LI-RADS US Surveillance Version 2024 for Surveillance of Hepatocellular Carcinoma: An Update to the American College of Radiology US LI-RADS Open Access 

Radiology, published 12/3/24 | Estimated Read Time: 25 minutes

The American College of Radiology's new LI-RADS US Surveillance Version 2024 overhauls HCC screening protocols, notably reducing follow-up for US-2 findings from 2 years to just two 3-6 month intervals and recommending alternative imaging for patients with repeated VIS-C scores. The updated guidelines demonstrate impressive validation metrics, with US surveillance showing a 15-35% positive predictive value for HCC detection (compared to mammography's 4.4% for breast cancer) and strong interreader agreement (intraclass correlation 0.7), while maintaining cost-effectiveness at a risk threshold of 0.8% per year.

Navigating the future of mammography: How women’s perceptions of AI may guide tomorrow’s screening practice

European Journal of Radiology, published 12/3/24 | Estimated Read Time: 15 minutes

Recent Norwegian study reveals promising acceptance of AI in mammography screening, with 64% (5,009/7,829) of participants willing to accept AI as a decision support tool and 58.5% (4,577/7,822) expressing confidence in AI when used alongside radiologists. While 76.6% (5,678/7,795) of women wanted to be informed about AI involvement in their screening, the study's 10% response rate and potential "healthy screenee" bias suggest the need for broader population sampling to confirm these findings.

The impact of 3D real-IR delayed post gadolinium MRI parameterisation on the diagnostic performance and optimal descriptor selection in Ménière’s disease Open Access 

European Radiology, published 12/3/24 | Estimated Read Time: 15 minutes

In a comparison of MRI techniques for diagnosing Ménière's disease (n=132 patients), traditional zero-point endolymph (ZPE) and optimized negative signal endolymph (NSE) sequences showed equivalent diagnostic accuracy (95.8% correct classification for both), despite NSE demonstrating superior reliability in 9/14 imaging descriptors. While NSE achieved higher signal-to-noise ratios (186.3 vs 118.3), ZPE surprisingly performed better for key diagnostic markers including saccular abnormalities (risk ratio 6.571 vs 6.300) and asymmetric enhancement (risk ratio 3.628 vs 2.903), suggesting both sequences remain viable options for clinical practice - though radiologists should note their sequence choice will determine which imaging features are most reliable for diagnosis.

Diffusion Tensor Imaging and Evaluation of Cardiac Ischemic Disorders: A Systematic Review 

Academic Radiology, published 11/29/24 | Estimated Read Time: 25 minutes

Systematic review of 16 studies reveals DTI's potential as a non-invasive alternative to cardiac biopsy, with infarcted regions showing significantly lower FA values in AMI (0.25 ± 0.03) compared to remote myocardium (0.36 ± 0.02, p < 0.001) and higher MD values (1.69 ± 0.14 vs 1.48 ± 0.12) in affected areas. The findings were consistent across both human and animal studies, though the research faced notable limitations including small sample sizes and incomplete histopathological validation, warranting larger confirmatory studies before clinical implementation.

Histopathologic confirmation of subchondral fracture in symptomatic pre-collapse osteonecrosis of the femoral head with bone marrow edema on magnetic resonance imaging

Skeletal Radiology, published 11/29/24 | Estimated Read Time: 25 minutes

In a histopathological study of 13 femoral heads, researchers found that bone marrow edema (BME) on MRI serves as a definitive indicator of subchondral fractures in pre-collapse osteonecrosis of the femoral head (ONFH), with fractures consistently occurring at the lateral junction between reparative and necrotic zones, demonstrating significantly higher bone volume fraction (83.1% vs 40.1%, P<0.001) and trabecular thickness (502.8μm vs 187.6μm, P<0.001) in the reparative zone. Despite study limitations, including a small sample size and time lag between imaging and surgery (average 10.7 weeks), these findings provide crucial evidence that BME on MRI should prompt clinicians to consider subchondral fracture in treatment planning, even when radiographs show no apparent collapse.

Microvascular Plug Embolization of Anterior Spinal Artery Bearing Segmental Arteries Prior Aortic Stenting: Technique and Safety Open Access 

Cardiovascular and Interventional Radiology, published 11/27/24 | Estimated Read Time: 15 minutes

In a retrospective study of microvascular plug (MVP) embolization for spinal cord protection, researchers achieved 100% technical and clinical success across 22 patients (117 segmental arteries), with zero cases of spinal cord ischemia during a median 10.9-month follow-up period. The technique, which involves a 10-minute test deployment before permanent placement, offers a potentially safer alternative to traditional coil embolization, which has shown paraparesis rates of up to 14% in previous studies. Although limited by its retrospective design and small sample size, the study demonstrates MVP's potential as a precise and retrievable option for protecting the anterior spinal artery during MISACE procedures.

Vessel deviation during stent retrieval predicts successful recanalization in stent-based mechanical thrombectomy for M2 occlusion

Neuroradiology, published 11/27/24 | Estimated Read Time: 25 minutes

Research reveals that lower vessel deviation during stent retrieval strongly predicts successful recanalization in M2 occlusion thrombectomy, with successful cases showing a median D/B ratio of 0.9 versus 2.5 in unsuccessful cases (p<0.001). The study of 25 patients demonstrated that advancing the aspiration catheter to M2 significantly reduced vessel deviation compared to M1/ICA positioning (median D/B ratio 0 vs 1.5, p=0.003), though the study was limited by its small sample size, retrospective design, and manual measurements of the D/B ratio.

Outcomes of Mechanical Thrombectomy for Acute Ischemic Stroke Following Cardiac Interventions: A Systematic Review and Meta-Analysis

Cardiovascular and Interventional Radiology, published 11/27/24 | Estimated Read Time: 25 minutes

New research demonstrates that mechanical thrombectomy may be an effective rescue treatment for acute ischemic stroke following cardiac interventions, achieving successful reperfusion in 79.4% of cases (95% CI 66.7-88.1%) and favorable functional outcomes in 42.7% of patients (95% CI 32.5-53.6%) at 90 days. In this largest meta-analysis to date examining 195 thrombectomy procedures, half of the cases occurred after TAVR/TAVI procedures, though mortality (30.3%) and symptomatic intracranial hemorrhage (11.6%) rates were higher than typically seen in non-cardiac surgery stroke cases - suggesting careful patient selection and monitoring is warranted.

“Efficacy of needle tract embolization using gelfoam in reducing incidence of Pneumothorax in CT-guided transthoracic Lung Biopsies: A Systematic Review and Meta-analysis.”

British Journal of Radiology, published 11/28/24 | Estimated Read Time: 25 minutes

CT-guided lung biopsy complications can be significantly reduced using gelfoam slurry for needle tract embolization, with a new meta-analysis of 1,885 patients showing a 59% reduction in pneumothorax (OR = 0.41, 95% CI= 0.25–0.66, p = 0.01) and a 63% reduction in chest tube placement (OR = 0.37, 95% CI= 0.20–0.69, p = 0.01). In the embolization group, only 3.6% of patients required chest tubes versus 10.3% in the control group.

Ethanol Chemical Gallbladder Ablation for Cholecystitis in Inoperable Elderly Patients

Cardiovascular and Interventional Radiology, published 11/27/24 | Estimated Read Time: 25 minutes

10-year study demonstrates the potential of ethanol chemical gallbladder ablation as a minimally invasive alternative for high-risk elderly patients with cholecystitis, reporting zero complications and a 95% success rate (19/20 patients) with a median survival of 603 days. When combined with cystic duct embolization, the technique achieved significantly higher gallbladder atrophy rates (83% vs 29% without embolization), though the researchers acknowledge limitations including the small sample size, retrospective single-center design, and lack of a control group. 

Large Language Models with Vision on Diagnostic Radiology Board Exam Style Questions Open Access 

Academic Radiology, published 12/3/24 | Estimated Read Time: 15 minutes

Research testing vision-enabled LLMs on 280 radiology board-style exam questions reveals that image analysis capabilities don't enhance performance, with GPT-4V and Gemini 1.5 Pro showing no significant difference across text-only, image-only, and multimodal formats (52-57% accuracy). Claude 3.5 Sonnet emerged as the top performer in text-only (63%) and multimodal formats (66%), significantly outperforming competitors (p < .01), but notably struggled with image-only inputs (48%, p < .001 compared to text/multimodal), highlighting the current limitations of AI vision capabilities in radiology diagnostics.

Applying Conformal Prediction to a Deep Learning Model for Intracranial Hemorrhage Detection to Improve Trustworthiness

Radiology: Artificial Intelligence, published 11/27/24 | Estimated Read Time: 20 minutes

Deep learning model demonstrates both high performance in detecting intracranial hemorrhages (F1 score = 0.92) and the ability to identify cases where it's uncertain, achieving 95.3% accuracy in flagging challenging cases that even experienced radiologists disagreed on. The model, which uses Mondrian conformal prediction to evaluate its own uncertainty, showed robust performance on external validation with the RSNA dataset, though it currently only works with 2D CT sections rather than complete 3D volumes.

AI-driven Patient-Selection For Preoperative Portal Vein Embolization For Patients With Colorectal Cancer Liver Metastases

Journal of Vascular and Interventional Radiology, published 12/2/24 | Estimated Read Time: 25 minutes

A multi-institutional study has demonstrated a novel AI approach that predicts portal vein embolization (PVE) outcomes with remarkable accuracy, achieving AUCs of 0.91/0.88 (internal/external) for future liver remnant prediction and internal/external accuracies of 85.81%/79.66% for patient selection. The algorithm successfully predicted total liver volume within a 12.56% margin (95% CI: ±4.20%, p=0.86) by leveraging radiomic features and shape analysis, though it was limited by binary outputs and lack of KRAS mutation data.

AI-Based Evaluation of Prostate MR Imaging at a Modern Low-field 0.55 T Scanner Compared to 3 T in a Screening Cohort Open Access 

Academic Radiology, published 12/2/24 | Estimated Read Time: 15 minutes

Research demonstrates that low-field 0.55T MRI scanners show comparable performance to 3T systems for prostate imaging, with volumetric measurements showing no significant differences between field strengths (PZ: 14.81 vs 14.29 mL, p=0.13; CG: 23.07 vs 22.77 mL, p=0.38) and ADC value variations remaining within an acceptable ±20% margin of difference. While this study of 25 participants validates the potential of more comfortable and cost-effective low-field imaging, particularly for claustrophobic patients, its key limitation is the small sample size and lack of comprehensive lesion detection analysis, warranting larger studies to confirm diagnostic equivalence using PI-RADS criteria.

Predicting Survival and Recurrence of Lung Ablation Patients Using Deep Learning-Based Automatic Segmentation and Radiomics Analysis

Cardiovascular and Interventional Radiology, published 11/27/24 | Estimated Read Time: 20 minutes

This study demonstrates that post-procedure CT radiomics can predict lung ablation outcomes with a c-index of 0.71 and AUC of 0.75 for survival prediction, significantly outperforming pre-procedure imaging's c-index of 0.56 in a cohort of 113 patients. The study's automated deep learning pipeline, which achieved a Dice score of 0.75 for tumor segmentation, showed that tumor size metrics ("Maximum 2D Diameter Slice" and "Maximum 3D Diameter") were the most predictive features for survival, while tumor shape characteristics (elongation, axis length, and surface area-to-volume ratio) best predicted recurrence with a c-index of 0.65 and AUC of 0.72.

Upcoming Conferences and Approaching Abstract Submission Deadlines

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

Thank you,

Emily Barnard