ASFNR clinical state-of-practice: For posterior circulation large vessel occlusion stroke, use noncontrast CT to exclude hemorrhage, CT angiography to identify occlusion, and MRI with DWI as reference standard. Imaging scores like PC-ASPECTS help select patients for thrombectomy.
The ACR Appropriateness Criteria 2026 update provides imaging guidelines for suspected or confirmed diffuse lung disease, covering initial evaluation, acute exacerbation, and surveillance.
New expert consensus framework for symptomatic B3 breast lesions: assess concordance, lower excision threshold for discordant/palpable lesions, conservative management for concordant non-atypical lesions, and risk-based surveillance for atypia.
ACR Appropriateness Criteria for indolent lung cancer: For patients with slow-growing subsolid nodules or atypical cysts where surveillance is chosen over intervention, imaging follow-up aims to detect changes that alter management, with modality selection guided by accuracy and…
CTisus presents a short imaging case challenging viewers to identify the most likely diagnosis. An educational quick hit for radiologists and trainees.
Educational video from CTisus reviewing the CT imaging patterns of tumor recurrence in patients with renal cell carcinoma, aimed at practicing radiologists and trainees.
A new paper in RadioGraphics argues that the optimal level of human supervision for radiology AI lies between fully autonomous use and excessive manual checks. The authors report strong consensus on the need for continuous post-deployment monitoring of AI tools.
European Radiology publishes an assessment of whether small intestinal contrast ultrasonography (SICUS) is ready for wider clinical implementation in Crohn's disease management.
Letter to the editor calls for moving beyond generic NLP metrics toward clinically meaningful evaluations of large language models (LLMs) for radiology report quality assurance.
Functional MRI techniques show promise for predicting outcomes in primary sclerosing cholangitis but require validation for routine clinical use.
In a study of 10 UK nuclear medicine radiologists, all explainable AI (XAI) methods significantly increased willingness to adopt an AI staging system for PET/CT lung cancer over a black-box model (p<0.05). Explanations were rated useful (p<0.001), but a depth-usability trade-off…
Radiology is becoming a data-centric platform for prevention and personalized care, but the clinical impact of AI and radiomics depends on robust digital infrastructure and true interoperability across heterogeneous systems, not just better algorithms.
Current evidence does not support LLM use in AD/MCI without clinician oversight. Systematic review of 11 studies: diagnostic models had high internal accuracy (0.94-0.97) but dropped on external validation (3-way accuracy -7% pts; MMSE R² 0.90→0.25).
A random forest model using MRI cartilage thickness and clinical variables predicted shoulder redislocation with 79% cross-validated accuracy (AUC 0.84) in a prospective 42-patient cohort. External validation needed.
Deep learning (DL)-based respiratory-triggered 3D MRCP required 68% fewer effective respiratory cycles than conventional triggered MRCP (19 vs 60 cycles) while yielding similar qualitative image quality and superior peripheral duct visibility over breath-hold compressed sensing…
Random forest model on screening MRI radiomics predicted future HCC sites in cirrhosis (AUC 0.80, 95% CI 0.78–0.81); AUC 0.86 in LR-3 precursor areas. Heatmaps identified the HCC location in 30% of patients.
In a 2873-patient HCM CMR registry, concomitant hypertension conferred a 32% higher long-term mortality risk (HR 1.32, CI 1.04–1.69). LGE extent >3 segments and septal LGE independently predicted death in both groups.
Multicenter stepped-wedge trial (LuAna, Brazil) will test if AI improves GPs’ chest X-ray detection of key findings (consolidation, pneumothorax, etc) vs unaided reads, with radiologist consensus as reference. Results expected 2027.
MARS, a foundation model pretrained on 336,476 multi-sequence MRI scans, ranked first in 41 of 44 downstream tasks (diagnosis, segmentation, etc.) with significant gains, showing generalizability across anatomy and sequences.
CT radiomics models tailored to neoadjuvant chemo or immunochemotherapy better predicted pathologic response in gastric cancer. The NAIC-specific model achieved AUC 0.753 vs 0.659 general; patients getting recommended treatment had higher response (e.g., 40.5% vs 19.4%).
A case of blunt thoracic aortic injury with full-thickness transection 2.5 cm distal to left subclavian artery, pseudoaneurysm, and mediastinal hematoma highlights the lethal impact of imaging/referral delays in resource-limited settings—patient died 12h post-injury.
An NLP-based alert system (FAS) integrated into CT reporting achieved 99.7% accuracy for identifying IVC filter patients and increased eligible patients scheduled for clinic from 21.6% to 58.0% and filter retrieval from 21.6% to 44.0%.
A systematic review of 134 studies on AI-based invasive coronary angiography analysis shows high internal accuracy for vessel segmentation and lesion assessment, but external validation remains limited. AI prediction of invasive physiology indices is a promising area.
FDA 510(k) clearance for Rapid Vessel Occlusion (Rapid VO) by iSchemaview; the software aids detection of vessel occlusions in radiology.
Commercial AI detected pulmonary embolism on CTPA with 89% sensitivity and 99% specificity, while on-call radiology residents had 97.6% sensitivity and 99.2% specificity (p>0.05). For proximal PE, AI sensitivity was 97.6%.
MSK Radiologist Opportunity in Colorado at Diversified Radiology Partner Track Job Opening in Denver, Colorado
A European consensus group published a decalogue of 10 principles to improve undergraduate education in allergy and clinical immunology, addressing a recognized gap in curricula.
An AJNR editorial argues flow diversion has reached biological maturity but not clinical maturity: priorities now shift to standardized imaging, predictive selection, surface-modified devices, and evidence-based antiplatelet regimens like the COATING trial for safer, more durabl…
Acute intermittent porphyria brain MRI shows a spectrum beyond PRES: lobar-cortical pattern (32%), hippocampal (21%), and 25% normal. Dysautonomia (44% of episodes) was associated with ileus (p<0.001) and hyponatremia (p=0.02).
CT-based radiomics predicted gene mutations in clear cell renal cell carcinoma with AUC up to 0.987; AUCs range 0.68-0.91. Models also assess hypoxia, immune markers. Multicenter prospective validation needed.
Diagnostic Imaging's weekly scan rounds up radiology news from July 5 to 11, 2025. (No further details available from source.)
Imaging-only deep learning PET/CT model (AUC 0.78) non-inferior to guideline-recommended Herder model (AUC 0.73, p=0.005) for pulmonary nodule malignancy; matched clinicians. Herder referred more benign nodules to direct treatment (26 vs 3/81). Retrospective, single-center.
In a prospective cohort of 64 patients, 18F-FDG-PET/CECT outperformed CECT alone for detecting local tumor progression after thermal ablation (AUC 0.97 vs. 0.73) and resection (AUC 0.96 vs. 0.69) of colorectal liver metastases. PET/CECT also better detected extrahepatic disease.
A preoperative model integrating a PSMA-PET deep learning score with D'Amico classification and SUVmax improved prediction of biochemical recurrence-free survival after radical prostatectomy (C-index 0.846 training, 0.806/0.774 external validation).
