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.
Epsilon Health announced a $28M raise and an AI-native radiology group that processes over 2,500 studies daily, aiming to review 1% of U.S. X-rays by 2026. It integrates AI with board-certified radiologists to speed diagnoses and address the radiology shortfall.
What syndrome does this patient likely have?
AI developer Harrison.ai is backing teleradiology venture Frontier Radiology, providing algorithms that speed report turnaround; the company says this physician-owned model is common and follows US rules.
Radiologists and a general-purpose multimodal AI model show comparable moderate accuracy for pleural effusion volume on CXR (CT reference). Mean absolute errors: radiologists 250-300 mL, AI 264 mL; ICCs 0.64-0.74 vs 0.66.
DLIR with high-frequency kernels reduced noise and improved detectability of small lesions vs iterative reconstruction in CT phantoms, best with DLIR-High. However, for 3mm 200mg/cc lesions, contrast decreased and advantage was lost below 3mGy (chest) & 7mGy (spine).
Combined clinical, radiomics, and deep learning model predicted 5-year growth of pulmonary subsolid nodules with AUC 0.81 (95% CI 0.70–0.92) at external validation.
Fusion model combining vision transformer MRI analysis and MGMT methylation predicted temozolomide resistance in glioblastoma with external validation AUC 0.848 (95% CI 0.673-0.964).
AI triage can sharply cut breast screening reading counts, but counts alone miss time savings and shifted work. Proposed metrics: total active human interpretation minutes per 1000 women screened, work shifted elsewhere, and diagnostic outcomes, via matched-session design.
MRI plays a pivotal role in osteoarthritis by evaluating the whole joint with direct 3D cartilage assessment; advanced compositional MRI and high-resolution morphological imaging will expand clinical utility with emerging disease-modifying therapies.
A cross-modal deep learning model combining CT and MRI achieved AUCs 0.754 and 0.790 in external validation for hepatocellular carcinoma grading, surpassing unimodal approaches (AUC 0.53-0.745).
Correlating Procedural Complexity with the IR Specialization and E&M Billing of Individual Radiologists and Radiology Practices.
Higher CT-derived liver volume (HR 1.42) and right adrenal gland volume (HR 1.34) were independently associated with worse overall survival in clear cell renal cell carcinoma patients; no association with recurrence-free survival after adjustment.
At the 11th International Breast Density Workshop (June 2025, Kaua'i), 85 abstracts from 14 countries highlighted maturing AI imaging biomarkers, a need for validation in diverse populations, and standardized density measures to support equitable screening.
GPT few-shot with RAG achieved F1 0.89-0.94 for named entity extraction vs fine-tuned 0.97 in Spanish mammography reports, but relation extraction lower (0.78 vs 0.99) without task training.
A generative AI model (ChatGPT-4) agreed with multidisciplinary team recommendations in 81% of 100 hepatocellular carcinoma cases, including complete agreement on treatment technique in 53%. Discordance (19%) was linked to older patient age and MDT preference for more aggressive…
Nerve block during CO2 angiography reduced pain (VAS 0 vs 5), improved image quality, and cut procedure time, fluoroscopy, and contrast use vs sedation, per retrospective study.
U.S. FDA cleared INVIA's 4DM under the 510(k) pathway; the radiology committee reviewed the device under product code QIH.
The FDA cleared Omniscient Neurotechnology's Quicktome Software Suite via 510(k) as a radiological image processing device (product code QIH).
Rethinking medical AI: from model performance to governed agentic clinical intelligence.
RayzeBio and Ratio Therapeutics join forces.
Siemens Healthineers completed a $43.6M expansion of its Rockford, TN, PET detector center, the company’s only worldwide manufacturing site, to increase production capacity and meet growing clinical demand for PET/CT in oncology, theranostics, neurology, and cardiology.
Rubric safeguards patient-friendly AI-generated imaging reports
Siemens expands PET detector manufacturing site
<![CDATA[Can MRI-Guided Reconstruction Improve Assessment of Amyloid Brain PET Scans?]]>
AI model on par with experts in finding femoral-neck fractures
Advocate Health Employer Profile - American College of Radiology Career Center
Mecklenburg Radiology Associates Employer Profile
Colorado Imaging Associates and TRA Medical Imaging form Affiliated Radiology, a physician-led network of nearly 200 radiologists, to share picture archiving and communication systems, AI, and cross-reading capacity across the Pacific Northwest and Rocky Mountain regions.
Allegheny Health Network (AHN) Employer Profile
Increasing tube voltage (kVp) reduces patient dose when detector exposure is held constant. Higher energy photons penetrate better, so mAs can be lowered, reducing total photon output despite each photon carrying more energy.
