Among 7 overlapping meta-analyses (AMSTAR 7–10), the highest-quality evidence favored ultrasound-guided corticosteroid injections for carpal tunnel syndrome: better Boston Carpal Tunnel Questionnaire function/symptom scores and fewer adverse events vs landmark-guided; grip stren…
Ellipsoid formula for MRI prostate volume shows excellent agreement with manual segmentation (ICC 0.985, mean absolute difference 5.5 cc) and yields sensitivities 0.86–0.95 and specificities 0.88–0.97 for PSA density cut-offs in 328 men. Underestimation occurs in glands >80 cc b…
Hemorrhage Within Atherosclerotic Plaques: From Binary Marker to Dynamic Process.
A ViT-radiomics fusion model achieved an AUC of 0.95 (95% CI 0.91-0.98) for classifying benign vs malignant lymph nodes on ultrasound in a multicenter study, outperforming radiomics alone (AUC 0.79). Domain adaptation boosted external performance from 0.73 to 0.85.
Sybil, a deep learning model using a single low-dose CT, predicted future lung cancer with area under the curve (AUC) 0.91 at 1 year (95% CI 87.7-94.6) and 0.82 at 6 years, with similar performance in Black and non-Black individuals in 5,883 LDCTs (64.9% Black).
Major pancreatic cyst guidelines had modest long-term malignancy prediction using radiologic features alone (Kyoto AUROC 0.655, ACR 0.639, European 0.623; p<0.001). High-risk classification was concordant; after 5 years of imaging stability, 8-year cancer incidence was 0.71%.
In a meta-analysis of nineteen studies and 2142 patients, 30.2% (95% CI 19.4-43.6) of previously headache-free patients developed new headache after endovascular treatment of an unruptured intracranial aneurysm; 71.4% with pre-existing headache improved, and 5.6% worsened.
In 214 robotic vs 636 manual diagnostic cerebral angiograms, there was no significant difference in room time, procedure time, fluoroscopy time, contrast volume, or radiation dose. Robotic cases required conversion to manual in 18% (39/214). Learning curve spanned ~50 cases.
Photon-counting CT 3D volume rendering agreed well with 2D for intracranial aneurysm dome (ICC 0.83) and neck size (ICC 0.82), but inter-reader reliability was moderate (neck ICC 0.55).
Prostate cancer patients with germline pathogenic variants had more PI-RADS 5 MRI lesions (OR 2.7; CI 1.5–4.9) and grade 3 extra-prostatic extension (OR 3.0; CI 1.5–5.8) than sporadic controls, though PI-RADS ≥3 rates were similar.
ESR Bridges: new developments in imaging and treatment of soft tissue sarcoma-a multidisciplinary view.
A new dual-branch 3D UXFormer model combining 3D UNET and Vision Transformer outperformed existing algorithms for multi-tissue segmentation on breast MRI, achieving higher Dice and IoU scores.
Transarterial microembolization (TAME) with imipenem/cilastatin cut median pain score from 9.0 to 6.0 at 6 months in 42 patients with refractory hip, knee, or shoulder pain; 71.4% achieved a clinically important pain reduction, with only minor complications.
In a two-center pediatric oncology study, 88.5% of families said virtual reality (VR) tumor models should be used for all pre-surgical consent discussions; satisfaction was high, though knowledge scores increased only minimally.
An MRI-based habitat deep learning fusion model predicted response to californium-252 neutron brachytherapy in cervical cancer, achieving an AUC of 0.892 (95% CI 0.821–0.963) in the validation set, significantly outperforming clinical and radiomics models.
For CTD-ILD, HRCT assessment still relies on conventional visual/semiquantitative scoring; quantitative CT and AI are emerging complementary tools. Most evidence in systemic sclerosis, RA, inflammatory myopathy. Validation needed.
New evidence-based guideline for pediatric empyema: routine CT not recommended over ultrasound (conditional recommendation, very low certainty). No clinical, radiological, or laboratory predictors of medical management failure identified.
Semi-automatic ML quantification of hepatobiliary lesion-to-liver enhancement ratio in benign liver tumors (n=23) correlated strongly with manual values (r=0.69–0.87) and effectively differentiated FNH from HCA, showing feasibility.
ASTRO: Elekta expands adaptive radiotherapy portfolio
ASTRO: Ultra-low-dose RT controls myeloma bone pain
ASTRO: Siemens unveils Accela radiotherapy system
ASTRO: Limb-sparing care now standard for extremity sarcoma
Breast Imaging Radiologist job opening in Scranton, Pennsylvania: 1 week on / 1 week off.
ASTRO: Low-dose RT tied to fewer knee replacements
ECOS-Net, a CT-based multitask AI model integrating habitat and deep learning features, predicted early disease control (AUC 0.817) and overall survival (C-index 0.720) in an external test set of immunotherapy-treated HCC, outperforming clinical models.
CAD software performance for cerebral aneurysm and brain metastasis detection improved when institutions were added sequentially in federated learning, with fine-tuning strategies achieving comparable gains to full fine-tuning using fewer parameters.
Across thirty studies, deep learning MRI segmentation of cervical cancer reached Dice coefficients 0.60–0.93; GRADE certainty low due to heterogeneity and limited external validation — promising but not yet routine.
A 2023 fully revised Japanese guideline on diabetic foot ulcers and gangrene uses GRADE methodology and makes weak recommendations for imaging procedures for osteomyelitis, topical negative-pressure wound therapy, and platelet-rich plasma therapy.
A benign-appearing breast mass was classified as BI-RADS 2 by an AI ultrasound tool based on shape. However, shear-wave elastography showed markedly elevated stiffness (max 135.48 kPa), leading to biopsy that revealed high-grade DCIS arising in a fibroadenoma. Malignancy can ret…
In 1707 CT-anchored chest radiographs, vendor A had higher sensitivity for fractures, nodules and pleural effusions, while vendor B had higher specificity. Only 55.4% of CT-confirmed abnormalities were visible on X-ray.