Academic Radiology最新文献

Rationale and objectives: Hypoperfusion and related hypoxia are critical factors that contribute to therapeutic resistance in solid tumors. Ultrasound-stimulated microbubble (USMB) has been approved to enhance tumor perfusion, albeit with limited efficacy. This study was aimed to investigate whether combining USMB with alprostadil, a vasodilatory agent, could further improve tumor perfusion and alleviate hypoxia, thereby enhancing drug delivery.

Materials and methods: Sixty-nine rabbits bearing VX2 tumors were included in this study. USMB treatment was conducted using a modified diagnostic ultrasound at low intensity (mechanical index 0.24). Tumor perfusion was assessed using contrast-enhanced ultrasound. Hypoxia was evaluated by measuring hypoxia-inducible factor-1α (HIF-1α) and D-lactic acid (D-LA) levels. A pathway inhibition experiment was conducted to explore underlying mechanisms. Doxorubicin was administered to evaluate drug delivery efficacy.

Results: Tumor perfusion was increased following combination therapy, USMB or alprostadil monotherapy, with the combination treatment producing the most pronounced improvement. Furthermore, the combined therapy resulted in the most significant reduction in HIF-1α and D-LA. The pathway inhibition study revealed that USMB led to elevated adenosine triphosphate (ATP) levels in tumors, while cyclic adenosine monophosphate levels were reduced upon pathway inhibition. Nitric Oxide production was highest after combination treatment and markedly decreased following pathway inhibition. Notably, the concentration of doxorubicin within the tumor was highest following combined therapy.

Conclusion: The combination of USMB and alprostadil alleviates hypoperfusion and hypoxia in solid tumors synergistically, which is most likely related to the ATP signaling pathway. This protocol is an effective approach of enhancing drug delivery.

Rationale and objectives: PD-L1 expression is a critical biomarker in guiding immunotherapy for gastric cancer (GC). This study aims to investigate the value of deep learning analysis based on dual-energy CT-derived iodine map for predicting the level of PD-L1 expression in GC.

Methods: A total of 267 GC patients who underwent gastrectomy and preoperative dual-energy CT from multiple centers were prospectively enrolled and categorized into training (TC, n=143), internal validation (IVC, n=60), and external validation cohort (EVC, n=64). A 50-layer Residual Network was used to extract deep learning (DL) features from tumor volumes of interest on the iodine map. Machine learning was employed to develop the DL feature signature model (DFSigM). Multivariable logistic regression was used to screen PD-L1-related clinical characteristics, then a clinical model and a DL-clinical fusion model were also built. Model performance was evaluated based on discrimination, calibration, and clinical utility. Model interpretability was achieved through SHAP and Grad-CAM.

Results: Following feature selection, 12 key DL features were identified and utilized to construct DFSigM. DFSigM achieved AUC values of 0.854 in TC, 0.836 in IVC, and 0.818 in EVC, outperforming the clinical model (AUCs of 0.785, 0.720, and 0.695), while comparable to the fusion model (AUCs of 0.858, 0.828, and 0.833). DFSigM provided a high net clinical benefit across a wide range of threshold probabilities, and also demonstrated good agreement between the predicted and actual probabilities. SHAP and Grad-CAM visualized the decision-making process of the model.

Conclusion: A deep learning model based on iodine map has been proven to be a valuable, reliable, and interpretable tool for non-invasive prediction of PD-L1 expression in GC.

Background: Prostate cancer is the second most common cancer in men, with rising mortality rates necessitating precise risk stratification. High-invasive biological features-specifically International Society of Urological Pathology (ISUP) grade, extracapsular extension (EPE), and positive surgical margins (PSM)-are critical for guiding treatment but are difficult to detect due to tumor heterogeneity. Current imaging modalities, including 18F-PSMA-1007 PET/CT and multiparametric MRI (mpMRI), have limitations in fully capturing these features. This study aims to develop a few-shot deep learning model (CL-MGNET) that integrates multimodal imaging and clinical data to predict high-risk biological features, optimizing performance even with limited training data.

Materials and methods: This retrospective, multicenter study analyzed data from 377 patients: 341 from a primary medical center (Center A) and 36 from an independent external validation cohort (Center B). The study utilized multimodal inputs (PET/CT, mpMRI) and clinical variables to predict ISUP grade, EPE, and PSM. A specialized few-shot deep learning network, CL-MGNET, was designed to fuse these data sources. The model was trained using a restricted subset of 30 patients and subsequently evaluated on both internal and external test sets to assess generalizability across different centers.

Results: CL-MGNET demonstrated excellent performance in predicting high-invasive biological features (defined as the presence of at least one high-risk feature: ISUP ≥ 3, EPE, or PSM), achieving an internal test AUC of 0.877 and an external validation AUC of 0.872, which significantly outperformed the clinical model with an AUC of 0.792. The model surpassed both single-modality models (PET/CT, mpMRI) and the clinical model. Furthermore, CL-MGNET exhibited strong generalization capability, effectively predicting various high-risk biological features. When clinical variables were integrated, the model's performance improved significantly, exceeding traditional methods.

Conclusion: The CL-MGNET model, leveraging multimodal imaging data and clinical variables with a few-shot learning approach, successfully predicts high-invasive biological features of prostate cancer with high accuracy, even with limited data. The model's performance across different biological features and medical centers shows its robust generalizability. This method holds great promise for improving prostate cancer diagnosis and risk prediction in data-limited environments.

Background: Paracentesis and thoracentesis are essential procedures increasingly performed by radiologists. Prior national studies demonstrated rising radiologist involvement, but contemporary, practice-level data are limited.

Methods: We performed a multicenter retrospective review of internal billing data from 92 U.S. practice sites (2014-2022). Annual totals and per-radiologist averages for paracentesis and thoracentesis were calculated. Results were compared to publicly available Medicare Physician & Other Practitioners datasets (2017-2022), capturing national volumes across all specialties.

Results: From 2014 to 2022, radiologists in the dataset performed 93,037 paracenteses and 50,357 thoracenteses. Annual paracentesis volume increased from 3105 (2014) to 16,891 (2022), while thoracentesis rose from 1943 (2014) to 9712 (2022). Yearly average procedures per radiologist increased substantially (paracenteses: 38.3 → 66.8; thoracenteses: 25.9 → 43.4) as did yearly average procedures per site 280.4 (2014) to 436.1 (2022). National Medicare totals remained stable or declined slightly from 2017 to 2022.

Conclusion: Radiologists now perform the majority of paracentesis and thoracentesis procedures in the United States, despite stable national volumes. This shift underscores the need for strategic workforce planning, training emphasis in radiology residency, and health policy adjustments to support the expanding procedural role of radiologists.

Summary sentence: Radiologists now perform a majority of paracentesis and thoracentesis procedures in the United States, reflecting a major shift in procedural burden with implications for workforce planning, training, and health policy.

Rationale and objectives: To evaluate the prognostic value of radiomic features derived from contrast-free cine cardiac magnetic resonance (CMR) in patients with ST-segment elevation myocardial infarction (STEMI).

Materials and methods: We retrospectively included 440 patients with acute STEMI (86.6% males, 56.9 ± 10.6 years of age), who underwent CMR one week after percutaneous coronary intervention. Patients were assigned by centers into a development cohort (n = 359) and a validation cohort (n = 81). Radiomic features were extracted from cine images. Feature selection was performed using random survival forest and least absolute shrinkage and selection operator (LASSO)-Cox regression to generate a radiomics-based risk score (RAD score). Discrimination was evaluated using logistic and Cox regression analysis.

Results: During the median follow-up period of 2.9 years, 88 patients experienced major adverse cardiovascular events (MACE). The RAD score provided incremental prognostic value over the clinical model in the internal (C-index 0.86 [0.79-0.92] vs 0.65 [0.55-0.78]; p < 0.001) and external cohort (C-index 0.80 [0.70-0.91] vs 0.63 [0.48-0.78]; p = 0.014), comparable to the clinical + LGE-CMR model (C-index 0.80 [0.70-0.91] vs 0.77 [0.65-0.89]; p = 0.547). Receiver operating characteristic analyses were consistent with C-index findings. After adjusting for established risk factors, RAD score-defined high risk remained independently associated with MACE (HR 11.30, 95% CI 4.96-21.44; p < 0.001).

Conclusion: Cine-CMR radiomics provides independent and incremental prognostic information after STEMI and attains predictive performance comparable to parameters from cardiac magnetic resonance with late gadolinium enhancement, supporting contrast-free, individualized risk stratification.

Rationale and objectives: Accurate contouring of the Gross Tumor Volume (GTV) in High-Grade Gliomas (HGGs) is a cornerstone of effective Radiation Therapy (RT) planning, as it influences tumor control and spares normal tissue, thereby directly impacting treatment precision. However, the standard manual approach to GTV contouring requires considerable time and is prone to inter-observer variability. Accordingly, this study presents a deep learning framework for automatic GTV contouring in HGG cases.

Materials and methods: A modified 3D U-Net architecture was employed and trained on 469 subjects sourced from the Brain Tumor Segmentation (BraTS) 2018-2019 challenges, with multi-sequence magnetic resonance imaging (MRI) to enhance feature learning. The GTV was delineated following the European Society for Radiotherapy and Oncology (ESTRO) and the European Association of Neuro-Oncology (EANO) guidelines, based on the contrast-enhancing region of the tumor on post-contrast T1-weighted images, excluding edema. This corresponds to the enhancing tumor and necrotic core labels in our dataset. The segmentation accuracy was assessed using the Dice Similarity Coefficient (DSC) and the 95th-percentile Hausdorff Distance (HD95).

Results: The proposed model yielded a DSC of 91.70% ± 4.62% (mean ± standard deviation) and an HD95 of 2.43 ± 1.30 mm, indicating a high degree of overlap with minimal boundary deviation.

Conclusion: The results of our study highlight the potential of deep learning as a promising and efficient solution for GTV contouring in HGGs, supporting RT planning, improving clinical workflow, and enhancing treatment accuracy.

Rationale and objectives: This study aimed to develop a deep learning model using a novel pixel-level radiomics approach based on two-dimensional (2D) and strain elastography (SE) ultrasound images to predict Ki-67 expression in breast cancer (BC).

Methods: This multicenter study included 1031 BC patients, who were divided into training (n = 616), internal validation (n = 265), and external test (n = 150) cohorts. An additional 63 patients were prospectively enrolled for further validation. The deep learning model, termed Vision-Mamba, predicts Ki67 expression by integrating ultrasound (2D and SE) images with pixel-level radiomics feature maps (RFMs). A combined model was subsequently constructed by incorporating independent clinical predictors. Model performance was assessed using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) were applied to enhance interpretability.

Results: We developed a Vision-Mamba-US-RFMs-Clinical (V-MURC) model that integrates ultrasound images, RFMs, and clinical data for accurate prediction of Ki-67 expression in BC. The area under the ROC curve (AUC) values for the internal validation, external test, and prospective validation cohorts were 0.954 (95% CI, 0.929 - 0.975), 0.941 (95% CI, 0.903 - 0.975), and 0.945 (95% CI, 0.883 - 0.989), respectively, demonstrating excellent discrimination and calibration. Compared with individual models, the V-MURC model achieved significantly superior performance across all datasets (Delong test, P < 0.05). Calibration curves and DCA further supported its clinical applicability. SHAP analysis provided visual interpretability of the model's decision-making process.

Conclusion: The V-MURC model based on pixel-level RFMs can accurately predict Ki-67 expression in BC and may serve as a valuable tool for individualized treatment decision-making in clinical practice.