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.

Objective: To establish a prediction model for high-risk cardiovascular disease (CVD) among residents aged 35 to 75 years, so as to provide the basis for improving CVD prevention and control measures. Methods: Permanent residents aged 35 to 75 years were selected from Dongcheng District, Beijing Municipality using the stratified random sampling method from 2018 to 2023. Demographic information, lifestyle, waist circumference and blood biochemical indicators were collected through questionnaire surveys, physical examinations and laboratory tests. Influencing factors for high-risk CVD among residents aged 35 to 75 years were identified using a multivariable logistic regression model, and a prediction model for high-risk CVD was established. The predictive effect was evaluated using the receiver operating characteristic (ROC) curve. Results: A total of 6 968 individuals were surveyed, including 2 821 males (40.49%) and 4 147 females (59.51%), and had a mean age of (59.92±9.33) years. There were 1 155 high-risk CVD population, with a detection rate of 16.58%. Multivariable logistic regression analysis showed that gender, age, smoking, central obesity, systolic blood pressure, fasting blood glucose, triglyceride and low-density lipoprotein cholesterol were influencing factors for high-risk CVD among residents aged 35 to 75 years (all P<0.05). The area under the ROC curve of the established prediction model was 0.849 (95%CI: 0.834-0.863), with a sensitivity of 0.693 and a specificity of 0.863, indicating good discrimination. Conclusion The model constructed by eight factors including demographic characteristics, lifestyle and blood biochemical indicators has good predictive value for high-risk CVD among residents aged 35 to 75 years.

This study aims to explore the role and mechanism of berberine in promoting the expression of aquaporin 4(AQP4) in astrocytes by regulating the expression of miR-383-5p in HepG2 cell-derived exosomes under insulin resistance(IR). The IR-HepG2 cell model was established with 1×10~(-6) mol·L~(-1) insulin. With metformin as the positive control, the safe concentrations of berberine and metformin were screened by cell counting kit-8(CCK-8) and lactate dehydrogenase(LDH) leakage assays, and the effect of berberine on the IR of HepG2 cells was evaluated by glucose consumption. NanoSight was used to measure the particle size and concentration of exosomes secreted by HepG2 cells in each group. HepG2 cell-derived exosomes in each group were incubated with astrocytes for 24 h, and the protein and mRNA levels of AQP4 in HA1800 cells were determined by Western blot and qRT-PCR, respectively. qRT-PCR was performed to determine the expression of miR-383-5p in HepG2 cell-derived exosomes and HA1800 cells after co-incubation. Western blotting was employed to determine the expression levels of miRNAs and proteins associated with exosome production and release in HepG2 cells. The results showed that 10 μmol·L~(-1) berberine and 1 mmol·L~(-1) metformin significantly alleviated the IR of HepG2 cells and reduced the concentration of exosomes in HepG2 cells. The exosomes of HepG2 cells treated with berberine and metformin significantly up-regulated the protein and mRNA levels of AQP4 in HA1800 cells. The mRNA level of miR-383-5p in HepG2 cell exosomes and HA1800 cells co-incubated with berberine and metformin decreased significantly. The intervention with berberine and metformin significantly down-regulated the expression of proteins associated with the production of miRNAs(Dicer, Drosha) as well as the production(Alix, Vps4A) and release(Rab35, VAMP3) of exosomes in IR-HepG2 cells. In conclusion, berberine can promote the expression of AQP4 in astrocytes by inhibiting the production and release of miR-383-5p in HepG2-derived exosomes under IR.
Humans ; MicroRNAs/metabolism* ; Berberine/pharmacology* ; Hep G2 Cells ; Exosomes/genetics* ; Aquaporin 4/metabolism* ; Insulin Resistance ; Astrocytes/drug effects*

ObjectivePrimary liver cancer, predominantly hepatocellular carcinoma (HCC), is a significant global health issue, ranking as the sixth most diagnosed cancer and the third leading cause of cancer-related mortality. Accurate and early diagnosis of HCC is crucial for effective treatment, as HCC and non-HCC malignancies like intrahepatic cholangiocarcinoma (ICC) exhibit different prognoses and treatment responses. Traditional diagnostic methods, including liver biopsy and contrast-enhanced ultrasound (CEUS), face limitations in applicability and objectivity. The primary objective of this study was to develop an advanced, light-weighted classification network capable of distinguishing HCC from other non-HCC malignancies by leveraging the automatic analysis of brightness changes in CEUS images. The ultimate goal was to create a user-friendly and cost-efficient computer-aided diagnostic tool that could assist radiologists in making more accurate and efficient clinical decisions. MethodsThis retrospective study encompassed a total of 161 patients, comprising 131 diagnosed with HCC and 30 with non-HCC malignancies. To achieve accurate tumor detection, the YOLOX network was employed to identify the region of interest (ROI) on both B-mode ultrasound and CEUS images. A custom-developed algorithm was then utilized to extract brightness change curves from the tumor and adjacent liver parenchyma regions within the CEUS images. These curves provided critical data for the subsequent analysis and classification process. To analyze the extracted brightness change curves and classify the malignancies, we developed and compared several models. These included one-dimensional convolutional neural networks (1D-ResNet, 1D-ConvNeXt, and 1D-CNN), as well as traditional machine-learning methods such as support vector machine (SVM), ensemble learning (EL), k-nearest neighbor (KNN), and decision tree (DT). The diagnostic performance of each method in distinguishing HCC from non-HCC malignancies was rigorously evaluated using four key metrics: area under the receiver operating characteristic (AUC), accuracy (ACC), sensitivity (SE), and specificity (SP). ResultsThe evaluation of the machine-learning methods revealed AUC values of 0.70 for SVM, 0.56 for ensemble learning, 0.63 for KNN, and 0.72 for the decision tree. These results indicated moderate to fair performance in classifying the malignancies based on the brightness change curves. In contrast, the deep learning models demonstrated significantly higher AUCs, with 1D-ResNet achieving an AUC of 0.72, 1D-ConvNeXt reaching 0.82, and 1D-CNN obtaining the highest AUC of 0.84. Moreover, under the five-fold cross-validation scheme, the 1D-CNN model outperformed other models in both accuracy and specificity. Specifically, it achieved accuracy improvements of 3.8% to 10.0% and specificity enhancements of 6.6% to 43.3% over competing approaches. The superior performance of the 1D-CNN model highlighted its potential as a powerful tool for accurate classification. ConclusionThe 1D-CNN model proved to be the most effective in differentiating HCC from non-HCC malignancies, surpassing both traditional machine-learning methods and other deep learning models. This study successfully developed a user-friendly and cost-efficient computer-aided diagnostic solution that would significantly enhances radiologists’ diagnostic capabilities. By improving the accuracy and efficiency of clinical decision-making, this tool has the potential to positively impact patient care and outcomes. Future work may focus on further refining the model and exploring its integration with multimodal ultrasound data to maximize its accuracy and applicability.

OBJECTIVE: To explore the relationship between serum chloride levels and prognosis in patients with hepatic coma in the intensive care unit (ICU). METHODS: We analyzed 545 patients with hepatic coma in the ICU from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. Associations between serum chloride levels and 28-day and 1-year mortality rates were assessed using restricted cubic splines (RCSs), Kaplan-Meier (KM) curves, and Cox regression. Subgroup analyses, external validation, and mechanistic studies were also performed. RESULTS: A total of 545 patients were included in the study. RCS analysis revealed a U-shaped association between serum chloride levels and mortality in patients with hepatic coma. The KM curves indicated lower survival rates among patients with low chloride levels (< 103 mmol/L). Low chloride levels were independently linked to increased 28-day and 1-year all-cause mortality rates. In the multivariate models, the hazard ratio ( HR) for 28-day mortality in the low-chloride group was 1.424 (95% confidence interval [ CI]: 1.041-1.949), while the adjusted hazard ratio for 1-year mortality was 1.313 (95% CI: 1.026-1.679). Subgroup analyses and external validation supported these findings. Cytological experiments suggested that low chloride levels may activate the phosphorylation of the NF-κB signaling pathway, promote the expression of pro-inflammatory cytokines, and reduce neuronal cell viability. CONCLUSION Low serum chloride levels are independently associated with increased mortality in patients with hepatic coma.
Humans ; Male ; Female ; Middle Aged ; Intensive Care Units ; Prognosis ; Chlorides/blood* ; Aged ; Coma/blood* ; Adult

Objective To investigate the alterations of cerebral blood flow(CBF)in type 2 diabetic mellitus(T2DM) patients without cognitive impairment by using arterial spin labeling(ASL)technique.Methods A total of 23 T2DM patients without cognitive impairment and 23 healthy controls(HC)matched by age,sex,and education attainment were recruited.Their clinical data were collected,and neuropsychological tests and cerebral magnetic resonance imaging were performed.Then,the outcomes of clinical features,neuropsychological tests,and global and regional CBF were compared between the two groups.The significant regional zCBF(z-transformed relative CBF)values were extracted and correlated with clinical data and neuropsychological scores in T2DM patients,controlling age,sex,and education.Results No significant difference was found in whole brain CBF between the two groups(P=0.155),while significantly higher CBF was identified in the left superior temporal gyrus and left insula in the T2DM group(Gaussian random field correction,initial threshold P < 0.001,cluster level P < 0.05).No correlation was observed between the significant regional zCBF values and the clinical data or the neuropsychological scores in T2DM patients(all P>0.05).Conclusion Alterations in cerebral hemodynamics may precede cognitive function changes in T2DM,suggesting that the ASL technique is promising for early monitoring of cerebral hemodynamic changes associated with cognitive impairment in patients with T2DM.
Humans ; Diabetes Mellitus, Type 2/physiopathology* ; Cerebrovascular Circulation ; Middle Aged ; Male ; Female ; Magnetic Resonance Imaging ; Case-Control Studies ; Cognitive Dysfunction ; Neuropsychological Tests ; Aged

Objective To evaluate the diagnostic value of transrectal contrast-enhanced ultrasound (CEUS) for rectal cancer with intestinal stenosis caused by tumors. Methods Forty-nine patients with rectal cancer underwent transrectal CEUS and magnetic resonance imaging (MRI) before surgery.Intraoperative tumor localization and postoperative pathological results were taken as the gold standard for diagnosis.The differences in T stage,localization,and tumor length of rectal cancer were compared between the two methods. Results The total accuracy rates of transrectal CEUS and MRI in diagnosing T stage were 75.5% (36/49) and 67.3% (33/49),which had no significant difference (χ²=0.8,P=0.371).The total accuracy rates of transrectal CEUS and MRI in judging tumor localization were 79.5% (39/49) and 77.5% (38/49),which had no significant difference (χ²=0.061,P=0.806).The measurement results of tumor length in pathological examination had no significant difference from the transrectal CEUS results (t=1.42,P=0.162) but a significant difference from the MRI results (t=3.38,P=0.001).Furthermore,transrectal CEUS detected 8 (16.3%) cases of colonic polyps among the 49 patients,while MRI did not detect colon lesions. Conclusions Transrectal CEUS has good consistency with MRI in T staging and localization judgement of rectal cancer with intestinal stenosis,and this method can more accurately evaluate the tumor length and simultaneously evaluate whether there is a lesion in the entire colon at the proximal end of stenosis.It can be used as a supplementary examination before rectal cancer treatment in clinical practice.
Humans ; Rectal Neoplasms/complications* ; Male ; Middle Aged ; Female ; Aged ; Contrast Media ; Ultrasonography ; Adult ; Magnetic Resonance Imaging ; Constriction, Pathologic/diagnostic imaging* ; Aged, 80 and over ; Intestinal Obstruction/etiology*