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.

To investigate the transcriptionally regulatory mechanism of the senp8 promoter in yellow cat-fish(Pelteobagrus fulvidraco);this study used P.fulvidraco as the research subject.Dual-luciferase re-porter assay and electrophoretic mobility shift assay were employed to analyze the functional activity of the promoter;coupled with in vivo experiments.The results indicated that the 2 045 bp senp8 promoter se-quence contained key transcription factor binding sites such as SP1;TATA-Box;CCAAT-Box;SREBP1;PPARα;and PPARγ.The binding sites of SREBP1(-901/-910 bp);PPARα(-1 291/-1 308 bp);and PPARγ(-1 292/-1 306 bp)in the senp8 promoter positively regulate its activity;and oleic acid or palmitic acid promote this binding.Furthermore;high-fat feeding promoted the expression of the senp8 gene and its protein in the liver of P.fulvidraco;oleic acid or palmitic acid treatment significantly en-hanced the activity of the senp8 promoter;and this enhancement could be achieved through the regulatory effects of SREBP1;PPARα;and PPARγ response elements.Additionally;high-fat feeding influenced the mRNA and protein expression levels of genes related to deSUMOylation modification in the liver of P.fulvidraco.This study provides new insights into the relationship between deSUMOylation modification and the regulation of lipid metabolism in the vertebrates.

To investigate the transcriptionally regulatory mechanism of the senp8 promoter in yellow cat-fish(Pelteobagrus fulvidraco);this study used P.fulvidraco as the research subject.Dual-luciferase re-porter assay and electrophoretic mobility shift assay were employed to analyze the functional activity of the promoter;coupled with in vivo experiments.The results indicated that the 2 045 bp senp8 promoter se-quence contained key transcription factor binding sites such as SP1;TATA-Box;CCAAT-Box;SREBP1;PPARα;and PPARγ.The binding sites of SREBP1(-901/-910 bp);PPARα(-1 291/-1 308 bp);and PPARγ(-1 292/-1 306 bp)in the senp8 promoter positively regulate its activity;and oleic acid or palmitic acid promote this binding.Furthermore;high-fat feeding promoted the expression of the senp8 gene and its protein in the liver of P.fulvidraco;oleic acid or palmitic acid treatment significantly en-hanced the activity of the senp8 promoter;and this enhancement could be achieved through the regulatory effects of SREBP1;PPARα;and PPARγ response elements.Additionally;high-fat feeding influenced the mRNA and protein expression levels of genes related to deSUMOylation modification in the liver of P.fulvidraco.This study provides new insights into the relationship between deSUMOylation modification and the regulation of lipid metabolism in the vertebrates.

Medical device related infections caused by bacteria are common complications in clinical practice,and preventing bacterial colonization on the surface of medical materials is one of the important challenges in the medical field.Therefore,there is an urgent need to construct medical coatings that combine antibacterial properties and biocompatibility.In this study,a γ-polyglutamic acid(γ-PGA)complex with long-chain alkyl quaternary ammonium salts formed by electrostatic and hydrophobic interactions was prepared,which was insoluble in water but soluble in organic solvents(e.g.,ethanol),and was capable of constructing antimicrobial coatings on the surfaces of medical materials in a simple and efficient manner.The bactericidal effect of the coating was verified using viable bacteria counting experiments,and the results showed that the bactericidal rate of the coated thermoplastic polyurethane(TPU)membrane against Staphylococcus aureus was greater than 99.9%compared with that of the uncoated TPU membrane.In addition,a cytotoxicity assay was performed using the L929 fibroblast and cell proliferation detection kit(CCK-8),which showed that the survival rate of L929 fibroblasts on coated TPU was greater than 90%.Meanwhile,the hemolysis rate of coated erythrocytes was tested using fresh rabbit red blood cells(RBCs),and the hemolysis rate on the coated TPU surface was 1.5%.The above results indicated that the coating had good biocompatibility.The preparation method of medical antibacterial coating reported in this study provided a new idea for preventing bacterial infections related to implantable/interventional medical devices.

This study investigated the effects of Rehmanniae Radix Praeparata on striatal neuronal apoptosis in rats with attention deficit hyperactivity disorder(ADHD) based on the B-cell lymphoma-2(Bcl-2)/Bcl-2-associated X protein(Bax)/caspase-3 signaling pathway. Twenty-four 3-week-old male spontaneously hypertensive rats(SHR) were randomly divided into a model group, a methylphenidate group(2 mg·kg~(-1)·d~(-1)), and a Rehmanniae Radix Praeparata group(2.4 mg·kg~(-1)·d~(-1)). Age-matched male Wistar Kyoto(WKY) rats were used as the normal control group, with 8 rats in each group. The rats were administered by gavage for 28 days. Body weight and food intake were recorded for each group. The open field test and elevated plus maze test were used to assess hyperactivity and impulsive behaviors. Nissl staining was used to detect changes in striatal neurons and Nissl bodies. Terminal deoxynucleotidyl transferase dUTP nick end labeling(TUNEL) fluorescence staining was used to detect striatal cell apoptosis. Western blot was employed to detect the expression levels of Bcl-2, Bax, and caspase-3 proteins in the striatum. The results showed that compared with the model group, Rehmanniae Radix Praeparata significantly reduced the total movement distance, average movement speed, and central area residence time in the open field test, and significantly reduced the ratio of open arm entries, open arm stay time, and head dipping in the elevated plus maze test. Furthermore, it increased the number of Nissl bodies in striatal neurons, significantly downregulated the apoptosis index, significantly increased Bcl-2 protein expression and the Bcl-2/Bax ratio, and reduced Bax and caspase-3 protein expression. In conclusion, Rehmanniae Radix Praeparata can reduce hyperactivity and impulsive behaviors in ADHD rats. Its mechanism may be related to the regulation of the Bcl-2/Bax/caspase-3 signaling pathway in the striatum, enhancing the anti-apoptotic capacity of striatal neurons.
Animals ; Male ; Apoptosis/drug effects* ; Rats ; Drugs, Chinese Herbal/administration & dosage* ; Caspase 3/genetics* ; Proto-Oncogene Proteins c-bcl-2/genetics* ; bcl-2-Associated X Protein/genetics* ; Rehmannia/chemistry* ; Attention Deficit Disorder with Hyperactivity/physiopathology* ; Signal Transduction/drug effects* ; Neurons/cytology* ; Rats, Inbred SHR ; Rats, Inbred WKY ; Humans ; Corpus Striatum/cytology* ; Plant Extracts

Molecular imprinting is a biomimetic technique that simulates the specific recognition of biological macromolecules such as antibody. Based on molecular imprinting and high-specificity affinity analysis,the biomimetic imprinting affinity analysis (BIA) possesses many advantages such as high sensitivity,strong tolerance,good specificity and low cost,and thus,it has shown excellent prospects in food safety detection,pharmaceutical analysis and environmental pollution monitoring. In this review,the construction methods of recognition interfaces for BIA were summarized,including bulk polymerization,electro-polymerization and surface molecular imprinting. The application of molecularly imprinted polymers in different analysis methods,such as radiolabeled affinity analysis,enzyme-labeled affinity analysis,fluorescence-labeled affinity analysis,chemiluminescence affinity analysis and electrochemical immunosensor was mainly discussed. Furthermore,the challenges and future development trends of BIA in practical application were elucidated. This review might provide new reference ideas and technical supports for the further development of BIA technique.

Objective　To collect data of fungi isolated from a fungus monitoring network in Hainan Province from 2013 to 2022, and analyze the drug resistance characteristics of Candida and the results of serological tests, with an aim to provide a basis for clinical diagnosis and treatment. Methods　In accordance with the National Fungal Drug Resistance Monitoring Network technical scheme, the qualifying fungal data were extracted from the microbial identification system database using SQL language, and the data information was then analyzed, with statistical processing done using SPSS 26.0 software. Results　Among 5 503 fungal isolates from clinical specimens between 2013 and 2022, cervical orifice secretions accounted for 30.37%(1 671 strains), mid-stream urine for 23.55%(1 296 strains), lower respiratory tract specimens for 25.24% (1 389 strains)[(sputum for 20.37%(1 121 strains) and alveolar lavage fluid for 4.87%(268 strains)], wound pus for 9.59%(528 strains), ascites for 5.60%(308 strains), blood for 3.67%(202 strains), cerebrospinal fluid for 0.38%(21 strains), and joint fluid for 0.04%(2 strains), with the highest number of strains isolated in 2022 and the lowest in 2013, the 2022 figure is about 2.6 times that of 2013. Among yeast-like fungi, Candida albicans had the highest proportion with 3 312 strains accounting for 60.2%; The highest resistance rate of Candida albicans was to fluconazole at 16.7%, with 2.5% being non-wild type (NWT) for amphotericin B; Candida tropicalis had the highest rate of resistance to fluconazole at 36.0%, with NWT at 41.1% for fluconazole and 3.1% for amphotericin B; Candida glabrata had a resistance rate to fluconazole of 2.8%, dose-dependent susceptibility (SDD) of 97.2%, NWT of 15.5% for fluconazole, and NWT of 8.6% for itraconazole; Candida parapsilosis had the highest resistance rate to fluconazole at 15.7% and and NWT of 8.3% for amphotericin B; Candida krusei had a 0.0% resistance rate to caspofungin; and Candida dubliniensis was 100.0% NWT to fluconazole. Of 70 cases of blood culture-positive specimens, 64 cases were detected by G test and 25 cases by Mn test, and the positive blood cultures were statistically significant when compared with the G test and Mn test, respectively (P<0.01). Conclusions　 Fungal serological test can make up for the deficiency of blood culture and distinguish fungal invasion and colonization, thus providing a basis for the effective control of fungal infection in clinical practice.

AIM: To analyze research hotspots and trends of artificial intelligence in diabetic retinopathy(DR)based on bibliometrics and high-impact papers.METHODS: Papers on artificial intelligence in DR research published in the Web of Science Core Collection(WoSCC)from January 1, 2012, to December 31, 2022 were retrieved. The data was analyzed by CiteSpace software to examine annual publication number, countries, institutions, source journal, research categories, keywords, and to perform an in-depth analysis of high-impact papers.RESULTS: A total of 1 009 papers on artificial intelligence in DR from 79 countries were included in the study, with 272 papers published in 2022. Notably, China and India contributed 287 and 234 papers, respectively. The United Kingdom exhibited a centrality score of 0.31, while the United States boasted an impressive H-index of 48. Three prominent institutions in the United Kingdom(University of London, Moorfields Eye Hospital, and University College London)and one institution in Egypt(Egyptian Knowledge Bank)all achieved a notable H-index of 14. The primary academic disciplines associated with this research field encompassed ophthalmology, computer science, and artificial intelligence. Burst keywords in the years 2021～2022 included transfer learning, vessel segmentation, and convolutional neural networks.CONCLUSION: China emerged as the leading contributor in terms of publication number in this field, while the United States stood out as a key player. Notably, Egyptian Knowledge Bank and University of London assumed leading roles among research institutions. Additionally, IEEE Access was identified as the most active journal within this domain. The research focus in the field of artificial intelligence in DR has transitioned from AI applications in disease detection and grading to a more concentrated exploration of AI-assisted diagnostic systems. Transfer learning, vessel segmentation, and convolutional neural networks hold substantial promise for widespread applications in this field.

Background and Objectives: Osteoblasts are derived from bone marrow mesenchymal stem cells (BMMSCs) and playimportant role in bone remodeling. While our previous studies have investigated the cell subtypes and heterogeneity in osteoblasts and BMMSCs separately, cell-to-cell communications between osteoblasts and BMMSCs in vivo in humans have not been characterized. The aim of this study was to investigate the cellular communication between human primary osteoblasts and bone marrow mesenchymal stem cells. Methods: and Results: To investigate the cell-to-cell communications between osteoblasts and BMMSCs and identifynew cell subtypes, we performed a systematic integration analysis with our single-cell RNA sequencing (scRNA-seq) transcriptomes data from BMMSCs and osteoblasts. We successfully identified a novel preosteoblasts subtype which highly expressed ATF3, CCL2, CXCL2 and IRF1. Biological functional annotations of the transcriptomes suggested that the novel preosteoblasts subtype may inhibit osteoblasts differentiation, maintain cells to a less differentiated status and recruit osteoclasts. Ligand-receptor interaction analysis showed strong interaction between mature osteoblasts and BMMSCs. Meanwhile, we found FZD1 was highly expressed in BMMSCs of osteogenic differentiation direction. WIF1 and SFRP4, which were highly expressed in mature osteoblasts were reported to inhibit osteogenic differentiation. We speculated that WIF1 and sFRP4 expressed in mature osteoblasts inhibited the binding of FZD1 to Wnt ligand in BMMSCs, thereby further inhibiting osteogenic differentiation of BMMSCs. Conclusions Our study provided a more systematic and comprehensive understanding of the heterogeneity of osteogenic cells. At the single cell level, this study provided insights into the cell-to-cell communications between BMMSCs and osteoblasts and mature osteoblasts may mediate negative feedback regulation of osteogenesis process.