Academic Radiology最新文献

Rationale and objectives: Post-SABR radiologic appearances in the lung are dynamic and may imitate local failure. Although radiologic changes after SABR have been described, the dynamic evolution and predictors of pseudoprogression remain unclear. We aimed to characterize serial computed tomography (CT) evolution after SABR and examine its association with pseudoprogression.

Materials and methods: We retrospectively included patients who underwent SABR and had ≥24 months radiologic follow-up without definite local recurrence. The maximum diameter of the evolving post-SABR lesion was referenced to the initial (pre-SABR) tumor. Pseudoprogression was defined as any >20% increase in maximum diameter during follow-up that was followed by radiographic stabilization and lacked pathologic or clinical evidence of recurrence.

Results: We evaluated 109 lesions from 104 patients (median follow-up, 53 months). Mass-like consolidation developed in 94/109 (86%) lesions. CT evolution followed three phases with typical onset: initiation-reticular infiltrates adjacent to the index tumor (median 3 months; range 1-9); development-increasing density with patchy/streaky opacities and/or fibrotic consolidation (median 4.5 months; range 1.5-15.5); and mass-like consolidation-organization/absorption or retraction culminating in a final mass-like consolidation (median 13 months; range 5.5-39). A total of 74 (68%) lesions showed a >20% increase vs the initial tumor, and 50% occurred during the third phase. Planning target volume (PTV) was an independent predictor of pseudoprogression (adjusted odds ratio 0.98, 95% CI 0.97-0.99; P = 0.002).

Conclusion: Dynamic CT evolution after SABR commonly manifests as mass-like consolidation with high rates of pseudoprogression. PTV independently predicts this phenomenon, underscoring the need for cautious interpretation of radiographic enlargement during follow-up.

Rationale and objectives: Current assessment of parametrial invasion (PMI) in cervical cancer relies on subjective gynecological palpation and conventional MRI, often leading to diagnostic inconsistencies and treatment inaccuracies. This study aimed to develop an objective model integrating Diffusion Kurtosis Imaging (DKI) parameters with clinical variables to improve PMI assessment.

Materials and methods: This prospective study enrolled 90 patients with cervical cancer. All participants underwent 3.0T MRI, including DKI. Clinical variables and DKI parameters (mean diffusivity-MD, mean kurtosis-MK, axial kurtosis-Ka, radial kurtosis-Kr) were analyzed. Predictors of PMI were identified using multivariable logistic regression with Akaike Information Criterion (AIC) backward selection. Three models were constructed and compared: a Clinical model, a Clinical+ADC model, and a Clinical+DKI fusion model. Model performance was evaluated by the area under the receiver operating characteristic curve (AUC) with 5-fold cross-validation, and interpretability was assessed using SHapley Additive exPlanations (SHAP).

Results: Multivariable analysis identified tumor size, MK, Kr and Ka as independent predictors of PMI (all P<0.05). The Clinical-DKI fusion model demonstrated significantly superior predictive performance, achieving an AUC of 0.87 (95% CI: 0.80-0.94), which was significantly higher than the Clinical-only model (AUC=0.77, P=0.017) and the Clinical+ADC model (AUC=0.77, P<0.013). The model's robustness was confirmed by 5-fold cross-validation (AUC=0.85). SHAP analysis confirmed MK as the most influential predictor CONCLUSION: The integration of DKI parameters, particularly MK, with clinical factors provides a non-invasive and objective tool for pretreatment PMI prediction, significantly improving upon conventional assessments.

Rationale and objectives: Predicting neoadjuvant chemotherapy (NACT) efficacy is vital for advanced nasopharyngeal carcinoma (LA-NPC) management. Existing models have limited generalizability. The combination of MRI-based deep learning features (DLF) and Vision Transformer (ViT) for this purpose remains unexplored. This study therefore aims to evaluate the value of multi-sequence MRI-based DLF combined with ViT for predicting NACT efficacy in LA-NPC.

Materials and methods: This study retrospectively enrolled 266 LA-NPC patients receiving standard NACT, categorized by RECIST 1.1 into CR and non-CR groups, and split into training and testing sets (3:1). Traditional radiomics and 2D/2.5D/3D deep learning models were built and compared. Select the optimal architecture as the feature extractor, features were reduced via PCA and input into ViT.

Results: The XGBoost model on T2-FS sequences performed best among traditional radiomics models, with a validation AUC of 0.760. For deep learning models, performance improved with model complexity: 2D models were least effective (AUC: 0.502-0.653), followed by 2.5D (best AUC: 0.713), while 3D models were optimal (best AUC: 0.755). Ultimately, we integrated the deep learning features extracted from the two optimal single models (2.5D-ResNet50 and 3D-DenseNet121) and input them into the ViT architecture for global context modeling. This fused model achieved superior performance, with a validation AUC of 0.926, accuracy of 0.903, and F1-score of 0.927, significantly outperforming all previous models (all P < 0.05).

Conclusion: The integrated model combining multi-sequence MRI DLF with ViT significantly enhances predictive performance for NACT efficacy in LA-NPC.

Rationale and objectives: To investigate the clinical value of time-dependent diffusion MRI(Td-dMRI) for predicting IDH and 1p/19q codeletion status in adult-type diffuse gliomas.

Materials and methods: 174 patients with diffuse glioma who underwent preoperative Td-dMRI and diffusion-weighted imaging (DWI) between June 2022 and September 2024 were prospectively enrolled. Td-dMRI-based parameters, including Diameter, V_in, Cellularity, and three apparent diffusion coefficients were quantified. DWI-derived ADC_tumor and normalized ADC_ratio were computed. The differences in Td-dMRI and DWI-derived-parameters between the different IDH and 1p/19q type gliomas were analyzed using t-tests and ANOVA followed by Bonferroni comparisons. Diagnostic performance was evaluated using the receiver operating characteristic curve (ROC) and the Delong test was used to compare the differences. Logistic regression model was constructed, with its performance evaluated using ROC curve, calibration curves, and the Hosmer-Lemeshow test. The correlation between histologically quantified and Td-dMRI-based parameters was assessed using Spearman's correlation.

Results: IDH wild-type gliomas exhibited smaller Diameter, higher V_in and higher Cellularity than IDH mutant-type gliomas. V_in yielded an AUC of 0.851 (95%CI:0.786-0.916), slightly higher than those of ADC_tumor (AUC = 0.777, 95%CI: 0.703-0.851) and ADC_ratio (AUC = 0.786, 95%CI: 0.714-0.858). Among IDH mutant-type glioma, 1p/19q codeleted glioma showed higher V_in and lower Diameter than non-codeleted glioma. Combined Diameter, V_in and Cellularity achieved the best performance (AUC = 0.908, 95%CI:0.836-0.979), while DWI failed to identify 1p19q genotypes. The Td-dMRI estimated parameters correlated well with the pathologic measurements (r = 0.78-0.79, p<0.001).

Conclusion: Td-dMRI is an effective method for predicting IDH and 1p/19q codeletion subtypes in adult-type diffuse gliomas.

Rationale and objectives: To develop a cluster-specific magnetic resonance (MR) radiomics model for predicting induction chemotherapy (ICT) response in nasopharyngeal carcinoma (NPC) with enhanced interpretability using Shapley Additive exPlanations (SHAP).

Materials and methods: In this single-center, retrospective study, 225 patients with NPC were randomly assigned to the training (n = 158) and test (n = 67) cohorts. Tumor burden reduction ratio (TBRR), derived from pre- and post-ICT MR images, was used to classify patients as responders or non-responders. Tumor volumes of interest were subdivided into three radiomics-defined clusters, and radiomics features were extracted from each cluster and the whole tumor. Dimensionality reduction was performed using the intra-class correlation coefficient, Pearson correlation coefficient, and recursive feature elimination. Cluster-specific, whole-tumor, and multicluster radiomics models were constructed using six machine learning classifiers. Model performance was assessed by area under the curve (AUC), calibration, and decision curve analysis. SHAP was applied to interpret feature contributions at the cohort and individual levels.

Results: Cluster 1, representing the vascularized tumor periphery, was the largest and most stable subregion and showed the strongest correlation between volume reduction and overall TBRR (R = 0.726, P < 0.001). The cluster 1-specific support vector machine model achieved the best performance (AUCs of 0.812 and 0.800 in the training and test cohorts, respectively), with good calibration and clinical utility. SHAP highlighted features reflecting intratumoral heterogeneity and voxel intensity, explaining inter-patient variability.

Conclusion: The cluster 1-specific MR radiomics model reliably predicted the ICT response in NPC and may offer interpretable prognostic insights regarding treatment beneficiaries.

Rationale and objectives: While the origins of neuroradiology scholarship can be traced back to individual pioneers, the current status of institutional level neuroradiology scholarship has not been explored. We aimed to examine present day academic institutional contributions to the neuroradiological literature in the United States.

Materials and methods: Neuroradiology fellowship programs in the US were identified, and a Scopus search for numbers of papers published, citations, and h-index for each institutions' neuroradiology faculty between July 2025 and August 2025 was conducted. Median h-index, mean h-index and median paper numbers by institution were recorded.

Results: We identified 94 neuroradiology fellowship programs with 1282 faculty producing 63,769 manuscripts and 2.4 million citations, with Mayo Clinic Rochester (MCR) over-represented in both publications and citations. Rankings varied depending on metric: while MCR, University of California San Francisco, Massachusetts General Hospital, and Washington University consistently ranked highest overall, smaller programs such as University of California Davis and Thomas Jefferson University rose in h-index due to highly prolific individual faculty. There were discrepancies between published commercial diagnostic radiology program rankings and neuroradiology-specific scholarly output.

Conclusion: Several institutions across the nation make disproportionate contributions to neuroradiology scholarship. While scholarship can influence program reputation, program rankings do not necessarily parallel neuroradiology academic productivity.