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.
Body Imager - Partnership Track (Abdominal Radiologist) Job Opening in Charlotte, North Carolina
A rule-based NLP algorithm reviewing spine imaging reports identified vertebral fractures with a PPV of 95.7% and NPV of 100% vs physician adjudication. Applied to 14,172 older adults with a diagnosis code, the algorithm confirmed fracture in 88.5%.
Accelerated brain MRI with deep learning reconstruction showed disproportionate quality loss in regions with abnormalities above 4-fold acceleration—all image-quality metrics worsened faster than normal regions (adjusted P<0.001).
Deep learning model on chest CT estimated rib fracture age with a mean absolute error of 7.94 days vs 10.65 days for manual reads, in external testing of 1,848 fractures.
In a meta-analysis of 5 studies (577 patients), AI models for trigeminal neuralgia detection on MRI achieved pooled sensitivity 71% (95% CI 65-76%), specificity 92% (88-94%), and AUC 0.91 (0.88-0.93). High specificity may aid preop planning.
Radiology should preserve an auditable pre-AI interpretation: capture what the radiologist concluded before AI exposure, then the AI display and response. Automation-bias research shows that the current blended record is consequential.
In head and neck squamous cell carcinoma, GPT-assisted MRI radiomics predicted pathological complete response to chemoimmunotherapy with AUC 0.741 (validation) and 0.706 (multicenter prospective), comparable to manual models (0.714 and 0.700).
MRI surveillance of perineural spread from cutaneous SCC shows frequent durable response to anti-PD-1 immunotherapy: in a retrospective cohort of 9 patients, 3 had complete response, 5 partial, none progressed (median 37 mo). Literature review of 101 cases: 35% complete response.
In four postpartum patients with cerebral venous thrombosis (CVT) after epidural anesthesia, all had coexisting spinal CSF leaks, and imaging signs of intracranial hypotension were present on initial brain MRI but overlooked in every case.
In a series of six patients, prominent dorsal perimedullary veins in spontaneous intracranial hypotension (SIH) mimicked spinal dural arteriovenous fistula (dAVF). Key distinguishing features of SIH included orthostatic headache, brain sagging, and absence of cord edema, avoidin…
In a retrospective review of 259 indeterminate T1-hypointense vertebral lesions on spine MRI, 77% of reports recommended further imaging, but only 49% specified timing; recommended intervals ranged from 2-4 weeks to 6-12 months, and modality recommendations were inconsistent.
In 1716 sepsis patients, automated CT-derived visceral-to-subcutaneous fat ratio independently predicted mortality (HR 2.117, 95% CI 1.946-2.301) and improved model C-index by up to 0.080 over clinical variables.
A spatially aware neural network (Spatial NCDE) reduced IVIM parameter estimation error by 81% (D) vs. least squares and 55% vs. UNet in simulations. In esophageal cancer patients, improvement in predicting chemoradiotherapy response was not statistically significant.
DLRC model integrating deep learning, radiomics, and clinicopathological features from preoperative breast MRI predicted 3-year recurrence-free survival with an AUC of 0.79 (95% CI 0.67-0.91) in external validation (n=196).
Direct Percutaneous Access of Renal Artery for Treatment of Hypertensive Emergency in a Patient with Solitary Kidney.
Final-year medical students at Qatar University showed strong ultrasound image acquisition skills but a clear gap in applying findings to clinical reasoning. A qualitative study of 21 students and 15 faculty found preclinical training built technical proficiency, but unstructure…
Multimodal AI integrating imaging, pathology, and genomics outperforms unimodal biomarkers for treatment prediction in lung cancer, according to a review of technical architectures and clinical studies.
Proof-of-concept: MobileNet-V3-large achieved AUC 0.97, sensitivity 0.959 for slice-level acute cholecystitis on contrast-enhanced CT; ResNet-101 had accuracy 0.915, specificity 0.857. Slice-level, single-center, not patient-level – external validation needed.
Researchers can use this guide to select AI reporting resources for dentomaxillofacial radiology, covering radiomics, segmentation, and diagnostic accuracy. It includes decision aids, tables, and examples to boost transparency and reproducibility.
Accelerated deep learning reconstruction knee MRI cut acquisition time by 49.81% (6 min 42 sec less per patient) with equivalent diagnosis, image quality, and artifacts versus conventional sequences in knee pain patients.
Diagnostic Imaging's Weekly Scan: September 6 — September 12
EDAN Instruments Reveals Upgrades to Acclarix LX9 Ultrasound System
Weekly Vigils Protest ICE at Metropolitan Detention Center
Neuroradiologist | Partnership Track + Flexible Employed Options | Raleigh, NC Job Opening in Raleigh, North Carolina
JPEG compression level ≥50 maintains AI output agreement for clinically important labels on panoramic radiographs; lower levels cause degradation, with periapical radiolucency positive agreement dropping to 0.614 at level 10, mainly due to missed findings.
In a conjoint study of 440 participants, lung cancer patients favoured combined human+AI chest X-ray result processing (part-worth +13.3, p<0.001), while other groups preferred human-only. Waiting time and next care-step decisions accounted for ~67% of preference weight.
Multi-habitat radiomics on fat-suppressed T2 MRI achieved AUC 0.883 for identifying active sacroiliitis in axial spondyloarthritis, significantly outperforming whole-lesion radiomics (AUC 0.714; P<0.001). Peripheral transition zone and hyperintense core were key drivers.
AI model detects neonatal pneumothorax on supine chest radiographs with AUC 0.975, sensitivity 87.6%, specificity 95.3%. PPV 68.5%, NPV 98.5%. Grad-CAM matched clinical side in 91.3% of right-lung vs 67.6% left-lung cases. External validation needed.
Cryoablation for extra-abdominal desmoid tumors showed 81.1% clinical success, 67.4% 2-year progression-free survival (95% CI 46.5-81.7%), and 5.3% major adverse events in a 38-patient retrospective study.