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.
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 multiphase CT angiography spot sign identified high-risk intracerebral hemorrhage: hematoma expansion occurred in 25/51 (49%) phase-1, 13/28 (46%) phase-2, and 4/9 (44%) phase-3 spots versus 12/71 (17%) with no spot sign.
AI chest radiograph flagged right apical pneumothorax after left-sided pacemaker; chest CT definitively excluded pneumothorax. Alert was clinically discordant: left-sided access, normal oxygen saturation, hemodynamic stability. AI is decision support, not autonomous diagnosis.
The third molar maturity index (I3M) on panoramic radiographs showed excellent discrimination for classifying age ≥18 vs. <18 years across five populations (AUC 0.917–0.942). Optimal cutoff values varied from 0.126 (Croatia) to 0.223 (China), and machine learning with leave-one-…
Infarct volume on diffusion-weighted imaging (DWI) increased after complete recanalization in 85.7% (median change 4.4 mL); each decile increase in volume change lowered odds of 90-day functional independence (aOR 0.81, 95%CI 0.68–0.97).
CTA radiomics of both plaque and lumen predicted 1-year stroke recurrence better than plaque-only features (testing AUC 0.74-0.81 vs 0.58-0.62) in 203 patients with symptomatic intracranial atherosclerotic stenosis.
Noncontrast abbreviated MRI after US visualization score C in cirrhosis: DWI adequate in 91.7%, T2WI in 95.1%; 94.4% (17/18) of detected HCCs were early-stage — retrospective study.
Among 58 patients undergoing venous sinus stenting for pulsatile tinnitus, balanced transverse sinus dominance on CTV/MRV was associated with 0% residual tinnitus, vs 23.7% unfavorable outcomes with non-balanced dominance (p=0.021).
A CT-based radiomics model (RaFiST) predicted tumor fibrosis in NSCLC (AUC 0.879 training, 0.813 test) and identified patients with worse survival (all p<0.01). Combined Clinical-RaFiST model achieved 5-year DFS AUC 0.827.
No synthetic data augmentation method (cVAE, cSD-VAE, cWGAN-GP) significantly improved balanced accuracy for CT-based emphysema subtype classification over baseline (0.522 ± 0.067) in 168 patches. Best method (cWGAN-GP at 1.0x) achieved 0.548 ± 0.068 (p=0.084).
Hospital physicians were more likely to implement a 3D interactive artificial intelligence (AI) radiology report (82.4%) than private practitioners (36.8%; p<0.001). Private practitioners more often felt AI aids must be disclosed (82.6% vs 76.9%).
New consensus from ESTRO, ESNR, and EORTC defines Treatment-Related Imaging Abnormality (TRIA) for post-radiotherapy MRI changes in intracranial CNS tumors, with clear definitions, imaging subcategories, and required MRI request elements to harmonize reporting and research.
In 90 oral/maxillofacial cases, three multimodal artificial intelligence chatbots were not clinically usable: panoramic-only accuracy for diseased cases was 0-60%, rising to 38-92% with added info. Cropped images helped most; misleading cues lowered accuracy.
This primer teaches endocrinologists how AI learns from pituitary imaging data, develops clinical applications, and how to critically appraise AI literature in pituitary medicine.
Generative AI model GPT-5 achieved 76.5% accuracy on Japan's oral and maxillofacial radiology board exam, significantly outperforming GPT-3.5 (40.3%) and GPT-4 (67.8%). It exceeded passing scores, especially on two-answer questions.
Abdominal vascular calcification on routine CT is a marker of systemic atherosclerotic burden, with semiquantitative (Agatston, volume) and length-adjusted metrics aiding risk stratification for cardiovascular events, renal dysfunction, and mortality. Standardization and AI vali…
Perspectives on the future of artificial intelligence in paediatric radiology.
Deep learning on full-head T1 MRI distinguished Alzheimer's from normal cognition with 91.07% accuracy, 95.16% AUC; saliency maps highlighted scalp and skull, suggesting extracranial signals are relevant. Hypothesis-generating, requires validation.
A multiparametric MRI-based 3D deep learning fusion model differentiated lung cancer from non-lung cancer brain metastases with an AUC of 0.925, outperforming single-sequence models (0.817, 0.802). Patient-level AUC was 0.942. Radiologist accuracy improved with AI assistance (tr…
Preoperative Pressure-Enabled Thyroid Artery Embolization to Reduce Blood Loss and Facilitate Transoral Thyroidectomy.
Intelligent organ recognition + AIIR in CAP CT improved image quality and cut effective dose 36% vs conventional reconstruction, with tube current down 45% (chest) and 33% (abdomen/pelvis) (p<0.001).
Ultrasound-derived fat fraction (UDFF) vs biopsy for steatosis grading in MASLD: cutoffs ≥6%, ≥15%, ≥21% for mild, moderate, severe; training AUROCs 0.966, 0.805, 0.929; externally validated (0.972/0.922, 0.948/0.934, 0.911/0.936).
5-Tesla magnetic resonance imaging (MRI) reveals atrophy of the cochlear nerve aperture segment in acquired severe-to-profound sensorineural hearing loss: aperture-to-meatus size ratio 0.66 vs 0.75 (p=0.045), and size correlates with hearing loss duration and severity.
A YOLOv10 model detected the nasopalatine canal on CBCT with 100% recall, and MobileNetV3Large classified its morphology with 81.48% accuracy in a retrospective study of 135 patients, showing feasibility for pre-procedural screening.
AI-assisted fracture detection on limb radiographs was associated with an 8.13-minute shorter emergency department length of stay (95% CI -11.52 to -4.70) in fracture-excluded patients, without increased early revisits.
Sodium fluoride (NaF) PET visualizes active arterial microcalcification below the resolution of CT, years before macrocalcification appears. This allows clinicians to monitor early response to statins or PCSK9 inhibitors, even when CT calcium scores stay unchanged.
A projection-based cascade CNN detected 34 dental landmarks on 3D digital models with mean error 0.66 mm; arch length discrepancy (ALD) error was 0.90 mm, comparable to intraobserver error (0.95 mm). Processing time <4 s per model. External validation ALD error: 1.70 mm.
In a preclinical study, an AI-designed dimeric 68Ga-PLEC PET tracer achieved tumor uptake of 3.8 ± 0.5%ID/g at 30 min in pancreatic cancer xenografts, significantly higher than a monomeric tracer, with receptor-specific blockade.
An integrated model combining AI-derived CT body composition, nutritional, and inflammatory markers predicted 90-day nutrition-related adverse events in NSCLC with AUCs of 0.842 and 0.816 in development and temporal validation cohorts, outperforming clinical-only models.
Your Donations in Action: Zofia M. Lasiecka, MD, PhD
Aero Simulation, Inc. (ASI) Employer Profile
Cortechs.ai wins FDA clearance for NeuroQuant PET