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.
A prototype-guided CT foundation model predicted lung cancer progression-free survival with AUC 0.765 (507 patients) and overall survival with AUC 0.822 (420 patients), outperforming clinical and deep-learning baselines.
CT morphometrics (loss of domain, muscle quality) and AI (automated segmentation, radiomics) show promise for predicting outcomes in ventral hernia repair, but evidence is limited to small retrospective studies without external validation.
In 279 excised breast lesions in patients under 20, none were malignant and 87% were fibroadenomas, yet 55% were Breast Imaging Reporting and Data System (BI-RADS) 4—adult estimates overpredict pediatric cancer risk.
Radiology trainees are using generative AI for clinical and study tasks, but limited institutional guidance risks data bias and cognitive deskilling. A panel advocates embedding 'AI-conscious' teaching instead of prohibitive policies.
Imaging science converts anatomical images into quantitative biomarkers and therapeutic evidence. A quantitative output becomes a biomarker only within a prespecified context of use, and a surrogate endpoint only after rigorous qualification. AI accelerates analysis but doesn't…
In 85,349 MRI exams, 4.8% had motion-related repeat sequences. Brain MRI repeats improved image quality and diagnostic confidence (p<0.001), but 19.5% didn’t help; MRCP repeats showed no significant improvement.
Intraoperative contrast-enhanced ultrasound predicted 12-month recovery after laminoplasty for degenerative cervical myelopathy: a visible central spinal artery pattern predicted better recovery than sparse enhancement (P=0.01), and higher arterial resistance index correlated wi…
Cone-beam CT hepatic arteriography-guided microwave ablation achieved 89% technical success (41/46) in liver lesions unseen on ultrasound, with 4.9% local progression after 11.9-month median follow-up and no major complications.
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.
Locally adapted chest X-ray AI for tuberculosis triage in Uzbekistan reached held-out ROC AUC 0.9902 (95% CI 0.9694–1.0000) but did not outperform the public baseline (p=0.4864); only 15 positive held-out cases limit conclusions.
SegmentAnyMuscle, a universal MRI muscle segmentation model, achieved a mean Dice score of 88.45% on routine sequences and 86.21% in challenging cases across 11 anatomical regions. It correlated with expert measurements (r>0.99).
A proposed framework for an internationally compatible Terminologia Sonographica aims to standardize ultrasound nomenclature, replace metaphors, and improve AI model reproducibility by aligning terms with physical acoustics.
A new CT-based software tool, CT:VQ, generates regional ventilation and perfusion maps from paired inspiratory-expiratory non-contrast chest CT, adding functional context without contrast or radiotracers.
Altis Labs raises $25M for CT AI tool
Remote Radiologists - Day, Swing, Overnight - New Premium Compensation - $600+/hour Job Opening in Nationwide
Ex-NFL QB Matt Hasselback points brain imaging study
Telix inks deal to acquire ITM for $1.65B
RSNA Membership: Built for How Radiologists Work
The proportion of U.S. radiologists devoting a majority of their practice to interventional radiology (IR) rose from 12.6% to 18.5% (2008–2023), while those spending >90% on IR nearly doubled from 4.1% to 8.8%.
***Remote/Hybrid Opportunity*** - Abdominal & Pelvic Imaging Radiologist for The OSU Wexner Medical Center
Five Takeaways for Breast Radiologists from Guidelines on Proliferative Breast Lesions with Atypia and LCIS
Understanding pleural reflections and recesses on imaging is key to localising thoracic disease. This lecture covers visceral and parietal pleura, pleural recesses, and their appearance on CT/X-ray, with practical clinical correlation.
<![CDATA[Is Private Practice Radiology as Important as an Amusement Park?]]>
Deep learning–reconstructed volumetric interpolated breath-hold examination (VIBE) improved image quality, reduced artifacts, and reduced noise versus standard VIBE in 60 women, especially without contrast; signal-to-noise and contrast-to-noise ratios did not differ (p≥0.35).
After percutaneous coronary intervention, elevated radial wall strain (RWSmax ≥13%) predicted 5-year non-target vessel events: adjusted hazard ratio 4.82 (95% CI 3.14-7.40, p<0.0001); AUC 0.73, with non-target revascularization AUC 0.92.
Portable Low-Field MRI of the Brain in Clinic and Community Settings.
Radiomics on sagittal T2 MRI differentiated pediatric ependymomas from other intramedullary spinal tumors in forty patients: balanced accuracy 0.834 (95% CI 0.723-0.955), AUC 0.838, sensitivity 0.909, specificity 0.759. Pilocytic astrocytomas could not be differentiated.
Teleradiology improves subspecialty access and can slash report turnaround times by up to 89%, but uneven internet, incompatible PACS, missing priors, and legal liability (82% in one survey rate it a top barrier) limit its reach.
Understanding AI fundamentals helps radiologists evaluate and implement AI. This review covers convolutional neural networks, U-Nets, vision transformers, and vision-language models for multimodal clinical decision support, plus challenges in validation, bias, and regulation.
