for identifying pulmonary nodules, yet interpreting scans remains time-intensive and demanding. Currently, artificial intelligence (AI) is expected to reduce reading times, but the effect of AI on reporting times in this setting is unknown. Purpose To evaluate the impact of a commercial AI software on radiologists' reading time for pulmonary nodule assessment on chest CT scans within a real-world clinical setting. Materials and Methods This retrospective study included patients who underwent chest CT examinations at a tertiary medical center between September 2021 and May 2024. The study period was divided into pre- and post-AI phases. The primary outcome was radiology reporting time. The association between AI implementation and reporting time was evaluated using a multivariable parametric Weibull shared frailty survival model adjusted for reader function, examination type, patient location, and requesting specialty, with clustering at the radiologist level. Interaction analyses assessed heterogeneity across prespecified subgroups. An exploratory extrapolation estimated projected workforce and financial impact. Results This study included 19 433 patients (mean age, 62 years ± 14.2 [SD]; 21 814 men; 39 323 chest CT examinations, 19 190 pre-AI, and 20 133 post-AI). AI implementation was associated with faster report completion (adjusted hazard ratio, 1.17; 95% CI: 1.14, 1.21; P < .001). The adjusted median reporting time decreased from 21.3 minutes pre-AI to 18.2 minutes post-AI (14.6% reduction; P < .001). Heterogeneity was observed across reader function (P < .001), examination type (P = .048), and requesting specialty (P = .03). The largest relative reductions were observed for CT thorax electrocardiogram-gated examinations (-41.1%; P < .001) and thoracic radiologists (-25.0%; P < .001), whereas emergency department examinations showed increased median reporting time (7.1%; P < .001). At institutional scan volumes (approximately 20 000-22 000 chest CT examinations annually), exploratory modeling suggested an approximate reduction of 0.5 full-time equivalent radiologist workload. Conclusion Implementation of commercial AI-assisted pulmonary nodule assessment on chest CT scans reduced radiologist reporting time in a real-world clinical setting. © The Author(s) 2026. Published by the Radiological Society of North America under a CC BY 4.0 license. Supplemental material is available for this article. See also the editorial by Iwasawa in this issue.
Background Ferumoxytol has been described as an alternative contrast agent for vascular suppression in MR neurography (MRN), but its diagnostic utility in patients has yet to be evaluated. Purpose To evaluate the impact of ferumoxytol on vascular suppression, nerve conspicuity, and evaluation of nerve abnormalities at three-dimensional (3D) brachial plexus MRN, compared with noncontrast and gadolinium-enhanced MRN, in participants with suspected Parsonage-Turner syndrome (PTS) or thoracic outlet syndrome (TOS). Materials and Methods This prospective study included participants who underwent 3D MRN with and/or without gadolinium chelate for clinical suspicion of PTS or TOS and subsequently underwent 3D MRN with ferumoxytol (within 3 months of the clinical examination). Two musculoskeletal radiologists qualitatively evaluated 3D short-tau inversion-recovery fast spin-echo scans for the degree of vascular suppression, nerve conspicuity, and presence of nerve abnormalities. Wilcoxon signed-rank or McNemar tests were used for comparing noncontrast and gadolinium-enhanced scans with ferumoxytol-enhanced scans. Results This study included 18 participants (mean age, 42 years ± 15.2 [SD]; 10 men). Ferumoxytol-enhanced scans demonstrated improved vascular suppression compared with both noncontrast scans (both raters, P < .001) and gadolinium-enhanced scans (both P = .04). For rater 2, ferumoxytol-enhanced acquisitions demonstrated improved conspicuity of several nerve segments relative to the noncontrast scan, including segments of the suprascapular (P = .01), axillary (P = .02), and long thoracic nerves (P = .004). The distribution of scores for these nerve segments for rater 1 also favored ferumoxytol-enhanced versus noncontrast scans, but the differences were not statistically significant (all P ≥ .06). There was no evidence of a difference in nerve conspicuity between ferumoxytol-enhanced and gadolinium-enhanced scans (P ≥ .17 for all nerve segments) and also no evidence of discrepancies in abnormal nerve findings between the acquisitions (all P ≥ .48). Conclusion Ferumoxytol improved vascular suppression compared with noncontrast and gadolinium-enhanced 3D short-tau inversion-recovery fast spin-echo sequences in brachial plexus MRN, enhancing the conspicuity of several nerve branches versus noncontrast scans, with similar detection of abnormal nerve findings. © RSNA, 2026 Supplemental material is available for this article.
Combination approaches using systemic immunotherapy agents are now standard of care for patients with advanced-stage hepatocellular carcinoma (HCC), resulting in improved overall survival. However, even with optimal systemic regimens, fewer than 40% of cases respond to treatment, presumably due to resistance mechanisms, including antidrug antibodies and acquired resistance related to alterations in the tumor immune microenvironment (TIME). As a result, new strategies are needed to improve immunotherapeutic efficacy in this setting. Early investigations into the local and systemic effects of yttrium 90 (90Y) radioembolization using resin and glass microspheres demonstrated activation of both innate and adaptive immune systems, leading to sustained therapeutic efficacy in a subgroup of patients with HCC undergoing curative surgical resection after downstaging procedures. Preliminary prospective and retrospective studies have confirmed the safety of combining liver-directed interventions with immunotherapy. Based on these findings, clinical trials are being designed to evaluate the efficacy of different therapeutic strategies combining 90Y radioembolization and immune checkpoint inhibitor therapy. The Society of Interventional Oncology is committed to advancing research on how local-regional therapies influence the TIME and systemic inflammatory response. This review aims to inform the interventional and medical oncology community about essential considerations and strategies for implementing these combination therapies involving 90Y radioembolization and immunotherapy.
Background Suboptimal nodal assessment in rectal cancer hampers pre- and postoperative assessment for treatment planning. A recently proposed standardized nodal evaluation system, Node Reporting and Data System 1.0 (Node-RADS), lacks robust evidence of diagnostic utility. Purpose To validate and optimize Node-RADS against European Society of Gastrointestinal and Abdominal Radiology (ESGAR) criteria in MRI-based nodal assessment in patients with rectal cancer, with subsequent patient-level external validation. Materials and Methods This retrospective study included consecutive patients with rectal cancer who underwent MRI and radical surgery between January 2016 and July 2025. Node-level analyses were performed in patients from center 1, and external patient-level validation was performed in patients from centers 2 and 3. Two radiologists independently evaluated mesorectal nodes according to Node-RADS and ESGAR criteria. The optimal threshold of the Node-RADS score for diagnosing malignancy was determined using the Youden index. Diagnostic performance and interobserver agreement were assessed using area under the receiver operating characteristic curve (AUC) and Cohen κ, respectively. Results This study included 780 patients (median age, 61 years [IQR, 53-68 years]; 477 men). Among 251 patients (1302 nodes) who underwent direct surgery, the optimal threshold of Node-RADS score was 3 or higher (Youden index, 0.65), enabling simplification to a modified model, the Simplified Nodal Assessment Process in Rectal Cancer (recSNAP). recSNAP outperformed the ESGAR criteria at both the node level (AUC, 0.83 vs 0.71; P < .001) and the patient level (AUC, 0.83 vs 0.71; P < .001). Similarly, among 284 patients (918 nodes) who received neoadjuvant chemoradiotherapy before surgery, Node-RADS with a threshold of 4 or higher outperformed the ESGAR criteria (AUC, 0.72 vs 0.66; P < .001). Interobserver agreement for Node-RADS at pre- and posttreatment MRI was substantial (Cohen κ, 0.76 and 0.64, respectively). Conclusion Node-RADS outperformed the ESGAR criteria for MRI-based rectal adenocarcinoma nodal assessment at both the node and patient levels and in external validation, and a simplified version of Node-RADS exhibited similar diagnostic performance. © RSNA, 2026 Supplemental material is available for this article.
MRI continues to be the foundation of multiple sclerosis (MS) diagnosis and management. The latest innovations in technology have paved the way for earlier diagnosis, more informed treatment selection, better means of evaluating the disease course, and personalized therapeutic responses. This review summarizes the most recent developments in advanced MRI for MS, highlighting biomarkers that were newly added in the 2024 revisions of the McDonald criteria, describing the ongoing importance of established imaging measures, reviewing the applications of artificial intelligence in clinical practice, and suggesting future directions. Special emphasis is placed on the official adoption of the optic nerve as a fifth site of dissemination in space, as well as on the central vein sign and paramagnetic rim lesions as key supporting evidence in MS. Additionally, ongoing limitations and challenges in integrating these markers into clinical practice across both academic and community settings are identified. The overall goal is to improve diagnosis and prognostication in MS through imaging approaches that have higher contrast resolution and are more quantitative and biologically informed.
Background Radiomics may preoperatively identify high-grade patterns (HGPs) in lung adenocarcinoma (ADC) and assist in clinical decision-making. Purpose To develop and evaluate a machine learning model based on preoperative contrast-enhanced CT images to predict HGPs and explore the model's prognostic value. Materials and Methods Patients with clinical stage I invasive ADC who underwent surgery (January 2017 to May 2025) were retrospectively enrolled from three centers. Binary (low risk: HGPs < 20%; high risk: HGPs ≥ 20%) and ternary (HGP0: HGPs = 0; HGP1: 0 < HGPs < 20%; HGP2: HGPs ≥ 20%) classification analyses were performed based on the proportion of HGPs. Multivariable logistic regression analysis was used to determine independent predictors of HGPs. XGBoost classifier-based radiomic models and combined models (radiomics-predicted probabilities plus clinical variables plus CT semantic features) were constructed and evaluated for discriminability, calibration ability, and clinical utility. Kaplan-Meier and Cox regression analyses were conducted to identify prognostic factors for overall survival (OS) and recurrence-free survival (RFS). Results A total of 1181 patients (median age, 61 years [IQR, 54-66 years]; 694 female) were allocated to the training (n = 667), internal test (n = 279), and external test (n = 235) sets. The combined model achieved the best discrimination in both binary (training: area under the receiver operating characteristic curve [AUC], 0.87 [95% CI: 0.84, 0.90]; internal test: AUC, 0.80 [95% CI: 0.74, 0.85]; external test: AUC, 0.84 [95% CI: 0.78, 0.90]) and ternary (training: microaverage AUC, 0.80 [95% CI: 0.78, 0.82]; internal test: microaverage AUC, 0.74 [95% CI: 0.70, 0.77]; external test: microaverage AUC, 0.72 [95% CI: 0.68, 0.76]) classification analyses. Model-predicted high-risk group was an independent prognostic factor for both OS (binary: hazard ratio [HR] = 1.98, P =.04; ternary: HR = 2.93, P =.03) and RFS (binary: HR = 3.33, P < .001; ternary: HR = 5.10, P < .001) and was consistently confirmed across subgroup analyses. Conclusion The combined model, integrating clinical variables, CT semantic features, and radiomics-predicted probabilities, effectively predicted high-grade patterns in lung ADC and showed strong potential for prognostic risk stratification. © RSNA, 2026 Supplemental material is available for this article. See also the editorial by Arita and Kocak in this issue.
Background Differentiating acute hemarthrosis from nonhemorrhagic effusion in patients with hemophilia is clinically challenging and may require invasive joint aspiration. Purpose To determine whether T1 or T2 mapping MRI at 3 T can noninvasively differentiate acute hemarthrosis from nonhemorrhagic joint effusion in patients with hemophilia. Materials and Methods In this prospective study, ex vivo validation was first performed using blood dilution series from three control participants to determine blood detection limits and 14-day stability for T1 and T2 mapping at 3 T. Consecutive patients with hemophilia and joint effusion scheduled for arthrocentesis or surgery were then recruited (September 2019 through December 2022) to quantify T2 relaxation times in vivo. Arthrocentesis- or surgery-based classification of joint fluid served as the reference standard. Groups were compared using the Mann-Whitney U test, and diagnostic performance was evaluated using receiver operating characteristic analysis. Results Thirty male participants were evaluated (mean age, 35 years ± 17 [SD]). Ex vivo validation demonstrated that T2 mapping enabled reliable discrimination of blood concentrations down to 0.39%, with results remaining stable over 14 days, whereas reliable discrimination with T1 mapping was limited to dilutions greater than 12.5%. In the clinical sample, in vivo T2 relaxation times were lower in participants with acute hemarthrosis (median, 185 msec; range, 80-322 msec) than in those with nonhemorrhagic effusions (median, 523 msec; range, 368-747 msec; P < .001). A T2 cutoff value of 361 msec differentiated hemorrhagic (14 of 30 [47%]) from nonhemorrhagic (16 of 30 [53%]) effusions with 100% sensitivity (14 of 14 [100%; 95% CI: 77, 100]) and 100% specificity (16 of 16 [100%; 95% CI: 79, 100]; area under the receiver operating characteristic curve, 1.00). Conclusion T2 mapping MRI at 3 T enabled noninvasive differentiation of hemarthrosis and nonhemorrhagic joint effusion in participants with hemophilia, with excellent diagnostic performance. © RSNA, 2026 See also the editorial by Oca Pernas in this issue.
Background Foundation models show promise in medical imaging but remain underexplored in three-dimensional modalities. Despite the adoption of digital breast tomosynthesis (DBT) in breast cancer screening, no dedicated foundation model currently exists for this modality. Purpose To develop and evaluate a foundation model for DBT (DBT-DINO) and assess the impact of domain-specific pretraining across multiple clinical tasks. Materials and Methods This retrospective study used DBT images from Mass General Brigham acquired between March 2011 and February 2024. Self-supervised pretraining was performed using Meta AI's DINOv2 methodology on more than 25 million two-dimensional sections from 487 975 DBT volumes from 27 990 patients. Three downstream tasks were evaluated: (a) breast density classification using 5000 screening examinations, (b) 5-year risk of developing biopsy-proven breast cancer using 106 417 screening examinations, and (c) lesion detection using 393 annotated volumes. The performance of DBT-DINO was compared with that of ImageNet-pretrained DINOv2 baselines using McNemar and DeLong tests. Results A total of 4981 patients (mean age, 57.76 years ± 11.40 [SD]; 4855 female) were included for density classification, 31 561 patients (mean age, 60.09 years ± 10.47; 31 559 female) were included for risk prediction, and 199 female patients were included for lesion detection. For breast density classification, DBT-DINO achieved 79% (786 of 997 examinations) accuracy, outperforming the DINOv2 baseline (73% [728 of 997 examinations]; P < .001). For 5-year breast cancer risk prediction, DBT-DINO had an area under the receiver operating characteristic curve (AUC) of 0.78 and DINOv2 had an AUC of 0.76 (P = .057), showing no evidence of a difference. In lesion detection, DINOv2 had an average sensitivity of 67% (91 of 136 lesions), whereas DBT-DINO had a sensitivity of 62% (84 of 136 lesions) (P = .60), again with no evidence of a difference. Conclusion DBT-DINO demonstrated strong performance in breast density classification; however, there was no evidence of a difference compared with the ImageNet baseline in 5-year breast cancer risk prediction or lesion detection, suggesting that domain-specific pretraining for localized detection tasks required further refinement. © RSNA, 2026 Supplemental material is available for this article. See also the editorial by Wu in this issue.
Background Vascular factors have been associated with osteoarthritis (OA), but the association between popliteal artery wall morphology and symptomatic and radiographic knee OA outcomes over time remains unclear. Purpose To evaluate the associations between MRI-derived measures of popliteal artery wall morphology (localized wall thickening and global wall burden) and symptomatic and radiographic knee OA outcomes. Materials and Methods In this secondary analysis of the Osteoarthritis Initiative (February 2004 to October 2015), baseline knee MRI scans were analyzed to quantify maximum wall thickness and normalized wall index (NWI). Symptomatic OA was assessed using Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain scores, and radiographic OA using Kellgren-Lawrence (KL) grade. Cross-sectional and longitudinal associations were assessed using linear regression and Cox proportional hazards models, respectively, adjusted for demographic and clinical covariates. Results A total of 9333 knees from 4710 participants (mean age, 61.1 years ± 9.2 [SD]; 2753 female) were included. Each 1-mm increase in maximum wall thickness was associated with higher WOMAC pain scores (β = 2.0; P < .001) and KL grades (β = 0.1; P < .001), as well as symptomatic worsening (hazard ratio = 1.10 [95% CI: 1.02, 1.19]; P = .02) and incident radiographic OA (hazard ratio = 1.21 [95% CI: 1.02, 1.44]; P = .044). Each 10% increase in NWI was associated with higher WOMAC pain scores (β = 6.7; P < .001) and symptomatic worsening (hazard ratio = 1.38 [95% CI: 1.20, 1.58]; P < .001) but was inversely associated with KL grade (β = -0.1; P = .001) and incident radiographic OA (hazard ratio = 0.60 [95% CI: 0.43, 0.83]; P = .004). Conclusion Localized popliteal arterial wall thickening was associated with symptomatic OA worsening and incident radiographic OA, whereas diffuse wall remodeling was associated with symptomatic OA worsening and was inversely associated with radiographic outcomes. Clinical trial registration no. NCT00080171
