ObjectiveDiscriminating atypical hepatocellular carcinoma (HCC) from other malignancies in liver nodules classified as Liver Imaging Reporting and Data System category M (LR-M) remains a significant diagnostic challenge on conventional ultrasound examination. The LR-M category, originally intended to capture non-HCC malignancies, paradoxically contains up to 63% of atypical HCCs that deviate from classic enhancement patterns, leading to potential misdiagnosis and suboptimal treatment planning. While deep learning has shown promise in HCC diagnosis, most existing models rely exclusively on single-modality ultrasound, overlooking the diagnostic benefits of integrating complementary information from multiple imaging sources. To address this gap, we propose a novel attention-weighted tri-modal ultrasound network (TUS-Net) that integrates contrast-enhanced ultrasound (CEUS), B-mode ultrasound (BUS), and time-intensity curves (TICs) to improve diagnostic accuracy for these clinically challenging lesions. MethodsOur framework incorporates a three-dimensional convolutional neural network (C3D) backbone to extract spatiotemporal features from CEUS videos, capturing dynamic vascular patterns critical for lesion characterization. To effectively fuse complementary modalities, we introduce a dual-channel feature fusion module (DCFFM) that adaptively combines features from CEUS and BUS through channel-wise attention mechanisms, allowing the model to dynamically weigh the contribution of each modality based on diagnostic relevance. Additionally, we propose a temporal intensity feature fusion module (TIFFM) that leverages quantitative hemodynamic information from TICs to guide the model’s attention toward diagnostically critical temporal phases, such as arterial wash-in and portal venous washout. The model is further enhanced by automated lesion localization using YOLOX and class activation mapping for interpretability, ensuring that predictions align with clinically meaningful imaging features. ResultsEvaluated on a tri-modal ultrasound dataset comprising 161 patients with pathologically confirmed LR-M nodules (131 atypical HCC and 30 non-HCC malignancies), our model achieved an accuracy of 86.83%, a sensitivity of 92.50%, a specificity of 75.50%, and an AUC of 89.32% in screening atypical HCC. Compared to single-modality baselines, TUS-Net demonstrated superior specificity, a clinically critical metric given the higher risk associated with misclassifying non-HCC malignancies. Ablation studies confirmed the contribution of each module, with the full model outperforming both standard C3D and 3D ResNet backbones integrated with attention mechanisms. A reader study involving junior and senior radiologists further validated the clinical utility of AI assistance, showing consistent improvements in specificity and inter-reader consistency, particularly for less experienced clinicians. ConclusionThese results surpass existing benchmark models and demonstrate the potential of our approach to enhance diagnostic precision in clinically specific cases. By intelligently fusing multi-modal ultrasound data with attention-guided mechanisms, TUS-Net offers a reliable and interpretable tool that holds promise for improving the non-invasive diagnosis of atypical HCC in challenging LR-M liver nodules.

ObjectiveDiscriminating atypical hepatocellular carcinoma (HCC) from other malignancies in liver nodules classified as Liver Imaging Reporting and Data System category M (LR-M) remains a significant diagnostic challenge on conventional ultrasound examination. The LR-M category, originally intended to capture non-HCC malignancies, paradoxically contains up to 63% of atypical HCCs that deviate from classic enhancement patterns, leading to potential misdiagnosis and suboptimal treatment planning. While deep learning has shown promise in HCC diagnosis, most existing models rely exclusively on single-modality ultrasound, overlooking the diagnostic benefits of integrating complementary information from multiple imaging sources. To address this gap, we propose a novel attention-weighted tri-modal ultrasound network (TUS-Net) that integrates contrast-enhanced ultrasound (CEUS), B-mode ultrasound (BUS), and time-intensity curves (TICs) to improve diagnostic accuracy for these clinically challenging lesions. MethodsOur framework incorporates a three-dimensional convolutional neural network (C3D) backbone to extract spatiotemporal features from CEUS videos, capturing dynamic vascular patterns critical for lesion characterization. To effectively fuse complementary modalities, we introduce a dual-channel feature fusion module (DCFFM) that adaptively combines features from CEUS and BUS through channel-wise attention mechanisms, allowing the model to dynamically weigh the contribution of each modality based on diagnostic relevance. Additionally, we propose a temporal intensity feature fusion module (TIFFM) that leverages quantitative hemodynamic information from TICs to guide the model’s attention toward diagnostically critical temporal phases, such as arterial wash-in and portal venous washout. The model is further enhanced by automated lesion localization using YOLOX and class activation mapping for interpretability, ensuring that predictions align with clinically meaningful imaging features. ResultsEvaluated on a tri-modal ultrasound dataset comprising 161 patients with pathologically confirmed LR-M nodules (131 atypical HCC and 30 non-HCC malignancies), our model achieved an accuracy of 86.83%, a sensitivity of 92.50%, a specificity of 75.50%, and an AUC of 89.32% in screening atypical HCC. Compared to single-modality baselines, TUS-Net demonstrated superior specificity, a clinically critical metric given the higher risk associated with misclassifying non-HCC malignancies. Ablation studies confirmed the contribution of each module, with the full model outperforming both standard C3D and 3D ResNet backbones integrated with attention mechanisms. A reader study involving junior and senior radiologists further validated the clinical utility of AI assistance, showing consistent improvements in specificity and inter-reader consistency, particularly for less experienced clinicians. ConclusionThese results surpass existing benchmark models and demonstrate the potential of our approach to enhance diagnostic precision in clinically specific cases. By intelligently fusing multi-modal ultrasound data with attention-guided mechanisms, TUS-Net offers a reliable and interpretable tool that holds promise for improving the non-invasive diagnosis of atypical HCC in challenging LR-M liver nodules.

Despite significant advances in the field of critical care medicine over the past three decades, veno-arterial extracorporeal membrane oxygenation (V-A ECMO) remains the primary temporary mechanical circulatory support modality for patients with acute severe circulatory failure. With the accumulation of clinical experience and the increasing maturity of operational techniques in V-A ECMO, its technical management—particularly hemodynamic management—has become a key factor influencing patient outcomes. To further improve patient survival, the Chinese Critical Care Ultrasound Study Group, in collaboration with the Hemodynamic Therapy of Critical Care Collaborative Group and the Critical Care Medicine Branch of the China International Exchange and Promotive Association for Medical and Health Care, organized experts in critical care medicine to develop the Consensus on Hemodynamic Management in Adult Veno-Arterial Extracorporeal Membrane Oxygenation (2026 Edition). Based on years of clinical experience and the latest evidence-based medical evidence, this consensus formulates 15 systematic recommendations covering the entire process of V-A ECMO, including initiation, management during support, and weaning, aiming to promote standardized application of hemodynamic management in V-A ECMO.

Objectives: . Idiopathic sudden sensorineural hearing loss (ISSNHL) is a significant cause of hearing loss. Intratympanic steroid injection (ITSI) is commonly used as an initial or salvage treatment; however, the lack of a standardized treatment protocol has resulted in variability in clinical practice. In addition, no efficient prediction model currently exists to support personalized management. Therefore, this study aimed to develop tailored management strategies for ISSNHL using a machine-learning model. Methods: . This retrospective multicenter cohort study was conducted between January 2015 and December 2020, with data analysis performed between January 2021 and March 2024. Patients were selected based on the International Classification of Diseases, 10th Revision criteria for ISSNHL, along with relevant medication and procedure codes. Patients with pure-tone audiogram results not meeting ISSNHL criteria, better initial hearing in the affected ear, an identifiable etiology, no post-treatment audiogram, or delayed treatment (>6 weeks) were excluded. We included 770 patients diagnosed with ISSNHL who received ITSI. The primary outcome was the area under the receiver operating characteristic curve for prediction performance. Recovery status was determined using the last pure-tone audiogram. Modeling was conducted on the Quanta for Medical Care AI platform using five machine-learning algorithms and a nested cross-validation framework, in which feature selection and hyperparameter tuning were performed in the inner folds and model performance was evaluated in the outer folds. Results: . A random forest classifier outperformed the other models in predicting hearing outcomes, achieving an area under the receiver operating characteristic curve of 0.788. Time to ITSI was the most influential treatment-related factor, with ITSI administered within 10 days of hearing loss being associated with better outcomes. This model can be used to provide personalized prognostic estimates under different treatment protocols. Conclusion . The machine-learning-based prediction model facilitates personalized treatment strategies and timely treatment adjustments for ISSNHL, thereby optimizing the likelihood of complete recovery.

Objective: To develop machine learning identification models for the relatively high risk of social-emotional problems among Chinese school-aged children, and compare internal validation performances across different algorithms, so as to provide reference for preliminary identification of relatively high risk in school-aged children s social-emotional problems. Methods: Based on the 2022 proxy-reported data for children from the China Family Panel Studies (CFPS), a total of 3 575 children aged 6-15 years from 2 709 families who completed the survey between May and December 2022 were included. Relatively high risk of social-emotional problems was defined using a composite score constructed from 12 social emotional related items in the CFPS 2022 general questionnaire. Data were split approximately 7∶3 (modeling group: 2 503, validation group: 1 072) by household identifier (ID) to ensure all children from the same family were assigned to the same dataset. Candidate variable selection, missing value imputation parameter fitting, least absolute shrinkage and selection operator (LASSO) feature screening, model training, and hyperparameter tuning were performed within the development set, while internal validation was conducted in the validation set. Weighted AUC, Brier score, sensitivity, specificity, calibration curves, and decision curves were compared across Logistic regression, random forest (RF), gradient boosting decision trees (GBDT), support vector machines (SVM), and extreme gradient boosting (XGBoost). The Shapley additive explanations (SHAP) method was used for exploratory interpretation of the RF model, which achieved the highest AUC value in the validation set. Results: A total of 941 (26.3%) were classified as having relatively high risk of social emotional problems. A total of 659 and 282 high risk cases were in the modeling group and validation group, respectively, with no cross set split of households. LASSO retained 10 features. In the validation set, the RF model achieved the highest AUC of 0.672 (95% CI =0.624-0.717) and a Brier score of 0.184. The Logistic regression model achieved a validation AUC of 0.670 (95% CI =0.624-0.716). The incremental AUC gain of RF over Logistic regression was 0.002 (family cluster paired bootstrap 95% CI =-0.016 to 0.020, P =0.856). Following household ID splitting, all models showed moderate internal validation performance. Although RF achieved the highest numerical value, its incremental gain over Logistic regression was limited. Conclusion The model can serve as a reference for preliminary identification research on the relatively high risk of social emotional problems in school aged children, though its stability and applicability require further validation in external samples.

Objective To investigate the preventive and therapeutic effects of ethanol extracts derived from three sources of traditional Chinese medicine: Stellera chamaejasme L., Euphorbia fischeriana Steud. and Euphorbia kansuensis Prokh., on imiquimod （IMQ）-induced psoriasis in mice. Methods Thirty-six male BALB/c mice were randomly divided into the following 6 groups with 6 mice in each group: blank control, model, Stellera chamaejasme, Euphorbia fischeriana, Euphorbia kansuensis, and calcipotriol. Psoriasis area and severity index（PASI） scores were used to record the changes of skin lesions in every group; Hematoxylin-eosin（HE） staining was used to observe the pathological morphology of skin and measure the thickness of the epidermis. Immunohistochemistry was used to detect the expression of nuclear antigen Ki67 in the skin tissues of mice. Results Compared with the model group, the three kinds of ethanol extracts could reduce the PASI score, inhibit epidermal thickening, and decrease expression of Ki67 in the psoriasis mice. Among them, the therapeutic effect of Stellera chamaejasme was the most significant and it was better than the commonly used topical drug calcipotriol. Conclusion The ethanol extract of Stellera chamaejasme had good anti-psoriatic activity, which could inhibit the abnormal proliferation of keratinocytes, reduce the expression of Ki67, and significantly improve psoriasis-like skin lesions.

AIM To improve the quality standard for Dahuang Sodium Bicarbonate Tablets.METHODS TLC was adopted in the qualitative identification of rhein,emodin,chrysophanol,aloe-emodin,physcion and rhapontin,after which UPLC fingerprints were established,content determination of aloe emodin-8-O-β-D-glucoside,rhein-8-O-β-D-glucoside,emodin-8-O-β-D-glucoside,chrysophanol-8-O-β-D-glucoside,physcion-8-O-β-D-glucoside,aloe-emodin,rhein,emodin,chrysophanol,physcion were performed.RESULTS The clear TLC plots demonstrated good specificity.There were 20 common peaks in the fingerprints for 8 batches of samples with the similarities of more than 0.535.Ten constituents showed good linear relationships within their own ranges(r≥0.997 0),whose average recoveries were 98.7％-104.1％with the RSDs of 0.59％-2.13％,whose content ranges were 0.274 1-2.005 9,0.547 8-5.192 2,1.209 2-3.936 3,1.039 3-3.330 2,0.307 0-1.248 1,0.118 2-0.674 7,0.115 1-0.970 3,0.224 5-1.789 3,0.881 4-4.882 6,0.801 4-5.099 1 mg/g,respectively.CONCLUSION This simple and stable method can effectively control the quality of Rhei Radix et Rhizoma in Dahuang Sodium Bicarbonate Tablets.

AIM To investigate the improvement effect of Qinhuo Formula on obese mice induced by high-fat diet.METHODS Obese mice model was induced by feeding with high-fat diet.The mice successfully established were randomly divided into model group,orlistat group(15.6 mg/kg)and low and high dose groups of Qinhuo Formula(13.5 and 22.5 g/kg),and normal mice were taken as control group,with 10 mice in each group,and the drugs were continuously intervened for 8 weeks.After the drug administration,the body weight,abdominal circumference,fasting blood glucose and glucose tolerance were measured,and area under the curve(AUC)of oral glucose tolerance test(OGTT),Lee's index and fat index were calculated.The serum levels of TC,TG,LDL-C and HDL-C were detected by blood biochemical analyzer.Serum levels of TNF-α,IL-1 β and IL-6 were detected by ELISA.HE staining was used to observe the white fat of epididymis and the pathological changes of liver tissue.RT-qPCR and Western blot were used to detect the mRNA and protein expressions of TLR4,MyD88 and NF-κB in liver tissue.RESULTS Compared with the model group,the body weight,abdominal circumference and Lee's index of mice in each dose group of Qinhuo Formula decreased(P＜0.01);OGTT-AUC decreased(P＜0.01);visceral fat index decreased(P＜0.05,P＜0.01);the levels of serum TG,TC,IL-6 and IL-1 β decreased(P＜0.01);the diameter of epididymal adipocytes decreased,arranged tightly and regularly,the lipid droplet cavities in liver tissue decreased,and the morphology of hepatocytes recovered;the mRNA and protein expressions of TLR4,MyD88 and NF-κB decreased(P＜0.05,P＜0.01).CONCLUSION Qinhuo Formula can effectively improve glucose-lipid metabolism and reduce inflammation in obese mice,and its mechanism may be related to the inhibition of TLR4/MyD88/NF-κB signaling pathway.

The blue fluorescent carbon dots(TMCDs)and cyan fluorescent carbon dots(TMCDs-H2O)were synthesized fromm-phenylenediamine and tricarballylic acid through air-assisted melting polymerization and one-step hydrothermal method,respectively.Air purging could effectively inhibit the side reactions and reduce the derivative structures in the carbon dots product.The structure and morphology of these two materials were systematically characterized using liquid nuclear magnetic resonance spectroscopy(NMR),mass spectrometry(MS),and transmission electron microscopy.Compared to TMCDs-H2O((3.12±0.63)nm),TMCDs showed a smaller average particle size(approximately(1.85±0.02)nm).The NMR and MS analysis revealed that although the main structure of both types of carbon dots was similar,TMCDs exhibited a simpler structure with higher degree of polymerization.These results suggested that supramolecular interactions might introduce numerous small molecule derivatives into TMCDs-H2O particles,resulting in lower polymerization degree,multiple substructures,and larger particle size characteristics for this material.When employed as chemical sensors for metal ion detection,in the linear range of 1×10-5-5×10-4 mol/L,the detection limits of Fe3+by TMCDs and TMCDs-H2O were 3.3×10-6 mol/L and 3.8×10-6 mol/L,respectively.The experimental results demonstrated that the recoveries of CDs and inductively coupled plasma optical emission spectrometer(ICP-OES)were similarity,whereas TMCDs displayed a considerable relative standard deviation.This study demonstrated that endogenously derived structures in CDs could enhance the performance of metal ion sensing.

In this work,with tris(4-aminophenyl)amine(TAPA)and 1,3,5-tris(4-formylphenyl)benzene(TFPB)as monomers,an imine-type porous organic polymer,TAPA-TFPB,was synthesized using a simple method under the catalysis of acetic acid.The material TAPA-TFPB was used as solid-phase extraction adsorbent and combined with ultra-performance liquid chromatography/quadrupole time-of-flight-tandem mass spectrometry(UHPLC-QTOF-MS)to establish a detection method for four kinds of nitrofuran metabolites(NFMs)residues in meat samples.The parameters of the adsorbent dosage,the pH value and volume of sample,and the type and volume of washing and eluent solvents were optimized,respectively.Under the optimal extraction conditions,low detection limits(0.11-1.60 μg/kg)were achieved for four kinds of NFMs.At three different spiked levels,the intra-day and inter-day precisions(Relative standard deviations)were 2.8% -10.9% and 4.3% -16.2%,respectively,and the spiked recoveries were 72.0% -107.2%.The results showed that the method chould be used for efficient extraction and analysis of trace NFMs residues in meat samples,indicating that TAPA-TFPB was a kind of promising SPE adsorbent.