The 2026 Chinese lung cancer guideline adds AI and opportunistic screening in low-dose CT, expands molecular testing, and updates treatment regimens for NSCLC, including perioperative, consolidation immunotherapy, and targeted therapy for EGFR, ALK, KRAS, HER2, and MET alteratio…
On 76 test cases, the AI framework achieved Cobb angle MAE 2.4°, Lenke curve type accuracy 0.895, and reduced review time from 8.2 to 1.5 min/case.
An AI model (EAGLE) analyzing noncontrast chest CT achieved 90.0% sensitivity for esophageal cancer and 98.5% specificity in external multicenter testing (11,466 patients), and 99.94% specificity in low-dose screening. Could enable opportunistic esophageal cancer screening.
PSAD remained below 0.11 ng/mL² across all ages in 1037 healthy German men, contrasting with continuously rising PSA. 95th percentile reference: PV 51.39 mL, PSA 2.79 ng/mL, PSAD 0.071 ng/mL².
Medicare’s move to MIPS Value Pathways in 2026 will block radiologists from using multispecialty group reporting, likely lowering their payment adjustments. Radiology-only MIPS reporters already averaged a 71.4 quality score, +0.14% adjustment, with 27.3% penalized.
Health Affairs Article IDR ‘Cost’ Claim Inflated Due to Reliance on Flawed QPA Benchmark
A large language model (ChatGPT-4o) tracked radiology residents' clinical reports with >91% precision and recall, then targeted missed pathologies. Median unique pathology exposure rose in every subspecialty evaluated without significantly reducing live case reads, except postgr…
Reply to the Letter to the Editor: Thyroid incidentalomas detected on 18F-fluorodeoxyglucose positron emission tomography with computed tomography in cancer patients: imaging clues for further evaluation.
A lightweight Transformer model with knowledge distillation analyzing spatiotemporal radiomics from ultrafast DCE-MRI achieved AUC 0.959 and 92% accuracy for benign vs malignant breast lesion differentiation, an 8% improvement over conventional temporal radiomics.
Multimodal AI combining dual-sequence CT, clinical, and echocardiographic data detected left ventricular myocardial scar with AUC 0.91, using cardiac MRI as reference in a retrospective study.
Multimodal brain radiomics (CT, CE-T1WI, T2WI) predicted pathological subtypes of lung cancer brain metastases with AUC 0.883. Whole-brain features outperformed tumor-only, and spatial extent was more critical than MRI sequence.
Agentic artificial intelligence (AI) coordinates multiple tools and data sources to move beyond narrow radiology algorithms, but it lacks real-world validation and still faces reliability, scalability, workflow, explainability, bias, regulatory, and governance barriers.
In ACR's Assess-AI registry, LLM prompts for extracting findings from radiology reports achieved 0.985 and 0.997 agreement on development cohorts for ICH and PE, but independent validation is lacking.
Artificial Intelligence in Radiology: Beyond Image Interpretation.
AI-driven CBCT segmentation achieves Dice scores >0.90 with faster processing for alveolar ridge preservation, but no study linked AI output to patient outcomes like implant survival or aesthetics.
Going Beyond O-RADS US to Substratify Ovarian Lesions Classified as Intermediate or High-risk of Malignancy.
Post-training mammography test scores rose from 17.64 to 75.73 (p=0.001) among 62 radiography students after VR training; simulation design and satisfaction/self-confidence scores were high.
A Document-Grounded AI Chatbot as a Bridge to Mentorship in Radiology Workforce Onboarding: A Qualitative Study.
Decommissioning AI: More Than Removing Software.
International Medical Graduates Among NIH-Funded Clinician-Scientists in Radiology: A 14-Year Analysis.
Unregulated Gatekeepers: The Oversight Asymmetry in AI Prior Authorization for Advanced Imaging.
A review of leptomeningeal collateral grading systems on single-phase CTA finds no consensus on which best captures anatomy, with high variability in reliability and feasibility; automated scoring may improve objectivity.
In a nationwide registry of 53,402 neuroendovascular procedures, iatrogenic vessel dissection incidence was 0.65%. Isolated dissection rarely caused worsening (89.5% no mRS decline); deterioration linked to secondary hemorrhage/infarction.
Re-evaluation of Risk Factors and Adverse Event Attribution for IR-Guided Peritoneal Dialysis Catheter Placement.
Practical guide: for vision-language models in radiology, prompt engineering — the instructional and contextual input — is the critical lever on output quality and clinical usefulness. Covers core principles, radiologic examples, and limitations.
Logistic regression model integrating AI-structured DBT features and blood inflammatory biomarkers classified breast masses with AUROC 0.884 (95% CI 0.807-0.950), sensitivity 0.806, specificity 0.878 in testing set.
In a 219-patient retrospective study of oral tongue squamous cell carcinoma, pre-treatment MRI radiomics combined with semantic, clinical, and pathology features predicted recurrence-free and overall survival, outperforming DeepSurv and reclassifying NCCN risk groups.
Radiomics model combining ADC and ASL MRI predicted ATRX status in glioblastoma: AUC 0.711, sensitivity 0.737, specificity 0.690 in 106 patients. SHAP showed balanced modality contribution. Low PPV limits to screening.
MRI shows how hormones affect mood
Radiologist pipeline trails increase in training positions
Collaborate on Quality, Safety and Informatics
Elucid nets $55M in funding round
Telix completes enrollment in BiPASS study
An RSNA Unscripted episode features Jean-Pierre Tasu, MD, PhD, on how global perspective and direct patient contact can improve radiology leadership and patient outcomes.
