TLDR Radiology

This week in Radiology 11/13/24

Diagnostic Radiology

Radiology, published 11/12/24 | Estimated Read Time: 15 minutes

In a comparison of 5T vs 3T cardiac MRI for myocardial fibrosis assessment (n=18), researchers demonstrated that 5T imaging achieved 46% higher signal-to-noise ratios (183.7 vs 125.8, P=.002) and superior contrast-to-noise ratios for normal myocardium (50.8 vs 16.5, P<.001) while maintaining equivalent diagnostic quality and fibrosis quantification (11.8% vs 12.6%, P=.81). Though limited by its small sample size and 6-month interval between comparative scans, this first clinical evaluation of 5T cardiac MRI suggests it could offer enhanced imaging capabilities without compromising diagnostic performance, potentially opening new possibilities for advanced cardiac imaging.

British Journal of Radiology, published 11/12/24 | Estimated Read Time: 25 minutes

Study of 446 patients demonstrated that abbreviated non-contrast MRI protocols achieve exceptional diagnostic accuracy for endometriosis detection, with perfect agreement (kappa=1.0) for ovarian, bladder, uterine, and cesarean scar lesions, and high concordance (kappa≥0.88) across all other pelvic regions when compared to full contrast-enhanced protocols. The streamlined approach reduced imaging time by 30% and costs by 45%, while DWI sequences successfully identified malignant transformation in endometriomas with a 100% detection rate (2/2 confirmed malignancies), though the study's single-reader, retrospective design warrants further validation.

Neuroradiology, published 11/11/24 | Estimated Read Time: 20 minutes

Single-center study of 171 high-grade glioma patients found that FET PET effectively differentiated tumor recurrence from treatment changes with 87.13% accuracy (91.6% sensitivity, 76.9% specificity) using a tumor-to-background ratio cutoff of 2.5. The findings, validated through histopathology in 18 patients and standardized clinical follow-up, demonstrated that 109 of 121 FET PET-positive cases were confirmed recurrences while 40 of 50 FET PET-negative cases were confirmed treatment-related changes, suggesting FET PET could serve as a valuable complementary tool to MRI in post-treatment glioma assessment.

Radiology, published 11/12/24 | Estimated Read Time: 20 minutes

Study involving 184 patients reveals dual-energy CT (DECT) achieves 91% sensitivity and 93% specificity in diagnosing lower limb stress injuries compared to MRI, with significantly better performance in foot versus leg injuries (AUC 0.97 vs 0.88, p<0.001). Even in cases without visible fracture lines, DECT maintained high diagnostic accuracy with 79% sensitivity and 93% specificity, though performance was notably reduced in young patients with thick cortical bone versus elderly patients with osteoporotic changes.

Radiology, published 11/12/24 | Estimated Read Time: 20 minutes

In a large-scale analysis of 121,652 digital breast tomosynthesis screenings performed by 15 radiologists, researchers found that batch reading significantly improved performance, with false-positive rates dropping from 11.5% to 9.4% (P<.001) and interpretation times decreasing from 3.2 to 2.7 minutes (P<.001), while maintaining cancer detection sensitivity. While these findings suggest potential benefits of organizing DBT screenings into moderate-sized batches of 4-10 cases, the study's limitations—including its single academic institution setting and higher-than-national-average recall rates—warrant further investigation in diverse clinical settings.

Interventional Radiology

Journal of Vascular and Interventional Radiology, published 11/10/24 | Estimated Read Time: 20 minutes

Single-center study of 1,075 PAE patients reports exceptional outcomes with IPSS scores dropping from 23 to 6 at 6-12 months (p<0.001) and maintained improvement through 60 months, while 94% of retention patients became catheter-free at 3 months. The study demonstrates PAE's durability with only 16% requiring re-intervention at 5 years and 65.5% achieving freedom from BPH medications at 1 year.

Journal of Vascular and Interventional Radiology, published 11/12/24 | Estimated Read Time: 25 minutes

LAVA trial demonstrates remarkable efficacy for Lava Liquid Embolic System, an ethylene vinyl alcohol (EVOH), in the treatment of peripheral arterial hemorrhage, with 94.3% clinical success across 141 treated lesions and zero major adverse events during 30-day follow-up, significantly exceeding the performance goal of 72%. While the study showed promising versatility across multiple bleeding sites, including visceral (31.1%), renal (26.3%), and GI (18.3%) vessels, key limitations include its single-arm design and that 25.7% of lesions required adjunctive embolotherapy. The 8.3% all-cause mortality rate (9/109 patients) and reliance on experienced EVOH users warrant consideration, but the results provide compelling evidence for this novel approach to managing peripheral arterial bleeding.

Cardiovascular and Interventional Radiology, published 11/6/24 | Estimated Read Time: 25 minutes

Researchers have developed a nanocarrier-mediated stem cell delivery system that achieved an 80.3% reduction in ablation cavity volume compared to 61.5% with standard MSCs (p<0.001) in porcine livers, with significantly higher GFP+ cell counts (498.6 vs 426.8, p=0.01) demonstrating improved targeted delivery to ablation margins. The technique, which uses dendrimer nanocarriers complexed with LFA-1 to bind to inflammatory markers at ablation sites, could potentially expand treatment options for patients with severe liver dysfunction who are currently ineligible for thermal ablation, though the study's key limitation is its use of healthy rather than cirrhotic livers.

Journal of Vascular and Interventional Radiology, published 11/8/24 | Estimated Read Time: 25 minutes

Meta-analysis comparing covered stents (CS) versus bare metal stents (BMS) in 1,861 limbs where CS demonstrated superior freedom from target lesion revascularization (OR=3.00, 95% CI: 1.05-8.51, p=0.04) and better ankle-brachial index improvement (MD=0.03, 95% CI: -0.00-0.06, p=0.04) in treating aortoiliac occlusive disease. Both stent types showed comparable primary patency rates (CS: 80.7% vs BMS: 76.3% at 60 months) and safety profiles.

Cardiovascular and Interventional Radiology,, published 11/7/24 | Estimated Read Time: 15 minutes

Comprehensive analysis of 3,723 blunt splenic injury cases across England's Major Trauma Centers (2017-2022) demonstrates splenic artery embolization's impressive 94.7% success rate, with significantly shorter hospital stays compared to splenectomy (median 10 vs 12 days, p=0.001). The study found that while embolization showed lower mortality rates than splenectomy (8.6% vs 12.3%) and conservative management (11.5%), these differences weren't statistically significant (p=0.129), though the data strongly supports embolization's growing role in trauma care.

AI and Tech

Neuroradiology, published 11/11/24 | Estimated Read Time: 20 minutes

Meta-analysis of 23 studies reveals promising accuracy for AI in detecting head and neck cancer lymph node metastasis, with PET/CT-based models showing significantly higher performance (AUC: 92%, sensitivity: 91.5%, specificity: 92.5%) compared to CT (AUC: 91%) and MRI (AUC: 84%). Both deep learning and hand-crafted radiomics approaches demonstrated comparable excellence (AUCs of 92% and 91% respectively).

Radiology, published 11/12/24 | Estimated Read Time: 25 minutes

In a study of 43,788 chest radiograph reports, researchers demonstrated that GPT-4 can autonomously perform complex statistical analyses and build predictive models that match human expert performance (AUC 0.80 vs 0.80, P=.51; accuracy 79% vs 78%, P=.27). While the AI excelled at visualization, basic statistics, and machine learning tasks, significant differences in sensitivity and specificity compared to human-developed models (P<.001) and several statistical errors highlight the continued need for expert oversight—suggesting these tools should enhance rather than replace radiologists' analytical workflows.

Abdominal Radiology, published 11/10/24 | Estimated Read Time: 25 minutes

Meta-analysis of 12 studies encompassing 8,321 colorectal cancer patients, machine learning models demonstrated superior performance in detecting lymph node metastasis with a pooled AUC of 0.87 (95% CI: 0.82-0.91), significantly outperforming both junior and senior radiologists (AUC 0.65, p<0.001). While prospective studies showed particularly strong results (AUC 0.95 vs 0.85 for retrospective, p=0.03), the significantly lower performance in externally validated studies (AUC 0.81 vs 0.89, p=0.01) suggest more diverse multicenter trials are needed before clinical implementation.

Neuroradiology, published 11/8/14 | Estimated Read Time: 15 minutes

AI assistance can significantly reduce MS lesion assessment time, with radiologists saving an average of 2.83 minutes per read (p < 0.001) compared to the standard 9.05-minute reading time, and achieving the greatest efficiency gains during follow-up examinations with approximately 3 minutes saved per scan. While these results suggest promising workflow improvements, the study's limitation to Siemens scanners (1.5T and 3T) and relatively small dataset indicates the need for broader validation across different scanner manufacturers before widespread implementation.

Radiology: Artificial Intelligence, published 11/6/24 | Estimated Read Time: 25 minutes

SCIseg is an open-source deep learning model that automatically segments both spinal cord and intramedullary lesions in T2-weighted MRI scans with Dice scores of 0.92 and 0.61 respectively. The model, validated on 191 patients across three centers, demonstrated comparable accuracy to manual measurements for critical biomarkers including lesion length (P=.42) and maximal axial damage ratio (P=.16), though it showed differences in lesion volume measurements (P=.003), with performance notably lower in cases with severe metal artifacts from surgical hardware.

Upcoming Conferences and Approaching Abstract Submission Deadlines

Conference

Date

Location

Abstract Submission Deadline

Website

Radiological Society of North America (RSNA)

12/1/24-12/4/24

Chicago, IL

closed

https://www.rsna.org/annual-meeting/

Society of Interventional Oncology (SIO)

1/30/25-2/3/25

Las Vegas, NV

closed

https://www.sio-central.org/Events/Annual-Scientific-Meeting 

American Society of Spine Radiology (ASSR)

2/12/25-2/15/25

San Diego, CA

closed

https://www.theassr.org/

Society of Abdominal Radiology (SAR)

2/16/25-2/21/25

Tucson, AZ

closed

https://abdominalradiology.org/sar-subpages/annual-meeting/

Society of Thoracic Imaging (STR)

3/1/25-3/5/25

Huntington Beach, CA

closed

https://thoracicrad.org/?portfolio=annual-meeting 

Association of Academic radiology (AAR)

3/11/25-3/14/25

Los Angeles, CA

closed

https://www.aur.org/aar-annual-meeting 

Society of Interventional Radiology (SIR)

3/29/25-4/2/25

Nashville, TN

closed

https://www.sirmeeting.org/ 

Society of Skeletal Radiology (SSR)

3/30/25-4/2/25

Santa Fe, New Mexico

closed

https://skeletalrad.org/2025-annual-meeting-call-abstracts

Society for Pediatric Radiology (SPR)

4/13/25-4/15/25

Miami, FL

closed

https://www.spr.org/events/spr2024a 

The Neurodiagnostic Society (ASET)

4/24/25-4/26/25

New Orleans, IL

3/1/25

https://www.aset.org/aset-conference-save-the-date/ 

American Roentgen Ray Society (AARS)

4/27/25-5/1/25

San Diego, CA

closed

https://www2.arrs.org/am25/ 

American College of Radiology (ACR)

5/3/25-5/7/25

Washington, D.C.

1/27/25

https://www.acr.org/Lifelong-Learning-and-CME/Meetings-and-Course-Calendar/ACR-Annual-Meeting/Call-for-Abstracts

International Society for Magnetic Resonance in Medicine (ISMRM)

5/10/25-5/15/25

Honolulu, HI

closed

https://www.ismrm.org/meetings-workshops/future-ismrm-meetings/

Global Embolization Oncology Symposium (GEST)

5/15/25-5/18/25

New York, NY

2/3/25

https://annual.thegestgroup.com/GEST25/Public/Content.aspx?ID=114905&sortMenu=106000

American Society of Neuroradiology (ASNR)

5/17/25-5/21/25

Philadelphia, PA

closed

https://www.asnr.org/asnr-2025-call-for-abstracts/ 

If you have any comments, questions, or feedback, please respond to this email!

Thank you,

Emily Barnard