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.

Background and aims:Early and accurate diagnosis of the coexistence of Wilson's disease(WD)and chronic hepatitis B(CHB)presents a significant challenge for clinicians.The objective of this study was to retrospectively analyse the characteristics of such patients to improve clinical practice and provide a reference for clinical management.Methods:From January 2011 to December 2022,35 patients with concurrent CHB and WD(CHB+WD group)were identified.A total of 127 patients with CHB(CHB group)and 168 patients with WD(WD group)were included in the control group between January 2016 and December 2021.Propensity score matching(PSM)was performed to balance the baseline values between groups.The Kaplan-Meier(K-M)survival analysis and log-rank test were performed to compare the prognoses.Results:In the cohort of 35 patients with concurrent CHB and WD,74.3％of patients(26 patients)faced a substantial delay of up to 10 years(range:0-40 years)in WD diagnosis following their CHB diagnosis.Twenty-three(65.7％)patients had cirrhosis at the time of WD diagnosis,and 26(74.3％)patients experienced liver failure.The levels of serum copper and uric acid were lower in patients in the CHB+WD group than in those in the CHB group.Patients in the CHB+WD group presented higher alanine transaminase and total bile acid levels compared to those in the WD group.K-M survival analysis indicated that patients with CHB and WD had poorer outcomes than those with CHB alone;however,the outcomes were similar to those of individuals with WD alone.The optimal cut-point of serum ceruloplasmin(CP)in identifying WD in CHB patients was 0.10 g/L before PSM and after PSM.Conclusions:The present study emphasizes the importance of clinicians being vigilant for concurrent CHB and WD diagnoses,as delays in WD diagnosis may adversely affect patient outcomes.CHB patients with serum CP below 0.10 g/L are highly recommended to screen for WD.

OBJECTIVES: To investigate the protective effects and underlying mechanisms of Gynostemma pentaphyllum (GP)ethanol extract on motor dysfunction in a mouse model of Parkinson's disease (PD). METHODS: Eighty C57BL/6 male mice were randomly divided into five groups: control group, model group, levodopa group (positive control group), low-dose GP group, and high-dose GP group, with 16 mice per group. The PD model was induced by injection of 6-hydroxydopamine into the substantia nigra pars reticulata of the mice. Two weeks after 6-hydroxydopamine, positive control group received intraperitoneal injection of levodopa 10 mg·kg^－1·d^－1, while low-dose GP and high-dose GP groups received GP extract 100 or 200 mg·kg^－1·d^－1 orally for three weeks. After a 3-week-treatment, the effects of GP on motor dysfunction in 6-hydroxydopamine-induced PD were assessed using open field and CatWalk gait tests, while the effects on muscle strength were evaluated by forelimb grip strength. Immunofluorescence staining was used to detect the number of tyrosine hydroxylase (TH) positive neurons. The levels of dopamine and serotonin in the midbrain were determined by enzyme-linked immunosorbent assay. In addition, Western blotting was performed to detect the expression of mitogen-activated protein kinase (MAPK) family proteins such as p-extracellular signal-regulated kinase (ERK)1/2, p-p38 and p-c-Jun N-terminal kinase (JNK)1/2, and mitochondrial apoptosis pathway proteins such as B-cell lymphoma (Bcl)-2, Bcl-2 associated X protein (Bax), and cleaved-cysteine aspartic acid specific protease (caspase)-3. RESULTS: Behavioral experiments showed that GP significantly improved the spontaneous activity and motor coordination of PD mice (P<0.05). The forelimb grip strength was also increased by GP treatment (P<0.05), compared to the PD model group. In addition, compared with the model group, the number of TH-positive neurons in substantia nigra pars reticulata region, the levels of dopamine and serotonin in midbrain and the expression of p-ERK1/2 were significantly increased by GP treatment (all P<0.05), whereas the expression of p-p38 and p-JNK1/2, the ratio of Bax/Bcl-2 and cleaved-caspase-3/caspase-3 were significantly decreased (all P<0.05). CONCLUSIONS The results indicate that GP might increase dopamine and serotonin levels in the midbrain and promote the survival of dopaminergic neurons in substantia nigra pars reticulata by regulating the expression of phosphorylation of MAPK family proteins and the expression of mitochondrial apoptosis-related proteins, thereby ameliorating motor deficits in PD mice.
Animals ; Mice ; Male ; Gynostemma/chemistry* ; Mice, Inbred C57BL ; Apoptosis/drug effects* ; Plant Extracts/therapeutic use* ; Parkinson Disease/metabolism* ; Disease Models, Animal ; Neurons/pathology*

Objective To compare imaging quality and semi-quantitative parameters of 18F-FDG whole-body images obtained with domestic NeuWise and Philips Ingenuity TF PET/CT scanners.Methods Thirty-four patients who underwent 18F-FDG whole-body scanning using NeuWise and Philips Ingenuity TF PET/CT systems respectively on the same day were enrolled.The imaging quality and semi-quantitative parameters of 2 kind images,also the mean standard uptake value(SUVmean)of normal tissue,the maximum standard uptake value(SUVmax),peak standard uptake value(SUVpeak),SUVmean of lesions,total lesion glycolysis(TLG)and metabolic tumor volume(MTV)were compared.Results No significant difference of imaging quality nor semi-quantitative parameters of lesions(all P＞0.05),while significant differences of SUVmean of aortic arch,liver,lumbar vertebra and spinal cord were found between 2 kind images(all P＜0.05).Strong correlations of SUVmax,SUVmean,MTV and TLG of lesions(r,=0.734-0.890,P＜0.001),and high correlation of SUVpeak(rs=0.919,P＜0.001)were found between 2 kind images.The consistency of SUVmax,SUVpeak,SUVmean,TLG and MTV at the lesion site between 2 kind images were very good to extremely good(ICC=0.891-0.986,all P＜0.001),and the differences of all above semi-quantitative parameters were within 95％confidence interval.Conclusion Imaging quality of 18F-FDG whole-body images obtained with domestic NeuWise and Philips Ingenuity TF PET/CT scanners could meet the requirements of clinical diagnosis and treatment,and semi-quantitative parameters obtained based on both images had good consistencies.

Objective To compare imaging quality and semi-quantitative parameters of 18F-FDG whole-body images obtained with domestic NeuWise and Philips Ingenuity TF PET/CT scanners.Methods Thirty-four patients who underwent 18F-FDG whole-body scanning using NeuWise and Philips Ingenuity TF PET/CT systems respectively on the same day were enrolled.The imaging quality and semi-quantitative parameters of 2 kind images,also the mean standard uptake value(SUVmean)of normal tissue,the maximum standard uptake value(SUVmax),peak standard uptake value(SUVpeak),SUVmean of lesions,total lesion glycolysis(TLG)and metabolic tumor volume(MTV)were compared.Results No significant difference of imaging quality nor semi-quantitative parameters of lesions(all P＞0.05),while significant differences of SUVmean of aortic arch,liver,lumbar vertebra and spinal cord were found between 2 kind images(all P＜0.05).Strong correlations of SUVmax,SUVmean,MTV and TLG of lesions(r,=0.734-0.890,P＜0.001),and high correlation of SUVpeak(rs=0.919,P＜0.001)were found between 2 kind images.The consistency of SUVmax,SUVpeak,SUVmean,TLG and MTV at the lesion site between 2 kind images were very good to extremely good(ICC=0.891-0.986,all P＜0.001),and the differences of all above semi-quantitative parameters were within 95％confidence interval.Conclusion Imaging quality of 18F-FDG whole-body images obtained with domestic NeuWise and Philips Ingenuity TF PET/CT scanners could meet the requirements of clinical diagnosis and treatment,and semi-quantitative parameters obtained based on both images had good consistencies.

[Objective] To explore the characteristics of lifestyle behaviors and mental health status among blood donors in Guangzhou, and to investigate the correlation between donation frequency and these factors. [Methods] A cross-sectional study was conducted among 13 042 whole blood donors from 17 street blood donation sites of Guangzhou Blood Center from May to August 2020. Descriptive analysis was used to describe the characteristics of lifestyle behaviors and mental health status among blood donors in Guangzhou. Ordinal multinomial logistic regression model was used to analyze the correlation between donation frequency and these factors. [Results] It was found that some of 13 042 blood donors had unhealthy habits, such as 6.8% (698/10 214，2 828 missing values) had severe tobacco dependence, 30.6% (3 997/13 042) had low exercise levels, 38.8%（5 056/13 042）had poor sleep quality, and 2.2% (271/12 159，883 missing values) had alcohol dependence. In addition, 2.8% (364/13 042) and 1.3% (172/13 042) of the donors may have moderate to severe depression and anxiety symptoms, respectively. The results of the ordinal multinomial logistic regression model showed that exercise level was significantly negatively correlated with the degree of depression and anxiety among blood donors. With the decrease in exercise level, the possibility of depression and anxiety among donors increased significantly. BMI, household income, education level, marital status, donation frequency, alcohol consumption and smoking had no significant correlation with the mental health status of donors. [Conclusion] Improving the exercise habits of blood donors may help enhance their mental health level. It is recommended that blood station staff strengthen the content of exercise when providing health education to blood donors to maintain a healthy lifestyle. It also suggests that there may be a certain degree of under-diagnosis of mental health problems in the process of health consultation before blood donation, and conducting more comprehensive and effective mental health assessments for blood donors is recommended.

OBJECTIVE To evaluate the innovation and sustainable development ability of traditional Chinese medicine(TCM) preparation in medical institutions, and to provide reference for the decision-making of administrative departments and the sustainable development of TCM preparation in hospitals. METHODS Based on the information data of 133 medical institutions in Zhejiang Province, analytic hierarchy process was used to construct the evaluation system of innovation and sustainable development ability of TCM preparation of medical institutions, and the prediction model of evaluation system of TCM preparation room/center was built by back propagation neural network. RESULTS The evaluation index system for innovation and sustainable development of TCM preparations in medical institutions included 4 second-level indexes and 19 third-level indexes. Among them, the number of varieties developed, the number of over one million varieties, the number of registered varieties, the way of research and development, and the area of preparation room had higher weight values, which were 0.15758, 0.12928,0.09343, 0.07879 and 0.07458, respectively. The entropy weight method and analytic hierarchy process index weights were used to construct the back propagation neural network. The overall correlation coefficients of the training models of the Levenberg-Marquardt algorithm and the Scaled Conjugate Gradient algorithm were 0.98983 and 0.93480, respectively. CONCLUSION This study establishes a scientific and effective comprehensive evaluation system for the innovation and sustainable development ability of TCM preparations in medical institutions, which can realize the prediction and evaluation of the comprehensive ability of TCM preparations in medical institutions.

Objective:To explore the value of salivary gland imaging based on deep learning and Delta radiomics in assessing salivary gland injury after 131I treatment in post-thyroidectomy thyroid cancer patients. Methods:A retrospective analysis on 223 patients (46 males, 177 females, age(47.7±14.0) years ) with papillary thyroid cancer, who underwent total thyroidectomy and 131I treatment in Affiliated Hospital of Guilin Medical University between December 2019 and January 2022, was conducted. All patients underwent salivary gland 99Tc mO 4- imaging before and after 131I therapy. The patients were categorized according to salivary gland function based on 99Tc mO 4- imaging results (normal salivary gland vs salivary gland injury), and divided into training and test sets in a ratio of 7∶3. A ResNet-34 neural network model was trained using images at the time of maximum salivary gland radioactivity and those based on background radioactivity counts for structured image feature data. The Delta radiomics approach was then used to subtract the image feature values of the two periods, followed by feature selection through t-test, correlation analysis, and the least absolute shrinkage and selection operator( LASSO) algorithm, to develop logistic regression (LR), support vector machine (SVM), and K-nearest neighbor (KNN) predictive models. The diagnostic performance of 3 models for salivary gland function on the test set was compared with that of the manual interpretation. The AUCs of the 3 models on the test set were compared (Delong test). Results:Among the 67 cases of the test set, the diagnostic accuracy of 3 physicians were 89.6%(60/67), 83.6%(56/67), and 82.1%(55/67) respectively, with the time required for diagnosis of 56, 74 and 55 min, respectively. The accuracies of LR, SVM, and KNN models were 91.0%(61/67), 86.6%(58/67), and 82.1%(55/67), with the required times of 12.5, 15.3 and 17.9 s, respectively. All 3 radiomics models demonstrated good classification and predictive capabilities, with AUC values for the training set of 0.972, 0.965, and 0.943, and for the test set of 0.954, 0.913, and 0.791, respectively. There were no significant differences among the AUC values for the test set ( z values: 0.72, 1.18, 1.82, all P>0.05). Conclusion:The models based on deep learning and Delta radiomics possess high predictive value in assessing salivary gland injury following 131I treatment after surgery in patients with thyroid cancer.