Academic Radiology最新文献

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.

Rationale and objectives: Breast cancer screening via mobile mammography units (MMUs) can improve access in medically underserved communities. This study aims to evaluate factors associated with screening site, recall rates, and time to diagnostic resolution.

Materials and methods: This retrospective study analyzed recall rates, time to diagnostic resolution, and sociodemographic factors in patients who underwent screening mammograms in an MMU versus urban hospital system during overlapping two-week periods in 2022 and 2023. For patients with BI-RADS 0 (incomplete) screening mammograms, our main analytic cohort, time intervals between screening and diagnostic imaging and, when indicated, between diagnostic imaging and biopsy, were measured. Diagnostic resolution was defined as time from screening to BI-RADS 1 (negative), 2 (benign), or 3 (probably benign) on diagnostic mammogram or, when indicated (BI-RADS 4 or 5 [suspicious or highly suspicious for malignancy, respectively]), from screening to biopsy. Chi-square, analysis of variance, and Kruskal-Wallis tests were performed to compare MMU- and hospital-screened women's characteristics. Cox regression analysis was used to assess factors associated with diagnostic resolution.

Results: In the MMU cohort (n = 97) versus the hospital-based cohort (n = 236), more patients identified as Non-Hispanic Black (68% versus 40%), were uninsured (71% versus 2.1%), and had no primary care provider (35% versus 9.8%, all p<0.001). The MMU cohort also had a higher recall rate (18.8% versus 9.9%, p<0.001). Among BI-RADS 0 screening mammograms (n = 333), time to diagnostic resolution was longer among MMU- versus hospital-screened women (median 28 [IQR 15-51] vs 11 days [IQR 7-20], p<0.001). Patients with no insurance had a lower likelihood of diagnostic resolution (HR 0.42, 95% CI [0.26,0.69], p = 0.001). In the MMU cohort, 17/97 (18%) did not return for the recommended diagnostic imaging versus 9/236 (3.8%) in the hospital-screened cohort (p<0.001).

Conclusion: Although MMUs can improve access, our pilot study highlights opportunities to promote timely and equitable follow-up of abnormal screening mammograms through improved patient navigation, social-work support, and financial assistance.

Purpose: This study aims to develop and validate an interpretable machine learning model that integrates clinical data, radiomics, and deep learning (DL) features extracted from ¹⁸F-AlF-NOTA-Pentixafor positron emission tomography/computed tomography (PET/CT) images for the non-invasive prediction of pathological subtypes in primary aldosteronism (PA).

Methods: In this single-center retrospective study, we included 89 patients diagnosed with PA or non-functioning adrenal adenomas who underwent ¹⁸F-Pentixafor PET/CT between February 2024 and May 2025. Predictive models were built by integrating clinical data, PET/CT radiomics, and DL features. A two-stage feature selection strategy was employed, which utilized the minimum redundancy maximum relevance method followed by stepwise regression based on the Akaike information criterion. Four distinct models were constructed using the support vector machine algorithm, and their hyperparameters were optimized via stratified five-fold cross-validation. Model performance was rigorously evaluated by the area under the receiver operating characteristic curve (AUC), calibration analysis, and decision curve analysis. Furthermore, model interpretability was achieved using Shapley Additive Explanations (SHAP) to elucidate feature contributions.

Results: The combined model demonstrated superior diagnostic accuracy in the test set, with an AUC of 0.907, perfect sensitivity (1.000), and an F1-score of 0.923. It significantly outperformed the clinical, radiomics, DL models individually (p<0.01). SHAP analysis identified lesion-to-adrenal ratio, maximum standardized uptake value, and selected PET/CT radiomics and DL features as key contributors, revealing biological alignment with CXCR4 and CYP11B2 expression.

Conclusion: An interpretable machine learning model can non-invasively predict surgically confirmed PA subtypes, defined by immunohistochemistry for CYP11B2. This approach may reduce the reliance on invasive adrenal vein sampling and facilitate personalized surgical decision-making.

Rationale and objectives: To determine the publication rates and characteristics of oral scientific presentations from the European Society of Gastrointestinal and Abdominal Radiology (ESGAR) meetings held between 2019 and 2022, and to identify factors associated with subsequent publication.

Materials and methods: This retrospective observational study analyzed 407 oral abstracts from ESGAR meetings (2019-2022). Abstract data were categorized by country, subspecialty, study design, and collaboration type. Publication searches were performed in PubMed. Publication time, journal name, journal impact factor (JIF), and citation counts were recorded. Statistical analyses included chi-square, logistic regression and Kruskal-Wallis tests.

Results: Of 407 oral presentations, 215 (52.8%) were subsequently published in PubMed-indexed journals, significantly higher than rate from ESGAR 2000-2001 (39.5%) (P < .001). Median publication time was 11.3 months. Country of origin was significantly associated with publication outcome (P < .001). No significant differences were found in publication rates among subspecialties (P = .577). Prospective studies had higher JIF than retrospective studies (P = .004). International collaborations had higher JIF than local collaborations (P = .027).

Conclusion: More than half of ESGAR oral presentations achieved publication within 3 years, showing a clear increase compared with earlier meetings and reflecting enhanced research productivity and dissemination in gastrointestinal and abdominal radiology.

Aim: This study aimed to develop and validate a clinical-MRI quantitative parameter model to predict clinically significant prostate cancer (csPCa) in PI-RADS score 3 lesions.

Methods: A retrospective analysis was performed on 151 patients with PI-RADS score 3 lesions, divided into csPCa and non-csPCa groups according to pathological results. Patients were randomly assigned into training and validation cohorts in a 7:3 ratio. Quantitative values of T1, T2, and proton density (PD) were obtained from the synthetic magnetic resonance imaging (syMRI) quantitative maps, while apparent diffusion coefficient (ADC) values were derived from ADC maps. Independent predictors were identified using univariate and multivariate logistic regression analyses, based on which a quantitative parameter model was established. Clinical risk factors were used to construct a clinical model, and a combined model integrating both clinical and imaging predictors was developed. The predictive performance of the models was evaluated using the area under the curve (AUC), calibration curves, and decision curve analysis (DCA). The DeLong test was applied to compare the diagnostic efficiency between models.

Results: Multivariate logistic regression analysis revealed that prostate volume (PV) and prostate-specific antigen density (PSAD) were independent clinical predictors for csPCa, while T2 and ADC values were independent imaging predictors. In the training cohort, the combined model achieved an AUC of 0.91 (95% CI: 0.86-0.97), outperforming the clinical model (AUC = 0.76, 95% CI: 0.66-0.85, P = 0.001) and the quantitative parameter model (AUC = 0.84, 95% CI: 0.76-0.93, P = 0.017). DCA demonstrated that the combined model provided greater net clinical benefit compared to either model alone.

Conclusion: The clinical-quantitative parameter combined model can effectively identify csPCa within PI-RADS score 3 lesions based on syMRI, thereby guiding biopsy decisions, reducing unnecessary invasive procedures, and improving patients' quality of life.

Rationale and objectives: Unresectable hepatocellular carcinoma (uHCC) remains a formidable clinical challenge owing to the scarcity of effective treatment options and unsatisfactory therapeutic responses. The current study explored a combined regimen of RALOX-HAIC, lenvatinib, and camrelizumab in patients with uHCC. In addition, a radiomics-based nomogram was created to predict treatment outcomes and support individualized decision-making.

Methods: A total of 98 patients with uHCC received RALOX-HAIC, along with lenvatinib and camrelizumab. Before initiating therapy, radiomics features were derived from pretreatment computed tomography (CT) images and subsequently integrated with clinical variables, such as HBV status and Child-Pugh score. A radiomics nomogram was generated and assessed based on the area under the receiver operating characteristic curve (AUC), calibration analysis, and decision curve analysis (DCA).

Results: Triple therapy yielded an objective response rate (ORR) of 52.0%, disease control rate (DCR) of 90.8%, and median progression-free survival (PFS) of 10.7 months (95% CI: 7.3-20.5). The radiomics-guided nomogram showed high accuracy in the training (AUC: 0.986) and validation (AUC: 0.873) sets. The calibration curves showed close agreement between the projected and observed outcomes, and DCA confirmed the notable clinical merit. The main grade ≥3 toxicities included neutropenia and thrombocytopenia (68.4%), consistent with the profiles observed in comparable therapies.

Conclusion: The integrated approach exhibited promising antitumor activity and an acceptable safety profile. Moreover, the radiomics nomogram is a valuable tool for refining patient selection and advancing personalized treatment strategies for individuals with uHCC.

Rationale and objectives: As of September 2024, the FDA requires that breast imaging practices inform women about their breast density. This study aimed to evaluate the impact of breast density on the performance metrics of screening digital breast tomosynthesis (DBT) examinations.

Materials and methods: We retrospectively reviewed screening DBT examinations performed from 2013 to 2019 at a single academic medical center. Performance metrics were calculated according to the 5th Edition of the BI-RADS Atlas. Associations between breast density and screening performance were examined using multivariable logistic regression with generalized estimating equations.

Results: The cohort included 111,143 women (mean age, 59 ± 11 years) with 301,400 DBT examinations. Breast density was almost entirely fatty (category A) in 8.8%, scattered areas of fibroglandular density (B) in 50.5%, heterogeneously dense (C) in 36.9%, and extremely dense (D) in 3.8%. Cancer detection rates (CDR, per 1000 exams) were 3.4, 5.6, 5.2, and 3.7 for categories A-D, respectively. Sensitivities were 92.8%, 90.1%, 81.0%, and 61.8%. Specificities were 96.7%, 94.4%, 92.5%, and 93.3%. Category D was associated with significantly lower sensitivity than each of the other categories (adjusted odds ratios [aOR] 0.19-0.43, p<0.01 for all). It was associated with significantly lower specificity than almost entirely fatty tissue (aOR 0.64, p<0.001) but not the other two density categories.

Conclusion: Dense breast tissue significantly decreases the sensitivity of screening DBT. These findings highlight the need to report and consider breast density in screening recommendations and necessitate further research on more effective screening regimens for women with dense breast tissue.