Follow-up analysis of lesion characteristics of enchondromas and atypical cartilaginous tumours of the knee and shoulder region on MRI Open Access 

European Radiology, published 10/16/24 | Estimated Read Time: 15 minutes

A study of 182 cartilaginous lesions provides new insights into differentiating atypical cartilaginous tumors (ACTs) from enchondromas (ECs) through MRI follow-up patterns. The research revealed statistically significant differences in growth rates - ACTs showed consistent growth (0.039-0.083 mm/month) while ECs remained stable or decreased in size (0 mm/month, p<0.01), potentially offering a new diagnostic tool for radiologists. The findings are limited by its retrospective nature and the lack of histopathological confirmation in most cases.

Internal hematoma architecture predicts subdural hematoma responsiveness to standalone middle meningeal artery embolization

Neuroradiology, published 10/18/24 | Estimated Read Time: 25 minutes

Middle meningeal artery embolization (MMAE) may eliminate the need for surgical evacuation in select subdural hematoma patients. Consisting of 164 cases, there was a remarkable 98.8% success rate in those with hematomas with a laminar/trabecular architecture compared to 87.6% in homogeneous/separated types (p=0.005). Though limited by its single-center retrospective design, this analysis suggests MMAE may offer a safer alternative for high-risk surgical patients, based on radiographic classification of the hematoma architecture emerging as a critical predictor of treatment success.

Multiparametric MRI-Based Deep Learning Models for Preoperative Prediction of Tumor Deposits in Rectal Cancer and Prognostic Outcome 

Academic Radiology, published 10/21/24 | Estimated Read Time: 25 minutes

Deep learning model using multiparametric MRI that could transform preoperative tumor deposit prediction in rectal cancer patients. The model, which combines clinical factors with deep radiomic features, achieved remarkable accuracy across multiple validation cohorts (AUC >0.9) and demonstrated significant prognostic value by effectively stratifying 3-year disease-free survival rates – with tumor deposit positive patients showing dramatically lower survival (66.9% vs 90.7%). While promising, the study's retrospective nature and exclusion of neoadjuvant therapy patients indicate the need for prospective validation.

Mechanical properties of pediatric low-grade gliomas in children with and without neurofibromatosis type 1 Open Access 

Neuroradiology, published 10/21/24 | Estimated Read Time: 15 minutes

Unique mechanical signatures of pediatric brain tumors using magnetic resonance elastography (MRE), showing that low-grade gliomas are significantly softer (10.9% decrease, p=0.010) and less viscous (17.3% lower damping ratio, p=0.009) than healthy tissue. Neurofibromatosis type 1 lesions were found to have distinct mechanical patterns, potentially offering clinicians a new tool for differentiating between tumor types in children. While promising, the study was limited by its sample size and tumor heterogeneity, suggesting the need for larger-scale validation before clinical implementation.

Clinical Effectiveness of Automated Coronary CT-derived Fractional Flow Reserve: A Chinese Randomized Controlled Trial Open Access 

Radiology, published 10/15/24 | Estimated Read Time: 25 minutes

In a large randomized control trial (n=5,297), adding CT-derived fractional flow reserve (CT-FFR) to routine coronary CT angiography led to a significant 19.4% reduction in 90-day invasive coronary angiography rates (10.0% vs 12.4%, P=.006) while maintaining similar 1-year major adverse cardiovascular event rates (2.9% vs 2.8%, P=.90). The study demonstrated fewer cardiac events at 1 year in the CT-FFR group, though key limitations include the single-vendor CT-FFR software and inability to establish causality between CT-FFR use and outcomes. This real-world evidence suggests CT-FFR may serve as an effective gatekeeper to invasive procedures while maintaining patient safety.

Systemic Immunological Changes After Yttrium-90 Radioembolization: A Pilot Prospective Observational Study—Clinical Insights 

Cardiovascular and Interventional Radiology, published 10/15/24 | Estimated Read Time: 25 minutes

Significant immunological changes following Y90 radioembolization in colorectal liver metastases patients were observed, with notable increases in CD3+ T cells (p=0.002) and Ki-67+ proliferating CD8+ T cells (p=0.02) at 3 weeks post-treatment, suggesting a shift toward favorable antitumoral immune response. The study, though limited by its small sample size (n=15), demonstrated intriguing cytokine dynamics, including significant IFN-γ decrease immediately post-TARE (p=0.006) followed by recovery at 6 weeks, and decreased immunosuppressive IL-10 at 3 weeks (p=0.025). These findings could pave the way for future combination therapies of TARE with immunotherapy in colorectal cancer patients, particularly given that 5/15 patients experienced sustained disease control for 17-29 months.

Trans-arterial embolization for treatment of acute lower gastrointestinal bleeding—a multicenter analysis Open Access 

European Radiology, published 10/16/24 | Estimated Read Time: 20 minutes

Multicenter analysis of 141 patients reveals trans-arterial embolization achieved 100% technical and 93.6% clinical success rates in treating acute lower GI bleeding, though with a 14% rate of ischemia-related complications requiring surgery. While microcoils were the most commonly used embolic (48.5%), the study found interesting trends suggesting increased bowel resection risk with glue (p=0.090) and higher 30-day mortality with less selective embolization (p=0.057), providing practical insights for embolic choice and technique.

Clinical outcomes following invasive treatment of femoropopliteal artery disease: a retrospective single-center cohort study Open Access 

Journal of Vascular and Interventional Radiology, published 10/18/24 | Estimated Read Time: 15 minutes

A compelling retrospective analysis of 676 patients challenges our understanding of femoropopliteal disease management. The study reveals that critical limb-threatening ischemia is the sole independent predictor of primary clinical patency loss (HR 1.25, P=0.008), while demonstrating superior secondary clinical patency with bypass surgery over endovascular approaches (HR 0.66, P=0.037). Though limited by its single-center design and pre-dating modern drug-coated devices, this study offers valuable insights into predicting outcomes, particularly highlighting chronic renal failure as a significant risk factor in CLTI patients.

Outcomes of First-line Microwave Ablation of Treatment-Naive Epidermal Growth Factor Receptor-Mutated Advanced Lung Adenocarcinoma Treated with Tyrosine Kinase Inhibitors

Journal of Vascular and Interventional Radiology, published 10/19/24 | Estimated Read Time: 20 minutes

First-line microwave ablation (MWA) combined with tyrosine kinase inhibitors (TKIs) significantly improved outcomes in treatment-naive EGFR-mutant advanced lung adenocarcinoma patients, showing substantially longer median progression-free survival (19.0 vs 10.0 months, P<0.001) and overall survival (41.0 vs 25.0 months, P=0.044) compared to TKIs alone. After propensity score matching, patients who achieved complete ablation showed even more impressive results with median PFS of 24.0 vs 10.0 months (P<0.001) and OS of 48.0 vs 25.0 months (P=0.012) compared to TKIs alone, though the study's retrospective nature and potential selection bias warrant further evaluation through prospective trials. The research suggests that early intervention with MWA before TKI resistance develops could be a promising strategy, particularly when complete ablation is achievable.

Percutaneous Cryoablation in the Liver: A Meta-Analysis and Review of Safety with a Focus on Incidence of Cryoshock and Major Complications Open Access 

Cardiovascular and Interventional Radiology, published 10/15/24 | Estimated Read Time: 25 minutes

A landmark meta-analysis of 4,029 patients reveals cryoshock - the most feared complication of hepatic cryoablation - occurred in only 0.265% of cases and exclusively in lesions >3cm. Overall major complications were low at 4.71%, with a significant trend toward improved safety in recent years (p=0.015). While the analysis was limited by high heterogeneity in complication reporting and inconsistent cryoshock definitions across studies, this comprehensive review provides compelling evidence for reconsidering cryoablation's role in liver tumor treatment.

Automated AI-based coronary calcium scoring using retrospective CT data from SCAPIS is accurate and correlates with expert scoring Open Access 

European Radiology, published 10/18/24 | Estimated Read Time: 10 minutes

AI-based coronary calcium scoring has shown exceptional promise in automating this time-consuming task. Using a  large dataset (4,567 patients), researchers found the AI achieved near-perfect correlation with expert readers (ICC=0.994, p<0.001) and 91.2% accuracy in cardiovascular risk classification. While the AI showed a slight tendency to overestimate scores in the lower ranges, an audit revealed it was actually correct in 20.1% of apparent misclassifications. The study's main limitation was its single-center, single-vendor design using only Siemens equipment, warranting further validation across different systems and institutions before widespread clinical implementation.

Streamlining Radiology Reporting: A hands-free approach with voice-to-text and generative AI

Journal of the American College of Radiology, published 10/16/24 | Estimated Read Time: 15 minutes

Combining voice-to-text technology with GPT-4 to automate radiology reporting, demonstrates a significant 38% reduction in report generation time (p < 0.0001) while maintaining excellent quality metrics (clarity: 4.64/5, accuracy: 4.59/5). While the system excelled with routine chest and MSK radiographs, it had limitations with complex cases and was tested only on a modest sample of 100 studies at a single institution - but the results suggest an intriguing path forward for reducing radiologist workload in routine cases.

Revolution or risk?—Assessing the potential and challenges of GPT-4V in radiologic image interpretation Open Access 

European Radiology, published 10/18/24 | Estimated Read Time: 25 minutes

Evaluation of GPT-4V's performance in radiology reveals concerning limitations, with diagnostic accuracy at just 8.3% without clinical context and 29.1% with context for immediate correct diagnoses (p≤0.001). The study, analyzing 206 cases across multiple imaging modalities, found that GPT-4V frequently fabricated findings (258 instances) and misidentified anatomical regions or modalities (65 instances), highlighting significant patient safety concerns. While diagnostic reasoning was largely plausible (92.7% of responses), the tool's self-reported confidence showed no correlation with accuracy (ρ=0.117, p=0.776), though study limitations include single-image analysis and single-center design.

AI in radiology: From promise to practice − A guide to effective integration

European Journal of Radiology, published 10/20/24 | Estimated Read Time: 30 minutes

Study reveals concerning accuracy issues with AI diagnostic systems: when AI incorrectly indicates no abnormality, radiologist false negative rates jump from 2.7% to 20.7-33.0%, and false positive rates surge from 51.4% to 80.5-86% when AI incorrectly flags abnormalities. Through case studies of misidentified fractures and hemorrhages, the study demonstrates how AI is unable to consider clinical history and prior imaging leads to diagnostic errors, particularly with post-operative changes and medical devices, where false positives can account for up to 24% of AI errors. The authors propose multimodal fusion of imaging and clinical data as a solution, offering a practical roadmap for more effective clinical integration.

Testing the Ability and Limitations of ChatGPT to Generate Differential Diagnoses from Transcribed Radiologic Findings Open Access 

Radiology, published 10/15/24 | Estimated Read Time: 25 minutes

ChatGPT's GPT-4 achieved a promising 66.1% accuracy in primary diagnosis interpretation—a significant improvement over GPT-3.5's 53.7% (p<.001). While GPT-4 showed marked reduction in hallucinations and false statements (dropping from 16.2% to 4.7%, p<.001), both models demonstrated inconsistent repeatability, particularly with differential diagnoses, suggesting these tools may currently be better suited for educational purposes rather than clinical work. Key limitations include the use of textbook cases rather than real-world scenarios, and the possibility that the source material may have been included in the models' training data.

Upcoming Conferences and Approaching Abstract Submission Deadlines

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

Thank you,

Emily Barnard