This preprint is retracted at the authors' request due to a patent application.

The binary diagnostic approach does not reflect the entire spectrum of metabolic dysfunction associated steatotic liver disease (MASLD. We used an elastography technology, dual elastography ultrasound (DEUS), to discriminate the different stages of MASLD. This prospective multicenter study was conducted from December 2020 to March 2022. All patients underwent DEUS scan, a liver biopsy, and a liver function laboratory test. The optimal model was developed (ModelDEUSC) with 10 machine learning algorithms by combining DEUS and selected clinical parameters and tested the diagnostic accuracy for distinguishing the three progression stages of MASLD: low-, intermediate-, and high-risk. The diagnostic ability of ModelDEUSC for MASH with advanced fibrosis (≥ F3) was compared with other four non-invasive tests. The study included 312 patients in the derivation cohort and 135 in the validation cohort (7:3). Combining DEUS and clinical parameters, a ternary classification of MASLD in the validation cohort achieved a macro-average AUC of 0.858 (95

The development of intelligent diagnostic models for ultrasound images is often hindered by data-related limitations, including limited sample size, class imbalance, and insufficient diversity, which impede training robust and generalizable diagnostic models. Recently, diffusion models have shown remarkable success in natural image generation, which motivates us to investigate their potential in addressing these limitations and enhancing ultrasonic diagnosis. Here we propose the Spatial-Aware Latent diffusion-empowered Image Augmentation (SALIA) approach, to generate diverse and high-quality synthetic ultrasound images and enhance the performance of tumor diagnosis. We ensure the generation of high-quality synthetic images by editing them in the latent space, which involves relocating the tumor under anatomical constraints with a tumor range restriction and seamlessly fusing it with the background. Moreover, we design a comprehensive series of methods for assessment of image quality and fidelity including the no-reference image quality assessment and radiologist-conducted visual evaluation, and we construct various classification models to validate the improvement in tumoral diagnostic performance by image augmentation with synthetic edited images. Experimental results on three ultrasound datasets of breast, thyroid and liver, indicate that our synthetic images closely mimic the appearance of real images, and they contribute to enhanced performance in tumor diagnosis with areas under the receiver operating characteristic curves (AUCs) of 0.984, 0.877 and 0.800 on the three datasets, respectively. Our SALIA approach provides a robust image augmentation framework for generating synthetic ultrasound images with high quality and fidelity, thereby advancing the accuracy and reliability of tumor diagnosis.

BACKGROUND:Quantitative ultrasound techniques enable noninvasive assessment of hepatic steatosis, fibrosis, and inflammation in metabolic dysfunction-associated steatotic liver disease (MASLD). This study performed a head-to-head comparison of dual-elastography and 2D shear wave elastography (2D-SWE) for comprehensive histologic evaluation. METHODS:A total of 186 biopsy proven MASLD patients were enrolled. Dual-elastography provided attenuation imaging (ATI), fibrosis (F-index), and inflammatory activity (A-index), while 2D-SWE offered attenuation coefficient (ATT), shear wave elasticity (SWE), and shear wave dispersion (SWD). Histologic grades of steatosis (S0-S3), fibrosis (F0-F3), and inflammation (A0-A3) served as the reference standard. Diagnostic performance was assessed using receiver operating characteristic (ROC) analysis. RESULTS:ATI outperformed ATT for detecting ≥S1 steatosis (AUROC 0.90 vs. 0.81, P = 0.036), while their performance for ≥S2 and ≥S3 was comparable. For significant fibrosis (≥F2), the F-index showed higher accuracy than SWE (AUROC 0.87 vs. 0.80, P = 0.046) with greater sensitivity (73.7%) and balanced specificity (85.3%). SWD demonstrated moderate diagnostic ability for inflammatory activity (AUROC 0.75 for ≥A2; 0.84 for ≥A3), and the A-index achieved AUROC 0.73 for detecting lobular inflammation grade ≥2. CONCLUSIONS:ATI and ATT are reliable for assessing steatosis, the F-index provides superior accuracy for significant fibrosis, and SWD and A-index reflect overall and lobular inflammation, respectively. These multiparametric ultrasound techniques enable comprehensive, noninvasive evaluation of key histologic features in MASLD.

To develop and validate a machine learning (ML) model integrating dual elastography, clinical features, and serum biomarkers for noninvasive prediction of severe drug-induced liver injury (DILI). This prospective multicenter study enrolled consecutive DILI patients undergoing liver biopsy and dual elastography. Severe DILI was defined as Scheuer inflammation grade plus fibrosis stage ≥ 5 (G + S ≥ 5). Dual elastography-derived activity index (A index) and fibrosis index (F index) correlated with pathological inflammation (G0–4) and fibrosis (S0–4) stages. The dataset was stratified and split 7:3 into training and test sets. LASSO regression was applied for feature selection. Eight ML models were constructed and compared, optimized using 5-fold cross-validation and Bayesian methods. Performance was evaluated by area under the curve (AUC), sensitivity, and specificity. SHapley Additive exPlanations (SHAP) were used to interpret the models. A total of 305 participants were included (median age 49 years, IQR 40–56; 98 male), comprising 55 with severe DILI and 250 without. A and F indices increased with inflammation grade and fibrosis stage, respectively (p < 0.01). Combining clinical and dual elastography features with serum biomarkers, the optimized regularized regression model performed best in the test set (AUC 0.862 [95 https://wznng666.shinyapps.io/RR55555/ .

BACKGROUND:Cerebral hyperperfusion syndrome (CHS) is a serious complication following revascularization in moyamoya disease (MMD) patients, yet reliable predictors remain scarce. This study aims to evaluate subcortical hemodynamics intraoperatively using a novel ultrasound imaging technique, synchronous arbitrary gate spectral Doppler (SAGSD), and to identify indicators associated with CHS. METHODS:A total of thirty adult MMD patients undergoing revascularization were included. Intraoperative SAGSD imaging, conventional Doppler ultrasound, and indocyanine green videoangiography (ICG-VA) were performed to assess subcortical hemodynamics at multiple sites before and after anastomosis. RESULTS:All thirty patients underwent revascularization, and five of the six CHS patients exhibited a distinctive "mountain" sign on SAGSD, characterized by three or more discrete velocity peaks within a single cardiac cycle observed simultaneously across two or more sampling sites. The CHS patients showed a significant increase in velocity-time integral (VTI) change (median Δ + 0.48 cm [95% CI, 0.31 to 0.65], p < 0.001) post-anastomosis on SAGSD, while conventional Doppler showed no significant change in flow velocity (all p > 0.05). Notably, in CHS patients, ICG-VA showed no statistical difference in flow velocity, delay, and time to peak after anastomosis (all p > 0.05), whereas SAGSD demonstrated hemodynamic changes such as VTI (median Δ + 0.67 cm [95% CI, 0.43-0.90], p < 0.001) via Doppler spectrum. CONCLUSION:SAGSD enables highly sensitive, multi-position, and contrast-free hemodynamic evaluation of deep cerebral microvasculature. The "mountain" sign, a novel spectral morphology corroborated by quantitative VTI elevation, is a specific intraoperative biomarker strongly associated with CHS, offering a potential predictor for early intervention and improved surgical outcomes.

OBJECTIVE:Renal microvascular rarefaction is an early pathological event in chronic kidney disease (CKD) that may precede a measurable decline in renal function. However, currently available non-invasive tools remain limited in detecting these early structural alterations. This study aimed to evaluate the value of super-resolution contrast-enhanced ultrasound (SR-CEUS) for assessing renal cortical microvascular impairment in early CKD with preserved estimated glomerular filtration rate (eGFR), and to compare its performance with conventional contrast-enhanced ultrasound (CEUS) and serum biomarkers. METHODS:In this prospective study, 39 patients with early CKD scheduled for renal biopsy and 41 healthy controls were enrolled. All participants underwent CEUS and SR-CEUS imaging during the same contrast-enhanced examination. Quantitative SR-CEUS parameters, including vessel density (VD), mean velocity (MV) and perfusion index (PI), as well as conventional CEUS time-intensity curve parameters, were analyzed. Logistic regression models were constructed based on variables with significant between-group differences, and diagnostic performance was assessed using receiver operating characteristic analysis. DeLong tests were used to compare model performance. Correlations between imaging-derived parameters and renal functional biomarkers across all participants, as well as histopathological interstitial fibrosis and tubular atrophy scores in the CKD group, were evaluated. RESULTS:Compared with healthy controls, patients with early CKD showed significantly lower VD, MV and PI (all p < 0.001). Among CEUS-derived time-intensity curve parameters, peak enhancement and wash-in perfusion were also significantly reduced (both p < 0.001), whereas time-related parameters showed no significant between-group differences. Among individual variables, VD yielded the best diagnostic performance (area under the receiver operating characteristic curve [AUC] = 0.86; 95% confidence interval [CI]: 0.74-0.95). The SR-CEUS model (VD + MV + PI) achieved the highest diagnostic accuracy (AUC = 0.94; 95% CI: 0.88-0.99), significantly outperforming the CEUS model (AUC = 0.82, p = 0.030) and eGFR (AUC = 0.75, p = 0.003). Across all participants, VD showed the strongest positive correlation with eGFR (r = 0.58, p < 0.001). In the CKD group, VD (r = -0.61), MV (r = -0.44), peak enhancement (r = -0.44) and wash-in perfusion (r = -0.36) were significantly negatively correlated with interstitial fibrosis and tubular atrophy scores. CONCLUSION:SR-CEUS enables non-invasive visualization and quantitative assessment of renal cortical microvascular alterations in early CKD. Compared with conventional CEUS and serum biomarkers, SR-CEUS demonstrated superior diagnostic performance for detecting microvascular impairment in patients with preserved eGFR. These findings suggest that SR-CEUS may serve as a sensitive imaging biomarker for the early non-invasive assessment of CKD-related microvascular injury.

Background:Immunoglobulin A nephropathy (IgAN) has heterogeneous clinical and pathological manifestations. Renal biopsy is the invasive diagnostic standard, and crescents mark active glomerular injury. Quantitative contrast-enhanced ultrasound (CEUS) can characterize renal microvascular perfusion. We aimed to develop a non-invasive nomogram integrating quantitative CEUS parameters and routine serological biomarkers to diagnose IgAN and predict crescent formation. Methods:In this retrospective single-center study, 184 patients with chronic kidney disease (CKD) underwent CEUS within 7 days before renal biopsy. Patients were classified into IgAN (n=94) and non-IgAN (n=90) groups, and the IgAN cohort was further stratified by crescent status (n=47 in each subgroup). Quantitative time-intensity curve (TIC) parameters were extracted from regions of interest (ROIs) using VueBox 7.0 software. Elliptical ROIs of approximately 4 mm × 3 mm were placed in the mid renal cortex and medulla closest to the probe to generate cortical and medullary TICs. Cortex-to-medulla normalization was applied. The analyzed parameters included peak enhancement (PE), wash-in rate (WiR), and wash-in perfusion index (WiP). Logistic regression models were constructed for IgAN diagnosis, crescent prediction in IgAN, and crescent prediction in the entire cohort. Receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA) were used to evaluate discrimination, calibration, and clinical net benefit. DeLong tests compared the combined model (CEUS parameters plus serological indicators), clinical-only model, and CEUS-only model to assess the incremental diagnostic value of CEUS. Subgroup analysis based on estimated glomerular filtration rate (eGFR) was performed for crescent prediction in IgAN to examine the influence of renal function on model performance. Results:IgAN patients showed significantly altered renal perfusion characteristics compared with non-IgAN patients. The combined model incorporating age, eGFR, urine albumin-to-creatinine ratio (UACR), and WiR achieved moderate discrimination for IgAN diagnosis, with an area under the curve (AUC) of 0.78 [95% confidence interval (CI): 0.72-0.85]. For crescent prediction in IgAN, the combined model based on age and WiP achieved an AUC of 0.74 (95% CI: 0.64-0.85). For crescent prediction in the entire cohort, the combined model based on age and rise time (RT) showed an AUC of 0.72 (95% CI: 0.64-0.81). DeLong tests showed that the combined models outperformed the corresponding clinical-only and CEUS-only models (all P<0.05), supporting the incremental but moderate diagnostic value of CEUS. Subgroup analysis showed stronger predictive performance in patients with preserved renal function. Conclusions:CEUS-derived microvascular information, when combined with clinical biomarkers, provides incremental diagnostic information and may assist pre-biopsy risk stratification and crescent formation prediction in CKD. These findings should be interpreted as exploratory and require external validation before routine clinical implementation.

Alterations in cerebral microcirculation during revascularization surgery may be associated with neurological outcomes, but methods for assessing subcortical microvasculature remain limited. Ultrasound localization microscopy (ULM) enables microvascular imaging beyond the conventional diffraction limit. We investigated its feasibility during surgery and examined whether measured microvascular changes are associated with clinical outcomes. In this prospective observational study, we analyzed 37 patients with Moyamoya disease who underwent cerebral revascularization at three centers. ULM images were reconstructed from contrast-enhanced ultrasound data acquired before and after bypass surgery. Vessel density, velocity, diameter, tortuosity and length were quantified. Associations between perioperative changes in ULM parameters and complications or neurological outcomes were assessed using multivariable logistic regression, correlation analysis, receiver operating characteristic analysis, and generalized estimating equations. Here, we show that intraoperative ULM enables visualization and quantitative assessment of subcortical cerebral microvasculature. Greater increases in vessel density and greater reductions in vessel velocity are independently associated with perioperative complications after adjustment for clinical covariates (adjusted odds ratios, 1.24 and 0.58, respectively). Both parameters show good discrimination for perioperative complications, with optimism-corrected areas under the curve of 0.93 and 0.90, respectively. Greater reductions in vessel velocity are also associated with poorer functional outcomes over 6-24 months of follow-up (adjusted odds ratio, 0.966). Intraoperative ULM enables quantitative assessment of subcortical cerebral microcirculation and may provide clinically relevant information for perioperative risk stratification and long-term prognostic assessment. Moyamoya disease causes narrowing of the arteries that supply the brain, reducing blood flow and increasing stroke risk. Revascularization surgery can improve circulation, but doctors have limited ways to examine how small brain vessels respond during the operation. In this multicenter study, we used ultrasound localization microscopy, which tracks contrast bubbles in the bloodstream, to map vessels beneath the brain surface in 37 adults before and after bypass surgery. Larger increases in vessel density and greater decreases in vessel velocity were associated with perioperative complications. Changes in vessel velocity were also related to neurological outcomes during follow-up. This ultrasound method may therefore provide useful information about brain microcirculation and help identify patients who may benefit from closer postoperative monitoring. Zhong, Wang, Zhang, Yan, Xu et al. use intraoperative ultrasound localization microscopy to quantify subcortical microvascular changes during cerebral revascularization in Moyamoya disease. Changes in vessel density and velocity identify people at higher perioperative risk, while velocity changes are associated with long-term functional outcomes.

Glioblastoma multiforme (GBM) remains an aggressive, treatment-refractory central nervous system malignancy, largely due to its invasiveness, vascular heterogeneity, and the blood-brain barrier (BBB). Conventional therapies suffer from poor specificity and insufficient intratumoral drug delivery. To address these challenges, we developed a tumor microenvironment (TME) and ultrasound (US) responsive biomimetic nanoreactor (CuTPD@M NPs) integrating bioorthogonal catalytic therapy (BCT) and sonodynamic therapy (SDT) for precision-targeted GBM treatment. The nanoplatform comprises a Cu(II)-coordinated porphyrinic metal-organic framework encapsulating azide and alkyne-functionalized prodrugs, cloaked with homologous GBM cell membranes, which enhances BBB penetration and homotypic tumor targeting. In response to elevated intracellular glutathione levels, Cu(II) is reduced to Cu(I), triggering an azide-alkyne cycloaddition to generate a combretastatin A-4 analog in situ, effectively disrupting tumor angiogenesis. Concurrently, US activates the porphyrin-based sonosensitizer, producing reactive oxygen species that induce mitochondrial dysfunction, glutathione peroxidase 4 downregulation, and ferroptosis. In murine GBM models, CuTPD@M combined with US achieved significant tumor growth inhibition, reduced angiogenic markers, and prolonged median survival. This work suggests a potentially transformative nanotherapeutic approach that leverages TME- response to co-activate BCT and SDT within a biomimetic nanosystem. The CuTPD@M platform demonstrates potential as an advance in precision nanomedicine for the treatment of intracranial malignant tumors.

Background:This study evaluates the diagnostic efficacy of high-resolution ultrasound (US) in identifying focal swelling and focal narrowing of the anterior interosseous nerve (AIN) fascicles within the median nerve around the elbow in spontaneous anterior interosseous neuropathy. Methods:This retrospective STROBE (Strengthing the Reporting of Observational Studies in Epidemiology)-compliant study evaluated 33 patients with spontaneous anterior interosseous neuropathy who underwent preoperative high-resolution US and surgery. Two radiologists jointly assessed the lesions, with intraoperative findings as the reference standard for calculating US diagnostic metrics. Intraclass correlation coefficients (ICCs) were analyzed for interrater reliability. Results:Surgery confirmed 4 focal swellings and 56 focal narrowings of the AIN fascicles within the median nerve around the elbow. Preoperative US correctly identified all swellings and 45 narrowings, yielding a sensitivity and overall accuracy of 81.7%, and a positive predictive value of 100%. Specificity and negative predictive value were not calculable due to zero false positives. The interrater reliability was good to excellent (intraclass correlation coefficient = 0.856). US also detected denervated muscle changes (n = 5), concomitant median nerve trunk compression (n = 2), and concomitant posterior interosseous nerve narrowing (n = 2). Conclusions:High-resolution US is a sensitive, reliable, and noninvasive modality for identifying focal swelling and focal narrowing of the AIN fascicles within the median nerve around the elbow in spontaneous anterior interosseous neuropathy. Its precision in lesion localization supports its utility as a first-line imaging technique for clinical evaluation and surgical planning.

Metabolic dysfunction-associated steatotic liver disease（MASLD）has become one of the leading causes of chronic liver diseases in China and even globally. Early non-invasive diagnosis and grading MASLD，along with timely intervention，are crucial for assessing the disease condition and slowing the disease progression. In recent years，non-invasive techniques for assessing liver steatosis content based on ultrasound have attracted significant attention. The ultrasound derived fat fraction（UDFF）is an emerging quantitative technique for liver fat assessment using ultrasound，which calculates the percentage value of the fat content by analyzing the radiofrequency signals reflected from the liver tissue，thereby quantifying the degree of liver steatosis. UDFF shows significant potential in assessing liver steatosis. However，the current application guidelines based on this technology do not yet fully meet the clinical needs. To further standardize its clinical application，institutions such as the Ultrasound Medicine Branch of the Chinese Medical Association，the National Clinical Research Center for Aging and Medicine，and the Institute of Ultrasound Medicine and Engineering of Fudan University，together with multidisciplinary experts ultrasound，endocrinology，radiology，and hepatology from across China have formulated this "Chinese expert consensus on the use of ultrasound derived fat fraction in the assessment of metabolic dysfunction-associated steatotic liver disease（2025 edition）" based on the lastest clinical application advances of the UDFF technology. This consensus standardizes the clinical scope of application and scenarios of this technology. It optimizes the technical operation process in various aspects，including pre- examination preparation，the operation process，quality control，the number of measurements，result presentation format，and influencing factors as well. Meanwhile，it summarizes current clinical research results，aiming to ensure the standardization of the UDFF technology in clinical applications.

Domain generalization poses a central challenge in medical ultrasound imaging: clinical centers differ substantially in equipment manufacturers, scanning protocols, and patient demographics, yet diagnostic models must remain robust across these variations. Existing style-based augmentation methods focus on perturbation strategies during training while overlooking the inherent discrepancy between training and test distributions, leading to poor generalization when deployed at previously unseen centers. To address this limitation, we propose the Multi-Center Adaptive Cross-Domain Style Alignment (MACS) framework, which bridges the train-test distribution gap through domain-specific style prototype-guided test-time adaptive normalization. MACS comprises four synergistic modules: 1) Domain-wise Correlation-Aware Style Augmentation (CASA) independently models the channel covariance structure for each source domain and generates correlation-aware style perturbations along principal variation directions; 2) Domain Style Prototype Learning (DSPL) accumulates representative style statistics for each source domain via exponential moving average, forming stable alignment anchors; 3) Style-based Domain Discriminator (SDD) identifies domain membership from channel-wise means and variances of feature maps; 4) Test-Time Adaptive Style Alignment (TTASA) fuses source domain style prototypes through discriminator-guided weighting, constructing an adaptive normalization target for each test sample. We validate MACS on an international multi-center elastography ultrasound dataset comprising 1,937 chronic liver disease patients from 17 clinical centers across China, Japan, and Europe. Extensive experiments show that MACS consistently outperforms existing domain generalization methods across diverse cross-domain evaluation settings, achieving superior accuracy and stability. Ablation studies further confirm the synergistic contributions of each module. As a plug-and-play solution with negligible computational overhead, MACS supports real-time inference and is well suited for practical deployment in multi-center clinical diagnosis.

Purpose To evaluate the clinical applicability of the US dual-distillation model (USDist) through comparative analysis with state-of-the-art models, ablation analysis of dual-distillation components, and assessment on portable US devices. Materials and Methods This retrospective multicenter study evaluated USDist using US video datasets collected from 16 medical centers (August 2016-December 2024) and two independent public datasets. The model integrates spatiotemporal dual-distillation and dynamic-static feature fusion to transfer feature representations from video and image foundation models. Diagnostic performance was assessed using receiver operating characteristic (ROC) analysis with DeLong testing and Holm correction, along with evaluation of computational efficiency and qualitative feature visualization. Results A total of 5033 patients were analyzed (mean age ± SD, 49 years ± 12; 5031 female). USDist achieved a mean area under the ROC curve (AUC) of 0.95 (95% CI: 0.93, 0.97) for breast cancer diagnosis across datasets. In the main cohort, USDist outperformed foundation models while using 98.3% fewer parameters. Across multicenter datasets, diagnostic performance was similar to that of foundation models (all P < .05). On a portable US device, USDist maintained an AUC of 0.92 (95% CI: 0.86, 0.95) with 4.1% of the computational cost of full-parameter fine-tuning. Conclusion USDist demonstrated high diagnostic performance for automated breast cancer diagnosis, with substantial parameter reduction compared with foundation models, and maintained performance across multicenter and portable US settings. Keywords: Ultrasound, Breast Cancer, Dual Distillation Model Supplemental material is available for this article. © RSNA, 2026 See also commentary by Whitman and Cohen in this issue.

OBJECTIVE:Isocitrate dehydrogenase (IDH) status is a critical biomarker for guiding glioblastoma treatment. This study aims to evaluate the effectiveness of a nomogram derived from histogram analysis of ultrafast ultrasound localization microscopy (ULM) for predicting IDH status. METHODS:Thirty-seven in situ glioblastoma rat models were established. Following craniotomy, microvascular morphology and hemodynamics were quantified using ultrafast ULM, and tumor regions were manually delineated via multimodal magnetic resonance imaging fusion registration. Tumors were classified as IDH-mutant or IDH-wildtype based on immunohistochemistry. A novel analytical approach based on an intensity histogram of the localization-density map was proposed to capture microvascular features within tumor regions. Univariate and multivariate logistic regression analyses were conducted to construct the nomogram, which was evaluated using calibration curves, decision curves, and receiver operating characteristic analysis. RESULTS:No significant differences were found in conventional ULM parameters between the IDH-mutant and IDH-wildtype groups (p = 0.162-0.915). Significant intergroup differences were observed in histogram features, including mean (p = 0.047), standard deviation (p = 0.004), skewness (p = 0.027), coefficient of variation (p = 0.011), and solidity (p = 0.007). Multivariate analysis identified minimum (p = 0.009), width (p = 0.025), skewness (p = 0.010), and coefficient of variation (p = 0.001) as independent predictors of IDH status. The model demonstrated strong predictive performance, with an area under the curve of 0.936 and accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of 0.876, 0.667, 0.934, 0.700, and 0.910, respectively. The Hosmer-Lemeshow test (χ² = 5.254, p = 0.730) confirmed a good fit between predicted and observed IDH status. Decision curve analysis confirmed the nomogram's clinical utility across mutation incidences ranging from 6% to 91%, indicating strong generalizability. CONCLUSION:ULM-derived histogram parameters provide a reliable method for predicting IDH status.

ObjectivesTo explore pathology and ultrasound features of breast cancers with different risk categories. To establish and validate a nomogram primarily based on grayscale ultrasound features for non-invasive preoperative prediction of high-risk breast cancers and for rapid individual risk assessment and clinical decision making.MethodsA total of 685 breast malignant lesions were enrolled in this study. All lesions were classified according to the St. Gallen risk categories criteria. The pathology and ultrasound features were compared among different risk groups. A multifactorial Logistic model and a nomogram primarily based on grayscale ultrasound were established. Then prediction ability was evaluated.ResultsIn training cohort, the ultrasound features with significant differences were selected again through Lasso regression. Then, age, maximum diameter in ultrasound, posterior echo attenuation, spiculate margin and suspicious axillary lymph nodes were selected to establish the prediction model and nomogram. The areas under the curve in training cohort and internal test cohort were 0.833 and 0.827. Diagnostic sensitivity, specificity, accuracy, positive likelihood ratio and negative likelihood ratio were 75.6%, 76.6%, 76.4%, 41.4% and 93.5%, respectively.ConclusionsBreast cancers with different risk categories exhibit distinct pathology and ultrasound features. The prediction model and nomogram have good and stable diagnostic efficiency.

OBJECTIVES:This study sought to design and verify a nomogram that utilizes ultrasonographic and clinical indicators to differentiate between intrahepatic cholangiocarcinoma (ICC) and hepatocellular carcinoma (HCC). METHODS:From November 2022 to September 2024, 136 patients with confirmed ICC or HCC were enrolled and randomly assigned to training and validation groups in a 7:3 ratio. Preoperative B-mode ultrasound, contrast-enhanced ultrasound, two-dimensional shear wave elastography features, and clinical indicators were retrieved and compared. Least Absolute Shrinkage and Selection Operator regression and multivariate logistic regression analysis were used to identify independent factors and develop a predictive nomogram. The model's evaluation focused on discrimination, calibration, and clinical utility. RESULTS:Significant predictive factors for ICC include a history of hepatitis, levels of alpha-fetoprotein and carbohydrate antigen 19-9, rim-like arterial phase hyperenhancement, and the stiffness ratio between the lesion and liver parenchyma. With AUC values of 0.987 (95% CI: 0.969, 1.000) for the training set and 0.926 (95% CI: 0.813, 1.000) for the validation set, the nomogram exhibited strong differentiation capabilities between the two entities. CONCLUSIONS:The nomogram combining multimodal indicators achieved high AUC values in both the validation and test sets (AUC = 0.926-0.987), demonstrating robust diagnostic accuracy for distinguishing ICC from HCC. This tool could aid in clinical decision-making for these challenging diagnoses.

Semantic segmentation of ultrasound (US) images with deep learning has played a crucial role in computer-aided disease screening, diagnosis and prognosis. However, due to the scarcity of US images and small field of view, resulting segmentation models are tailored for a specific single organ and may lack robustness, overlooking correlations among anatomical structures of multiple organs. To address these challenges, we propose the Multi-Organ FOundation (MOFO) model for universal US image segmentation. The MOFO is optimized jointly from multiple organs across various anatomical regions to overcome the data scarcity and explore correlations between multiple organs. The MOFO extracts organ-invariant representations from US images. Simultaneously, the task prompt is employed to refine organ-specific representations for segmentation predictions. Moreover, the anatomical prior is incorporated to enhance the consistency of the anatomical structures. A multi-organ US database with segmentation labels, comprising 7039 images from 10 organs across various regions of the human body, has been established to develop and evaluate our model. Results demonstrate that the MOFO outperforms single-organ methods in terms of the Dice coefficient, 95% Hausdorff distance and average symmetric surface distance with statistically sufficient margins. Our experiments in multi-organ universal segmentation for US images serve as a pioneering exploration of improving segmentation performance by leveraging semantic and anatomical relationships within US images of multiple organs.

We studied the microvascular structure and function of in situ glioblastoma using ultrasound localization microscopy (ULM). The in vivo study was conducted via craniotomy in six Sprague–Dawley rats. Capillary pattern, capillary hemodynamics, and functional quantitative parameters were compared among tumor core, invasive zone, and normal brain tissue with ex vivo micro-computed tomography (micro-CT) and scanning electron microscopy. Correlations between quantitative parameters and histopathological vascular density (VD-H), proliferation index, and histopathological vascular maturity index (VMI-H) were evaluated. Kruskal–Wallis H, ANOVA, Mann–Whitney U, Pearson, and Spearman correlation statistics were used. Compared to the tumor core, the invasive zone exhibited higher microvascularity structural disorder and complexity, increased hemodynamic heterogeneity, higher local blood flow perfusion (p ≤ 0.033), and slightly lower average flow velocity (p = 0.873). Significant differences were observed between the invasive zone and normal brain tissue across all parameters (p ≤ 0.001). ULM demonstrated higher microstructural resolution compared to micro-CT and a nonsignificant difference compared to scanning electron microscopy. The invasive zone vascular density correlated with VD-H (r = 0.781, p < 0.001). Vessel diameter (r = 0.960, p < 0.001), curvature (r = 0.438, p = 0.047), blood flow velocity (r = 0.487, p = 0.025), and blood flow volume (r = 0.858, p < 0.001) correlated with proliferation index. Vascular density (r = -0.444, p = 0.044) and fractal dimension (r = -0.933, p < 0.001) correlated with VMI-H. ULM provided high-resolution, noninvasive imaging of glioblastoma microvascularity, offering insights into structural/functional abnormalities. ULM technology based on ultrafast ultrasound can accurately quantify the microvessels of glioblastoma, providing a new method for evaluating the effectiveness of antiangiogenic therapy and visualizing disease progression. This method may facilitate early therapeutic assessment.