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 validate the effectiveness of the leaf position data of multi-leaf collimator (MLC) recorded in the Infinity medical linear accelerator log file (LF).Methods:In this study, the establishment of the film dose calibration curve involved two steps. Initially, the pixel values obtained from scanning the film were converted into net optical density values. Subsequently, a sixth-order polynomial fit was performed on the net optical density and dose data. The picket fence (PF) tests were performed on the electronic portal imaging device (EPID) and film, and the corresponding acquired PF test images were analyzed using the relative leaf travel distance, relative leaf position, relative leaf pair full-width half -maximum (FWHM), and relative leaf pair neighbor region width analysis index, and compared with the relevant results in LF. In addition, to investigate the effect of gravity on the recorded leaf position data by the Infinity medical linear accelerator LF, the above PF tests were executed at 0°, 90°, 180°, and 270° gantry angles, respectively. Intergroup differences were assessed with the Wilcoxon rank-sum test.Results:The film dose calibration curve demonstrated high goodness-of-fit, with a correlation coefficient ( R2) of 0.9994. Across all gantry angles, the dispersion of overall leaf position data for EBT3 film, EPID, and LF followed a consistent pattern based on four analysis metrics: EBT3 film > EPID > LF. Moreover, for each analysis metric, inter-tool differences in standard deviations of results were consistently below 0.1 mm, and this pattern was invariant to gantry angle. Conclusions:The leaf position data recorded in the Infinity medical linear accelerator LF have a high degree of accuracy, and can be used as reference for the actual position of the MLC leaf. Furthermore, gravitational forces exhibit negligible impact on leaf position data acquired via LF.

OBJECTIVE: The relationship between fish consumption and stroke is inconsistent, and it is uncertain whether this association varies across predicted stroke risks. METHODS: A cohort study comprising 95,800 participants from the Prediction for Atherosclerotic Cardiovascular Disease Risk in China project was conducted. A standardized questionnaire was used to collect data on fish consumption. Participants were stratified into low- and moderate-to-high-risk categories based on their 10-year stroke risk prediction scores. Hazard ratios ( HRs) and 95% confidence intervals ( CIs) were estimated using Cox proportional hazard models and additive interaction by relative excess risk due to interaction (RERI), attributable proportion (AP), and synergy index (SI). RESULTS: During 703,869 person-years of follow-up, 2,773 incident stroke events were identified. Higher fish consumption was associated with a lower risk of stroke, particularly among moderate-to-high-risk individuals ( HR = 0.53, 95% CI: 0.47-0.60) than among low-risk individuals ( HR = 0.64, 95% CI: 0.49-0.85). A significant additive interaction between fish consumption and predicted stroke risk was observed (RERI = 4.08, 95% CI: 2.80-5.36; SI = 1.64, 95% CI: 1.42-1.89; AP = 0.36, 95% CI: 0.28-0.43). CONCLUSION Higher fish consumption was associated with a lower risk of stroke, and this beneficial association was more pronounced in individuals with moderate-to-high stroke risk.
Humans ; China/epidemiology* ; Male ; Female ; Stroke/etiology* ; Middle Aged ; Prospective Studies ; Incidence ; Aged ; Animals ; Fishes ; Risk Factors ; Diet ; Seafood ; Adult ; Cohort Studies

Objective:To investigate the self-perceived cognitive decline status in the community population,and to explore the association between different dietary habits and self-perceived cognitive decline.Methods:A cross-sectional study was carried out in 11 879 community residents in the three regions of Weifang,Jining,and Zoucheng in Shandong Province.The Ascertain Dementia-8 and dietary habits information questionnaire were used to assess self-perceived cognitive decline and dietary habits,and their association were analyzed using single factor and multivariate logistic regression.Results:The detection rate of self-perceived cognitive decline was 21.4％.Lo-gistic regression showed that smoking in the past was positively associated with self-perceived cognitive decline(OR=1.40,95％CI:1.14-1.73).However,intake of fruits(often,OR=0.70,95％CI:0.52-0.94;everyday,OR=0.60,95％CI:0.44-0.81),nuts(daily,OR=0.62,95％CI:0.44-0.88),mushrooms(often,OR=0.74,95％CI:0.57-0.92)and high tryptophan foods(sometimes,OR=0.79,95％CI:0.68-0.91;everyday,OR=0.54,95％CI:0.34-0.87)were negatively associated with self-perceived cognitive decline.Conclusion:Smoking history might be a risk factor for self-perceived cognitive decline,and high frequency intake of fruits,nuts,mush-rooms,and high tryptophan foods might protective factors for it.

To understand the background of Escherichia coli(E.coli)carried by artificially bred sika deer and the biological characteristics of the isolated strains,such as drug resistance and pathogenicity,in April 2024,we collected 184 fresh deer fecal samples from four deer farms in Luxiang Township,Shuangyang District,Changchun City,Jilin Province,for isolation and cultivation of E.coli.The isolates were tested for drug resistance and biochemical identification with a BD PhoenixTM-100 Automated Microbiology System.The virulence genes were detected with PCR,and the strains were molecularly typed with ERIC-PCR.A total of 165 E.coli strains were isolated from 184 samples of deer feces,with an isolation rate of 89.67%.Twenty strains had a drug resistance phenotype,and the drug resistance rate was 12.12%;these strains included 15 strains of multi-drug resistant bacteria and 11 strains of ESBL-producing bacteria.Virulence gene detection indicated that the sika deer isolates carried multiple diarrhea-associated virulence genes,such as EAST-1(12.12%),eae(1.21%),stx1(7.88%),stx2(7.27%),and STa(1.82%).ERIC-PCR demonstrated that the isolates showed high polymorphism.The ESBL-producing E.coli carried by sika deer are likely to spread drug resistance in the community and livestock population.Some isolates carried multiple diarrhea-associated virulence genes,thus posing a human transmission risk.Therefore,monitoring of drug resistance and virulence genes must be strengthened,and antibiotics must be used reasonably during the breeding process to avoid excessive use and misuse.

Objective : To construct a humanized mouse model of the interleukin-9 receptor(IL-9R) coding DNA sequence(CDS) gene and to verify the genotype and IL-9R expression in mice. Methods : The CRISPR/Cas9 genome editing technology was used to replace the exon 2-7 fragment of the il-9r gene in mouse embryonic stem cells with the corresponding human IL-9R sequence. After verifying the completion of the gene fragment replacement, tetraploid embryos were constructed and microinjected back into the oviducts of surrogate mice. Through surrogacy by female mice, homozygous humanized mice were obtained. DNA was extracted from the homozygous humanized mice IL-9R CDS gene, and their genotypes were identified by agarose gel electrophoresis after PCR amplification. Western blot was used to detect the expression of IL-9R in the spleen and thymus of homozygous humanized mice with either wild-type(WT) or IL-9R gene humanization. Results : Gel electrophoresis after PCR amplification showed that mice with only a 1 805 bp band amplified using WT primers were wild-type, while mice with 2 553 bp and 2 340 bp bands amplified using 5KI and 3KI primers, respectively, were homozygous humanized mice with IL-9R CDS gene. Western blot results indicated that the tissues of homozygous humanized mice model with IL-9R CDS gene expressed IL-9R significantly. Conclusion The humanized mouse model with IL-9R CDS gene has been successfully constructed and characterized.

Objective: To explore the breeding and genotyping of CCR2 knockout mice, and to verify the applicability of the polymerase chain reaction(PCR) method for genotype detection of CCR2 knockout mice. Methods: The introduced CCR2 pure male mice and wild-type female mice were mated and bred to produce the offspring generation, the obtained F1 generation heterozygous mice were continued to be mated. DNA was extracted by clipping the tail tissues of the mice at the age of 2 weeks, the target gene fragment was amplified by PCR, and the genotypic results were determined by agarose gel electrophoresis. The proportion of purebred progeny carrying the CCR2 knockout gene was increased by genetic crosses, the effect of CCR2 knockout in the progeny mice was verified by using Western blot against major immune cells and key organs, and flow cytometry was used to detect whether the knockout of the CCR2 gene had any effect on the function of the immune system by targeting the major immune cells. Results: CCR2 knockout mice were successfully bred and characterized, and three genotypes of F2 generation mice were obtained: CCR2+/+, CCR2+/-, and CCR2-/-. The offspring genotypes were identified by PCR, and Western blot showed extremely low CCR2 protein expression in CCR2 knockout mice. Flow analysis showed that CCR2 knockdown reduced the expression of CD4+T and Th1 cells in mouse spleen-derived T cells, but did not affect macrophage function. Conclusion Correct breeding and identification are important ways to get the pure CCR2 knockout mice, and PCR method for identifying mouse genotypes is simple, fast and reliable.