Current Medical Imaging Reviews最新文献

Introduction The Breast Imaging Reporting and Data System (BI-RADS) category 4 is subdivided into 4A, 4B, and 4C to reflect varying levels of suspicion for malignancy. However, the predictive consistency of these subcategories remains debated, especially in underrepresented populations. This study aims to assess the correlation between BI-RADS 4 subcategories and histopathological outcomes in Mexican women, identifying additional demographic and imaging predictors of malignancy. Materials and Methods This retrospective cross-sectional study included 173 female patients with BI-RADS 4 lesions who underwent mammography and/or ultrasound, followed by histopathological confirmation. Data were collected from the Hospital General de México between January 2023 and May 2024. Associations between BI-RADS subcategories and malignancy, age, lesion laterality, and imaging features were analyzed using chi-square tests and ANOVA.

Results: Among 173 patients, 41.6% had BI-RADS 4A lesions, 35.8% had 4B, and 22.5% had 4C. Malignancy rates increased progressively across subcategories: 7.5% (4A), 40.0% (4B), and 85.0% (4C) (p < 0.001). The mean age rose with BI-RADS level (42.1, 47.8, and 55.3 years for 4A, 4B, and 4C, respectively), although this trend was not statistically significant (p = 0.063). Nodules were the most frequent imaging finding (83.2%), and fibroadenomas were the most common benign diagnosis. Left-sided lesions were more frequently malignant (p = 0.034).

Discussion: The BI-RADS 4 subcategorization showed a clinically meaningful, although not statistically significant, trend in malignancy risk. Lesion laterality emerged as a potential independent predictor of malignancy, warranting further investigation. The findings reinforce the complementary role of demographic and imaging variables in risk assessment.

Conclusion: The BI-RADS 4 subclassification aligns with increasing malignancy risk, supporting its clinical utility. However, variability in diagnostic outcomes suggests the need to integrate histopathological and demographic data. Lesion laterality may represent a novel factor in malignancy prediction among breast lesions. Sustainable Development Goals (SDGs) Keywords SDG 3 Good Health and Well-being; SDG 5 Gender Equality; SDG 10 Reduced Inequalities; SDG 9 Industry Innovation and Infrastructure; SDG 4 Quality Education; SDG 17 Partnerships for the Goals.

Introduction: Primary Cutaneous Lymphomas (PCLs) are rare extranodal non-Hodgkin lymphomas that present in the skin without extracutaneous disease at diagnosis; they are exceptionally uncommon in children and frequently mimic benign dermatoses, delaying recognition. We report a case series of three pediatric patients managed at a national referral center.

Case presentation: Case 1: A 16-year-old with erythrodermic mycosis fungoides (stage IIIA) refractory to multiple systemic and skin-directed therapies who achieved remission after haploidentical allogeneic hematopoietic stem-cell transplantation. Case 2: A 3-year-old with aggressive cytotoxic CD8-positive Tcell lymphoma with EBV-associated disease and severe infectious complications during CHOP-based therapy, culminating in death despite salvage treatments and EBRT. Case 3: A 10-year-old with CD30-positive primary cutaneous anaplastic large cell lymphoma with a relapsing course, treated with BFM-90 chemotherapy, gemcitabine, pegylated doxorubicin plus total skin electron beam therapy, and ongoing PUVA with good dermatologic control. Across cases, diagnosis relied on clinicopathologic correlation with immunohistochemistry and staging CT; serum IgE was elevated in all three children.

Conclusion: Pediatric PCLs show heterogeneous behavior and therapeutic responses. Early biopsy of atypical or treatment-refractory eruptions, comprehensive histopathology and immunophenotyping, targeted EBV testing when suggested, and appropriate use of skin-directed radiotherapy (EBRT/TSEBT) and transplantation in selected refractory disease are essential. Multidisciplinary management and equitable access to specialized therapies are critical to optimize outcomes.

Introduction: Bone scintigram is a highly effective medical imaging technique widely used for the rapid screening of bone metastases, contributing significantly to early disease detection and prognosis assessment. Neural architecture search enables automated design and optimization of network structures by leveraging data characteristics and task-specific requirements.

Objective: For the automated and accurate diagnosis of bone metastasis in bone scintigrams, this study proposes an improved differentiable neural architecture search framework to address two key limitations of the original DARTS method: (1) the insufficient representation capability of standard candidate operations for characterizing bone metastasis lesions, and (2) architecture degeneration.

Methods: Based on the DARTS framework, two novel candidate operations were developed, and the training supernet architecture was optimized. (1) The channel-attention-integrated residual operation incorporates synergistic channel-wise intelligent weighting and gradient stabilization mechanisms, providing an efficient yet highly discriminative feature transformation module for architecture search. (2) The spatial-attention-enhanced multibranch operation combines multi-scale feature fusion with spatially adaptive focusing, significantly improving the model's ability to localize critical regions and detect lesions of varying sizes. (3) A dual-path convolutional structure is introduced into the training supernet to dynamically optimize both channel-wise and spatial dimensions of shallow features, thereby generating highly discriminative representations for subsequent cell architecture search.

Results: Comprehensive evaluations on clinical datasets demonstrate the effectiveness of the proposed method, achieving an accuracy of 0.8451, precision of 0.8700, recall of 0.8447, F1-score of 0.8423, and an AUC of 0.92.

Conclusion: The proposed neural architecture search method effectively detects bone metastasis in bone scintigrams and outperforms existing state-of-the-art approaches. The newly developed candidate operations and supernet optimizations successfully address the limited representational capacity of standard convolutions in the original DARTS operations, leading to substantial improvements in classification performance.

Background: Precision meningioma diagnosis requires MRI advancements, yet faces three barriers: (1) limited clinical translation, (2) inconsistent multimodal data standards, and (3) mismatched algorithm-resource allocation. A bibliometric analysis can guide evidence-based innovation.

Methods: We conducted a comprehensive bibliometric analysis of 4,280 Web of Science articles using CiteSpace, Bibliometrix, and SciExplorer, with dual screening to ensure data quality.

Results: Meningioma MRI research exhibited an S-shaped growth pattern. Research hotspots are transitioning toward AI applications. 502 core authors contributed to 83% of publications, with notable cross-disciplinary collaboration. The U.S. and China dominated production, while Europe demonstrated exceptional efficiency. Institutions, including Harvard, led development. Seventeen core journals conformed to Bradford's law, with the knowledge foundation established by highly cited papers in the field.

Discussion: The findings reveal an AI-guideline temporal gap and a field-strength validation deficit, underscoring the need for equitable, low-field-compatible AI tools.

Conclusion: We systematically delineate three evolutionary stages: structural imaging, functional integration, and intelligent analysis. AI-driven models have achieved enhanced diagnostic accuracy (AUC 0.82-0.97), but remain limited by heterogeneous data standards, low algorithm interpretability, and uneven global resources. Multicentre standardized protocols, interpretable AI frameworks, and lightweight algorithms compatible with ≤1.5 T scanners should be prioritised. By integrating burst-sigma mapping with global equipment metrics, we provide quantitative evidence supporting a field-strength-agnostic strategy for equitable AI deployment.

Introduction: This study aimed to investigate the value of intratumoral and peritumoral radiomics in predicting the risk grade of gastrointestinal stromal tumors (GISTs) using contrast-enhanced computed tomography (CT) images.

Methods: A total of 217 pathology-confirmed GISTs were retrospectively enrolled and divided into low-risk and high-risk groups. Significant predictors were selected from clinical and radiological characteristics to build a prediction model. Radiomics features were extracted from the intratumoral region, the 3-mm peritumoral region, and the 5-mm peritumoral region. After ANOVA and LASSO feature screening, logistic regression was applied to construct the radiomics model. The Rad-score of the optimal radiomics model was calculated and combined with the selected radiological characteristics to develop a combined model and a nomogram. ROC curves were used to assess the predictive performance of each model, while calibration curves and decision curve analysis were used to evaluate their clinical utility. The SHapley Additive Explanations (SHAP) method was applied to perform interpretability analysis of the optimal model.

Results: A radiological model (RM), five radiomics models, and a combined radiological characteristics plus Rad-score model (CRM) were constructed. In the validation set, the AUCs of the RM and CRM were 0.839 and 0.924, respectively. The intratumoral plus 3-mm peritumoral radiomics model (ITV+PTV3) achieved the best performance in the validation set, with an AUC of 0.934.

Discussion: The ITV+PTV3 model shows strong potential for objective GIST risk stratification but requires multi-center prospective validation to ensure generalizability beyond the limitations of this retrospective dataset.

Conclusion: Radiomics models based on intratumoral and peritumoral regions perform well in predicting the risk grade of GISTs and may effectively guide accurate preoperative diagnosis and treatment planning.

Background: Early detection of breast cancer and accurate assessment of lesions are key goals of imaging evaluation. Ultrasound is widely used, but its diagnostic performance is influenced by complex image features, noise, and operator experience. Reducing operator dependence and improving accuracy are critical clinical issues.

Methods: In this retrospective study, 7,025 breast ultrasound images from our center were annotated based on pathology and split into training, validation, and internal test sets (8:1:1). The Dataset of Breast Ultrasound Images was used as the external test set. YOLO-v7 and YOLO-v8 models were trained through transfer learning after data augmentation and balancing the classes. Performance was compared on internal and external test sets and was evaluated against a reader study.

Results: YOLO-v7 and YOLO-v8 reached optimal performance at epochs 294 and 135, respectively. YOLO-v7 slightly outperformed YOLO-v8 on the internal test set, while YOLO-v8 achieved higher accuracy, recall, specificity, precision, and F1 score on the external test set. Both models showed significantly higher accuracy, specificity, and precision than the senior radiologist, with YOLO-v8 achieving a significantly higher F1 score.

Discussion: YOLO-v8 demonstrated better generalization due to its anchor-free mechanism and deeper architecture, while YOLO-v7 showed signs of overfitting. Both models outperformed the junior radiologist and approached or exceeded the diagnostic performance of the senior radiologist, indicating potential to assist less experienced readers.

Conclusion: YOLO-v7 and YOLO-v8 effectively classified breast lesions. YOLO-v8 showed faster convergence and higher diagnostic efficiency, suggesting strong potential for clinical application.

Background Bow hunter's syndrome (BHS), also known as rotational vertebral artery occlusion syndrome, is a hemodynamic disorder caused by mechanical compression of the vertebral artery during head rotation or hyperextension. This compression may lead to transient occlusion or significant stenosis, resulting in posterior circulation ischemia. BHS is relatively rare in clinical settings, and most reported cases occur in middle-aged or elderly patients. Its etiology is commonly associated with degenerative cervical conditions, such as osteophyte formation or disc herniation. In addition, the condition most often involves unilateral vertebral artery compromise. Case Presentation In this study, we report a rare case of BHS in a 15-year-old adolescent without noticeable cervical degenerative changes who experienced recurrent, unexplained posterior circulation cerebral infarctions. Dynamic cervical magnetic resonance angiography (MRA) revealed significant compression of both vertebral arteries during head and neck rotation, confirming a diagnosis of bilateral BHS. This presentation differs from the conventional understanding that BHS predominantly affects adults and is typically unilateral, suggesting that the diagnosis should also be considered in young patients, even in the absence of typical cervical lesions. Conclusion This study is the first report identifying bilateral BHS as the cause of recurrent posterior circulation infarction in a teenager using dynamic MRA. Although dynamic digital subtraction angiography remains the gold standard for diagnosis, this case highlights the practical value of dynamic MRA in diagnosing BHS.

Introduction: This study aimed to evaluate the impact of varying slice thickness on quantitative values using the Magnetic Resonance Image Compilation (MAGiC) sequence.

Methods: In this retrospective study, 23 healthy subjects underwent the MAGiC sequence (at 3.0 T) with three slice thicknesses: 3 mm (TH3), 4 mm (TH4), and 5 mm (TH5). The T1, T2, and PD values were measured in various knee joint cartilage regions by two experienced radiologists, including the lateral femoral condyle (LFC), lateral tibial plateau (LTP), medial femoral condyle (MFC), medial tibial plateau (MTP), patella (PAT), and trochlea (TRO). The effects of varying slice thicknesses (TH4 vs. TH3 and TH5 vs. TH3) were analyzed using paired t-tests or Wilcoxon signed rank tests, with statistical significance set at P < 0.025. Intra-rater and inter-rater reliability were also assessed.

Results: Measurements of T1, T2, and PD values demonstrated high intra- and inter-rater reliability. Minimal differences were observed across slice thicknesses for T1WI, T2WI, and PDWI images. T2 and PD values showed little variation, while T1 mapping revealed significant differences. T2 values were consistent across regions, except for the LFC.

Discussion: TH4 and TH5 can replace TH3 for knee joint scanning while reducing scan time, with minimal differences in anatomical depiction across sequences. MAGiC technology significantly improves efficiency by acquiring quantitative data in a single scan, demonstrating stable T2 values unaffected by slice thickness, though T1 and PD values are thickness-dependent. This technique holds clinical value for cartilage injury assessment but requires further research on the applicability of multiplanar imaging.

Conclusion: T2 values obtained with the MAGiC sequence are stable across TH3, TH4, and TH5, allowing for reliable cartilage T2 quantification using TH5 to reduce patient scan time.

Introduction: Lymphovascular invasion (LVI) is a critical prognostic factor in breast cancer, typically diagnosed via postoperative pathology. This study aimed to evaluate whether fat-suppressed T2-weighted imaging (FS T2WI)-based breast edema score (BES) combined with clinicopathological features could preoperatively predict LVI status.

Materials and methods: This retrospective study enrolled 574 breast cancer patients who underwent MRI and surgery from January 2021 to December 2023. Patients were classified as LVI-positive (n=174) or LVI-negative (n=400) based on postoperative pathology. Breast edema on FS T2WI was scored from 1 to 4 (BES 1, no edema; BES 2, peritumoral edema; BES 3, prepectoral edema; and BES 4, subcutaneous edema). Univariate and multivariate binary logistic regression analyses were performed to identify risk factors associated with LVI. A clinicopathological model and a combined BES-clinicopathological model were constructed, and diagnostic performance was evaluated using the area under the receiver operating characteristic curve (AUC) and the DeLong test.

Results: Multivariate analysis revealed that BES and clinicopathological factors, including age, axillary lymph node metastasis, and tumor size, were independent predictors of LVI. Compared with BES 1, tumors with BES 2, BES 3, and BES 4 were associated with a 1.825-, 2.047-, and 4.341- fold increased LVI risk, respectively. The combined BES-clinicopathological model outperformed the clinicopathological model alone (AUC, 0.765 vs. 0.778; P<0.05).

Discussion: Higher BES was independently associated with increased LVI risk. The predictive model integrating BES with clinicopathological variables outperformed single-parameter models, suggesting that BES may provide complementary imaging biomarkers for assessing tumor aggressiveness. Validation in larger, multicenter cohorts is warranted.

Conclusion: FS T2WI-based BES combined with clinicopathological features may improve preoperative prediction of LVI in breast cancer and support individualized treatment planning.