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…
Comment on "Comparison of MRI-guided in-bore and fusion biopsy for small PI-RADS 4 prostate lesions".
Opportunistic Coronary Artery Calcium Screening: Time for Clinical Implementation.
Implementing a Department-wide Peer Learning Program Across a Large Academic Radiology Practice.
In a review of five case series (139 patients, 202 procedures) of transarterial embolization for refractory Achilles tendinopathy, pain scores dropped 60-83% from baseline, with 82-100% achieving ≥50% pain reduction and no major adverse events.
30-day mortality was 35% in 20 patients who had mechanical thrombectomy on VA-ECMO for high-risk pulmonary embolism (PE); complications occurred in 45%, a single-center case series reports.
Arterial spin labeling showed decreased cerebral blood flow in patients with unresponsive wakefulness syndrome, consistent with FDG-PET metabolism reductions. Positive correlations occurred globally and in most functional networks, supporting ASL as a noninvasive alternative.
A deep learning model trained on paired contrast-enhanced and non-contrast MRI segmented kidneys on non-contrast T1 with Dice 0.950. Total kidney volume had mean absolute error 8.76 mL (2.5% of mean) and concordance correlation 0.983, enabling contrast-free renal volumetry.
Oncoformer, a multimodal transformer trained on 3.67M individuals, achieved a pan-cancer diagnosis AUROC of 0.956 and predicted future cancer up to 1 year before diagnosis with AUROC 0.869, using longitudinal EHR and chest X-ray data.
A Bayesian deep learning framework (µGUIDE) applied to gray matter diffusion MRI models shows that exchange time and soma radius estimates are often unreliable, while extracellular diffusivity and neurite signal fraction are robust. Uncertainty-aware methods help flag poor estim…
A deep cross-branch multimodal fusion network (DCMFNet) using sMRI and rs-fMRI outperformed traditional ML and state-of-the-art DL models in six binary Alzheimer's disease subtype classification tasks, potentially aiding early diagnosis.
The FDA has cleared the Philips EPIQ Series and Affiniti Series diagnostic ultrasound systems for general imaging use via the 510(k) pathway.
The FDA has cleared the ND Needle Guidance System via the 510(k) pathway for use in interventional radiology procedures.
A deep learning framework integrating 3D T1 brain MRI with demographics and genetic data shows promise for predicting six dementia-related neuropathologies, potentially enabling non-invasive profiling, according to an autopsy-confirmed study.
A CT-derived intratumoral heterogeneity model achieved AUCs of 0.824 (training) and 0.792 (validation) for predicting major pathological response to neoadjuvant immunochemotherapy in resectable NSCLC. The model combined 2D and 3D heterogeneity scores and systemic inflammation in…
Radiomics on panoramic radiographs moderately differentiated odontogenic cysts; adding perilesional/contextual features raised four-class balanced accuracy from 0.416 to 0.487 (p=0.002). Best pairwise accuracy was 0.838 for dentigerous versus unspecified odontogenic cyst.
Adding deep learning-augmented zero echo time (ZTE DL) MRI to standard hand/foot protocols at 1.5T improved diagnostic confidence for bone findings (p<0.05 vs conventional MR and, for most lesions, vs radiographs) in 40 patients, though diagnostic accuracy wasn't assessed.
In 183 patients from public ACRIN 6698/BMMR2, early longitudinal breast MRI prediction of pathologic complete response was modest and comparable: handcrafted radiomics best AUROC 0.670; frozen deep embeddings best AUROC 0.672 and best AUPRC 0.472. Models remain exploratory.
FDA rejected Ascelia Pharma’s Orviglance, an oral manganese-based MRI contrast for focal liver lesions, because only 1 of 3 initial readers showed superiority of contrasted over noncontrast images. The agency recommends a new controlled study with rigorous reader training.
Wellumio raised $9.75M to advance its portable, AI-powered MRI brain imaging system for point-of-care stroke diagnosis. The 110-lb device requires no cryogen or contrast and has shown agreement with conventional MRI in early studies.
Skagit Radiology Employer Profile
A machine learning model using routine CT dose report metrics estimated adult body weight with a mean absolute error of 2.83 kg; 88.8% of exams were within 10% of measured weight. Findings from single-center, 2496-patient study.
Radiology Imaging Associates: Remote Overnight General Teleradiologist: W2 or 1099 Job Opening in Nationwide
<![CDATA[Diagnostic Imaging's Weekly Scan: July 19 — July 25]]>
<![CDATA[Study: AI Measurement of Coronary Inflammation from CT Scans Can Identify High CV Risk Patients with Low or Zero CAC Scores]]>
A multimodal deep learning model combining MRI and clinical features predicted positive surgical margins after radical prostatectomy with an AUC of 0.843 (95% CI 0.770-0.911) in an external validation cohort of 155 patients from six centers, outperforming imaging-only and clinic…
Radiomic scores from cardiac MRI T1 and ECV maps had an AUC of 0.834 and 0.718, respectively, for predicting left ventricular reverse remodeling in dilated cardiomyopathy, improving AUC of late gadolinium enhancement from 0.618 to 0.903 when combined.
In early CKD detection, multiparametric MRI combining all four sequences (T1WI, DWI, T1 mapping, BOLD) achieved AUC 0.866. DWI alone had AUC 0.874; T1WI+DWI balanced performance and feasibility (accuracy 0.844, F1 0.815).
An AI-supported workflow increased patient notification of incidental pulmonary nodules from 75% to 88.4% (p=0.0001) and follow-up from 53.5% to 67.9% (p=0.0012) in the ED.
A denoising autoencoder restored CT perfusion image quality at up to 90% simulated radiation dose reduction (SSIM 0.81, PSNR 43 dB) and reduced infarct core volume error from 10-30 mL to 4-16 mL, with high agreement (R²>0.85).
AI can improve radiation protection in medical imaging and radiotherapy through referral support, image reconstruction, and QA, but data bias and opaque models necessitate robust validation and AI literacy training, per an ICRP Committee 3 perspective.
