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 analyze the efficacy, safety, and learning curve of the three-lobe holmium laser enucleation of the prostate(HoLEP)in the treatment of benign prostatic hyperplasia(BPH)in a county-level hospital.Methods A retrospective analysis was conducted on the clinical data of 65 BPH patients who underwent the three-lobe HoLEP performed by a single surgeon at the Department of Urology, Yining County People's Hospital, during Dec.2023 and Jun.2024.The enucleation efficiency was calculated by dividing the weight of the enucleated prostatic tissue by the enucleation time.A case scatter diagram of enucleation efficiency was plotted according to the chronological order of the operations, and the learning curve was analyzed after fitting.Based on the inflection points of the learning curve, the learning process was divided into the initial learning, mastery, and proficiency phases.The basic clinical data, perioperative indicators, postoperative complications, and follow-up indicators were compared among the different learning phases.Results All 65 procedures were successfully completed.Marked by the enucleation efficiency reaching a plateau, cases 1-20 were defined as the initial learning phase, cases 21-40 as the mastery phase, and cases 41 onwards as the proficiency phase.The prostate volume in the proficiency phase was significantly larger than that in the initial learning and mastery phases(both P＜0.05).There were no statistically significant differences in other baseline characteristics among the three groups(all P＞0.05).The operation time [(105.50±19.12)min vs.(85.25±26.92)min vs.(69.00±23.58)min] and enucleation time [(76.90±14.19)min vs.(63.70±22.24)min vs.(48.80±20.48)min] showed a decreasing trend across the three groups(all P＜0.05).The enucleation efficiency in both the mastery and proficiency phases was significantly higher than that in the initial learning phase [(1.16±0.44)g/min vs.(0.85 ±0.25)g/min, P＜0.05;(1.36±0.49)g/min vs.(0.85±0.25)g/min, P＜0.05].The enucleation efficiency in the proficiency phase was slightly higher than that in the mastery phase, but the difference was not statistically significant(P= 0.389).There was no significant difference in the incidence of perioperative complications among the three groups(all P＞0.05).At the 6-month follow-up, the international prostate symptom score(IPSS), post-void residual(PVR), and maximum urinary flow(Qmax)were significantly improved compared to preoperative values in all three groups(all P＜0.05);however, no significant differences were observed among the three groups(all P＞0.05).Conclusion The three-lobe HoLEP for the treatment of BPH is safe and effective in a county-level hospital setting.Surgeons with some experience in endoscopic surgery can preliminarily master this technique after a learning period of approximately 20 procedures.

Objective:This study aimed to validated the diagnostic accuracy of Global Leadership Initiative on Malnutrition(GLIM)criteria for malnutrition in pancreatic cancer patients undergoing pancreaticoduodenectomy and to evaluated its prognostic value for postoperative outcome.Methods:A retrospective analysis was conducted on 230 consecutive pancreatic cancer patients who underwent pancreaticoduodenectomy at the Department of Pancreatobiliary Surgery,Sun Yat-sen University Cancer Center,between January 2018 to January 2024.Patients were stratified into malnutrition group and non-malnutrition group using Nutritional Risk Screening 2002(NRS 2002)and GLIM criteria.Multivariable logistic regression identified independent risk factors for postoperative morbidity.Results:GLIM criteria identified malnutrition in 96 patients(41.7％).Compared with the non-malnourished group,the number of preoperative nutritional support(t=20.038,P<0.001),the number of preoperative enteral nutrition support(t=8.377,P=0.004),the number of preoperative parenteral nutrition support(t=22.302,P<0.001),the number of anemia(t=8.037,P=0.005)and preoperative parenteral nutrition use days(t=-2.898,P=0.009),the difference was statistically significant.There were statistically significant differences in C-reactive protein(t=10.944,P=0.008),NLR(t=-2.523,P=0.012)and PNI(t=-2.397,P=0.017)between the two groups before surgery.Preoperative BMI(t=-4.410,P<0.001)was significantly lower in the malnourished group.The number of postoperative parenteral nutrition days(Z=-2.283,P=0.022)and amino acid supplementation during postoperative hospitalization were significantly higher in the malnourished group(Z=-2.309,P=0.021).The incidence of malnutrition was higher in patients with Clavien-Dindo grade≥Ⅲ(P=0.030)and intra-abdominal infections(P=0.049).Multivariable analysis identified preoperative weight loss(OR=2.154,95％CI:1.158～4.005;P=0.015)and BMI reduction(OR=0.175,95％CI:0.040～0.775;P=0.022)as independent predictors of postoperative complications.Conclusions:The GLIM standard effectively characterize malnutrition status in pancreatic cancer patients after pancreaticoduodenectomy patients and demonstrate superior predictive performance for postoperative morbidity.It has good predictive performance and clinical application value.

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.

With the rapid development of generative artificial intelligence(GAI)technologies,their widespread application in academic research and writing is continuously expanding the boundaries of sci-entific inquiry.However,this trend has also raised a series of ethical and regulatory challenges,inclu-ding issues related to authorship,content authenticity,citation accuracy,and accountability.In light of the growing involvement of AI in generating academic content,establishing an open,controllable,and trustworthy ethical governance framework has become a key task for safeguarding research integrity and maintaining trust within the academic community.This expert consensus outlines ethical requirements across key stages of AI-assisted academic writing-including topic selection,data management,citation practices,and authorship attribution.It aims to clarify the boundaries and ethical obligations surrounding AI use in academic writing,ensuring that technological tools enhance efficiency without compromising in-tegrity.The goal is to provide guidance and institutional support for building a responsible and sustainable research ecosystem.

Objective To summarize the changing prevalence of carbapenem resistance in Enterobacterales based on the data of CHINET Antimicrobial Resistance Surveillance Program from 2015 to 2021 for improving antimicrobial treatment in clinical practice.Methods Antimicrobial susceptibility testing was performed using a commercial automated susceptibility testing system according to the unified CHINET protocol.The results were interpreted according to the breakpoints of the Clinical & Laboratory Standards Institute(CLSI)M100 31st ed in 2021.Results Over the seven-year period(2015-2021),the overall prevalence of carbapenem-resistant Enterobacterales(CRE)was 9.43％(62 342/661 235).The prevalence of CRE strains in Klebsiella pneumoniae,Citrobacter freundii,and Enterobacter cloacae was 22.38％,9.73％,and 8.47％,respectively.The prevalence of CRE strains in Escherichia coli was 1.99％.A few CRE strains were also identified in Salmonella and Shigella.The CRE strains were mainly isolated from respiratory specimens(44.23±2.80)％,followed by blood(20.88±3.40)％and urine(18.40±3.45)％.Intensive care units(ICUs)were the major source of the CRE strains(27.43±5.20)％.CRE strains were resistant to all the β-lactam antibiotics tested and most non-β-lactam antimicrobial agents.The CRE strains were relatively susceptible to tigecycline and polymyxins with low resistance rates.Conclusions The prevalence of CRE strains was increasing from 2015 to 2021.CRE strains were highly resistant to most of the antibacterial drugs used in clinical practice.Clinicians should prescribe antimicrobial agents rationally.Hospitals should strengthen antibiotic stewardship in key clinical settings such as ICUs,and take effective infection control measures to curb CRE outbreak and epidemic in hospitals.

Objective To characterize the changing species distribution and antibiotic resistance profiles of respiratory isolates in hospitals participating in the CHINET Antimicrobial Resistance Surveillance Program from 2015 to 2021.Methods Commercial automated antimicrobial susceptibility testing systems and disk diffusion method were used to test the susceptibility of respiratory bacterial isolates to antimicrobial agents following the standardized technical protocol established by the CHINET program.Results A total of 589 746 respiratory isolates were collected from 2015 to 2021.Overall,82.6％of the isolates were Gram-negative bacteria and 17.4％were Gram-positive bacteria.The bacterial isolates from outpatients and inpatients accounted for(6.0±0.9)％and(94.0±0.1)％,respectively.The top microorganisms were Klebsiella spp.,Acinetobacter spp.,Pseudomonas aeruginosa,Staphylococcus aureus,Haemophilus spp.,Stenotrophomonas maltophilia,Escherichia coli,and Streptococcus pneumoniae.Each microorganism was isolated from significantly more males than from females(P＜0.05).The overall prevalence of methicillin-resistant S.aureus(MRSA)was 39.9％.The prevalence of penicillin-resistant S.pneumoniae was 1.4％.The prevalence of extended-spectrum β-lactamase(ESBL)-producing E.coli and K.pneumoniae was 67.8％and 41.3％,respectively.The overall prevalence of carbapenem-resistant E.coli,K.pneumoniae,Enterobacter cloacae,Pseudomonas aeruginosa,and Acinetobacter baumannii was 3.7％,20.8％,9.4％,29.8％,and 73.3％,respectively.The prevalence of β-lactamase was 96.1％in Moraxella catarrhalis and 60.0％in Haemophilus influenzae.The H.influenzae isolates from children(＜18 years)showed significantly higher resistance rates to β-lactam antibiotics than the isolates from adults(P＜0.05).Conclusions Gram-negative bacteria are still predominant in respiratory isolates associated with serious antibiotic resistance.Antimicrobial resistance surveillance should be strengthened in clinical practice to support accurate etiological diagnosis and appropriate antimicrobial therapy based on antimicrobial susceptibility testing results.

OBJECTIVE: To investigate chronic hepatitis C virus (HCV) infection's effect on gestational liver function, pregnancy and delivery complications, and neonatal development. METHODS: A total of 157 HCV antibody-positive (anti-HCV[+]) and HCV RNA(+) patients (Group C) and 121 anti-HCV(+) and HCV RNA(-) patients (Group B) were included as study participants, while 142 anti-HCV(-) and HCV RNA(-) patients (Group A) were the control group. Data on biochemical indices during pregnancy, pregnancy complications, delivery-related information, and neonatal complications were also collected. RESULTS: Elevated alanine aminotransferase (ALT) rates in Group C during early, middle, and late pregnancy were 59.87%, 43.95%, and 42.04%, respectively-significantly higher than Groups B (26.45%, 15.70%, 10.74%) and A (23.94%, 19.01%, 6.34%) ( P < 0.05). Median ALT levels in Group C were significantly higher than in Groups A and B at all pregnancy stages ( P < 0.05). No significant differences were found in neonatal malformation rates across groups ( P > 0.05). However, neonatal jaundice incidence was significantly greater in Group C (75.16%) compared to Groups A (42.25%) and B (57.02%) ( χ ² = 33.552, P < 0.001). HCV RNA positivity during pregnancy was an independent risk factor for neonatal jaundice ( OR = 2.111, 95% CI 1.242-3.588, P = 0.006). CONCLUSIONS Chronic HCV infection can affect the liver function of pregnant women, but does not increase the pregnancy or delivery complication risks. HCV RNA(+) is an independent risk factor for neonatal jaundice.
Humans ; Female ; Pregnancy ; Adult ; Pregnancy Complications, Infectious/epidemiology* ; Retrospective Studies ; Pregnancy Outcome ; Infant, Newborn ; Viremia/virology* ; Hepatitis C ; Hepacivirus/physiology* ; Hepatitis C, Chronic/virology* ; Young Adult ; Alanine Transaminase/blood*

OBJECTIVE: To investigate the association between long-term glycemic control and cerebral infarction risk in patients with diabetes through a large-scale cohort study. METHODS: This prospective, community-based cohort study included 12,054 patients with diabetes. From 2006 to 2012, 38,272 fasting blood glucose (FBG) measurements were obtained from these participants. FBG trajectory patterns were generated using latent mixture modelling. Cox proportional hazards models were applied to assess the subsequent risk of cerebral infarction associated with different FBG trajectory patterns. RESULTS: At baseline, the mean age of the participants was 55.2 years. Four distinct FBG trajectories were identified based on FBG concentrations and their changes over the 6-year follow-up period. After a median follow-up of 6.9 years, 786 cerebral infarction events were recorded. Different trajectory patterns were associated with significantly varied outcome risks (Log-Rank P < 0.001). Compared with the low-stability group, Hazard Ratio ( HR) adjusted for potential confounders were 1.37 for the moderate-increasing group, 1.23 for the elevated-decreasing group, and 2.08 for the elevated-stable group. CONCLUSION Sustained high FBG levels were found to play a critical role in the development of ischemic stroke among patients with diabetes. Controlling FBG levels may reduce the risk of cerebral infarction.
Humans ; Cerebral Infarction/blood* ; Middle Aged ; Male ; Female ; Blood Glucose/analysis* ; Fasting/blood* ; Aged ; Prospective Studies ; Risk Factors ; Diabetes Mellitus/blood* ; Adult ; Proportional Hazards Models