Radiology最新文献

Background Percutaneous CT-guided lung core-needle biopsy is a frequently performed and generally safe procedure. However, with advances in the management of lung cancer, there is a need for a greater amount of tissue for tumor genomic profiling and characterization. Purpose To determine whether the number of core samples obtained with percutaneous CT-guided lung biopsy is associated with postprocedural complications. Materials and Methods This retrospective study included consecutive patients who underwent percutaneous CT-guided coaxial lung core-needle biopsy for suspected primary lung cancer between November 2012 and August 2023 at an academic tertiary referral hospital. Patient data from medical records were collected, including demographics, lesion size and distance from pleura, and number of obtained biopsy samples. Postprocedural complications of pneumothorax, chest tube placement, perilesional hemorrhage, and hemoptysis were recorded. Multivariable logistic regression models were used to assess whether the number of cores was a predictive factor for lung biopsy complications. Results A total of 827 patients (mean age, 70.9 years ± 9.6 [SD]; 474 [57.3%] female patients) were included. The median lesion size was 22 mm (IQR, 15-34 mm), with 517 of 827 (62.5%) patients diagnosed with lung adenocarcinoma. Pneumothorax was noted in 171 of 827 (20.7%) patients, with a chest tube placed in 32 of 827 (3.9%), perilesional hemorrhage in 353 of 827 (42.7%), and hemoptysis in 20 of 827 (2.4%) patients. The median number of samples obtained was four (range, one to 12). Multivariable analysis showed no evidence of an association between the number of core samples obtained and any complications: pneumothorax (coefficient, -0.02; P = .81), chest tube (coefficient, 0.18; P = .26), perilesional hemorrhage (coefficient, -0.03; P = .63), or hemoptysis (coefficient, -0.10; P = .60). Conclusion In patients suspected of having lung cancer who underwent percutaneous CT-guided coaxial lung core biopsy, there was no evidence of an association between the number of core biopsy samples obtained and any postprocedural complications. © RSNA, 2024 See also the editorial by Zuckerman in this issue.

With the rising incidence of hepatocellular carcinoma, there has been increasing use of local-regional therapy (LRT) to downstage or bridge to transplant, for definitive treatment, and for palliation. The CT/MRI Liver Imaging Reporting and Data System (LI-RADS) Treatment Response Assessment (TRA) algorithm provides guidance for step-by-step tumor assessment after LRT and standardized reporting. Current evidence suggests that the algorithm performs well in the assessment of tumor response to arterial embolic and loco-ablative therapies and fair when assessing response to radiation-based therapies, with limited data to validate the latter. Both evidence-based and expert-based refinements of the algorithm are needed to improve its diagnostic accuracy after varying types of LRT. This review provides an overview of the challenges and limitations of the LI-RADS TRA algorithm version 2017 and discusses the refinements introduced in the updated 2024 LI-RADS algorithm for CT/MRI.

Background Limited statistical knowledge can slow critical engagement with and adoption of artificial intelligence (AI) tools for radiologists. Large language models (LLMs) such as OpenAI's GPT-4, and notably its Advanced Data Analysis (ADA) extension, may improve the adoption of AI in radiology. Purpose To validate GPT-4 ADA outputs when autonomously conducting analyses of varying complexity on a multisource clinical dataset. Materials and Methods In this retrospective study, unique itemized radiologic reports of bedside chest radiographs, associated demographic data, and laboratory markers of inflammation from patients in intensive care from January 2009 to December 2019 were evaluated. GPT-4 ADA, accessed between December 2023 and January 2024, was tasked with autonomously analyzing this dataset by plotting radiography usage rates, providing descriptive statistics measures, quantifying factors of pulmonary opacities, and setting up machine learning (ML) models to predict their presence. Three scientists with 6-10 years of ML experience validated the outputs by verifying the methodology, assessing coding quality, re-executing the provided code, and comparing ML models head-to-head with their human-developed counterparts (based on the area under the receiver operating characteristic curve [AUC], accuracy, sensitivity, and specificity). Statistical significance was evaluated using bootstrapping. Results A total of 43 788 radiograph reports, with their laboratory values, from University Hospital RWTH Aachen were evaluated from 43 788 patients (mean age, 66 years ± 15 [SD]; 26 804 male). While GPT-4 ADA provided largely appropriate visualizations, descriptive statistical measures, quantitative statistical associations based on logistic regression, and gradient boosting machines for the predictive task (AUC, 0.75), some statistical errors and inaccuracies were encountered. ML strategies were valid and based on consistent coding routines, resulting in valid outputs on par with human specialist-developed reference models (AUC, 0.80 [95% CI: 0.80, 0.81] vs 0.80 [95% CI: 0.80, 0.81]; P = .51) (accuracy, 79% [6910 of 8758 patients] vs 78% [6875 of 8758 patients], respectively; P = .27). Conclusion LLMs may facilitate data analysis in radiology, from basic statistics to advanced ML-based predictive modeling. © RSNA, 2024 Supplemental material is available for this article.

Background Cardiac MRI at 5 T has recently become available and potentially improves tissue contrast enhancement at gadolinium chelate-enhanced T1-weighted imaging. Purpose To evaluate the feasibility of 5-T myocardial late gadolinium enhancement (LGE) MRI in assessing myocardial fibrosis by comparing image quality and LGE quantification with reference-standard 3-T myocardial LGE MRI. Materials and Methods Consecutive patients with confirmed myocardial fibrosis on previous 3-T MRI scans between January 2023 and July 2023 prospectively underwent follow-up imaging from August 2023 to November 2023. Each participant underwent follow-up 5-T imaging using an identical dose of contrast agent. Radiologist scoring of image quality using a Likert scale (range, 1-5), signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), contrast ratio, and semiautomatic quantitative LGE assessment were obtained and reported as medians and IQRs. Paired Wilcoxon signed rank tests were used to compare characteristics derived at 3-T and 5-T imaging. Results A total of 18 participants (mean age, 49 years ± 17 [SD]; nine male participants) were included, with a mean interval of 6.2 months ± 2.3 between undergoing 3-T and 5-T MRI. Median image quality scores were 4.0 (IQR, 3.0-4.2) at 3 T and 4.0 (IQR, 3.0-4.4) at 5 T (P = .45). SNR at 5 T was higher than at 3 T (183.7 [IQR, 147.2-255.9] vs 125.8 [IQR, 108.2-171.6], respectively; P = .002). Median CNR at 5 T was higher than at 3 T in normal myocardium (50.8 [IQR, 35.4-67.9] vs 16.5 [IQR, 11.3-24.6], respectively) and pericardial fat (21.4 [IQR, 7.1-29.3] vs -5.0 [IQR, -16.4 to -2.3], respectively) (both P < .001). There was no evidence of a difference in the percentage of LGE quantified between 5 T and 3 T (median, 11.8% [IQR, 7.7%-20.5%] vs 12.6% [IQR, 6.6%-20.4%], respectively; P = .81). Conclusion Myocardial LGE MRI at 5 T was found to be feasible, with no evidence of differences in subjective image quality and myocardial fibrosis quantification compared with 3-T myocardial LGE MRI. Furthermore, with use of identical contrast agent doses, SNRs and CNRs were improved at 5 T. Published under a CC BY 4.0 license. Supplemental material is available for this article. See also the editorial by Czum in this issue.