Objective To analyze the effect of releasing the lower pulmonary ligament on right residual lung expansion after right upper lobe resection under different body mass index (BMI) levels. Methods The clinical data of patients who underwent thoracoscopic right upper lobe resection in the First Affiliated Hospital with Nanjing Medical University from 2021 to 2022 were retrospectively analyzed. Patients were divided into a group A (17 kg/m2＜BMI≤23 kg/m2), a group B (23 kg/m2＜BMI≤29 kg/m2) and a group C (BMI＞29 kg/m2) according to BMI. The presence of residual cavity was judged by chest X-ray at 7-10 days after operation, the degree of compensation change of the right main bronchus angle was measured, and the changes in lung volume were determined by CT three-dimensional reconstruction. Results A total of 157 patients who underwent thoracoscopic right upper lobe resection were included, including 71 males and 86 females, with an average age of (59.7±11.2) years. There were 50 patients in the group A, 75 patients in the group B, and 32 patients in the group C. In the group A, compared with those without releasing the lower pulmonary ligament, patients with releasing had a lower incidence of postoperative residual cavity (P=0.016), greater changes in bronchus angle (P<0.001), and smaller changes in lung volume (P<0.001). In the group B and C, there was no significant effect of releasing the lower pulmonary ligament on postoperative residual cavity, bronchus angle, and lung volume changes (P>0.05). Conclusion For patients with thin and long body shape and low BMI, releasing the lower pulmonary ligament is helpful to promote the expansion of the residual lung after right upper lobe resection and reduce the occurrence of postoperative residual cavity in patients.

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 machine learning-based precision blood donor recruitment model at the county level and assess its generalizability and applicability. Methods: A retrospective study was conducted using blood donation and SMS recruitment data from the Taicang Branch of the Suzhou Blood Center between 2019 and 2024. Multiple machine learning algorithms were employed, including extreme gradient boosting, support vector machine, k-nearest neighbor, logistic regression, decision tree, random forest, and multilayer perceptron. These were combined with techniques such as synthetic minority oversampling, undersampling, and cost-sensitive learning (using MFE and MSFE loss functions). Model parameters were optimized through grid search to identify the best-performing model. Results: In a prospective comparative study against conventional methods, the machine learning models increased the recruitment success rate among high-willingness donors by an average of 129.15%, and the recruitment efficiency per SMS improved by 125.02% compared with the traditional method. Under full-scale SMS sending, the recruitment rate per SMS increased by 42.61%, and SMS sending efficiency improved by 31.77%, significantly enhancing recruitment performance. Conclusion: This study represents the first application of a machine learning-based precision donor recruitment model at the county-level in China. The precise recruitment framework not only improves recruitment efficiency and reduces recruitment costs but also demonstrates strong scalability and generalizability. It provides a scientific and feasible intelligent pathway to ensure the safety and sustainability of the blood supply.

Objective To investigate the efficacy of varying doses of methylprednisolone(MP)on mice with acute exacerbations of chronic obstructive pulmonary disease(AECOPD)induced with influenza A virus(IAV).Methods Mouse model of COPD was established using LPS combined with smoking for 12 weeks,and then these COPD mice were treated with administration of 40 μL IAV via nasal drip to establish a AECOPD model.A total of 15 AECOPD mice were randomly divided into low-,medium-and high-dose MP groups,oseltamivir group and blank group.The body weight and survival time were monitored within 10 d after IAV infection.On days 1,3,and 5 post-treatment,lung function was assessed using whole-body plethysmography(WBP),inflammatory factors in bronchoalveolar lavage fluid(BALF)were quantified with ELISA,viral titers in BALF were determined using plaque assays,and colony-forming units were evaluated with blood agar plates.Immunofluorescence analysis:① Pulmonary immunofluorescence assay:Mice were randomly categorized into(n=4):LPS 1-day group,LPS 3-day group,and LPS+MP treatment group.All groups received an initial dose of LPS via atomization;subsequently,the LPS+MP treatment group received a single gavage dose of MP.Lung tissues were harvested from the 1-day LPS group on 1 d post-treatment,and from the 3-day LPS and LPS+MP groups on 3 d for immunofluorescence staining.② Cellular immunofluorescence assay:Mouse bone marrow neutrophils were classified into blank control(no intervention),LPS stimulation(LPS group),MP intervention with LPS stimulation(LPS+MP group),and MP intervention alone(MP group).The above cells were collected in 4 h after corresponding interventions for subsequent cellular immunofluorescence analysis.Results ①The medium-dose MP group demonstrated the most significant improvement in survival rate,weight recovery,and lung function when compared to other groups(P＜0.05).② Treatment of medium-dose MP obviously reduced the levels of IL-6 and neutrophil extracellular traps(NETs)(P＜0.05),while,elevated inflammatory factors and NETs were observed in the high-dose MP group on day 5 post-treatment.③ Notable decline in the lung injury score was found in the medium-dose MP group than the other groups(P＜0.05).④The high-dose MP group exhibited substantial bacterial proliferation and delayed viral clearance since day 5 after treatment.Conclusion Medium-dose MP shows best efficacy in treatment of IAV-induced AECOPD,and the dose neither delays viral clearance nor increases the risk of bacterial infection following viral infection.

The rapid and non-destructive detection of constituent content of fresh tea leaves shows an important reference value for quality identification of tea.Visible near infrared(Vis-NIR)spectroscopy has been used for qualitative and quantitative analysis of chemical components in plant samples with the advantages such as simple,rapid and non-destructive detection.In this study,residual attention convolutional neural network(RACNN)was used to predict the internal constituent content of fresh tea leaves.Firstly,the reflectance spectral data of the samples in the Vis-NIR band range and the constituent contents of gallic acid(GA),gallocatechin(GC),epigallocatechin(EGC),and epigallocatechin gallate(ECG)in fresh tea leaves were collected.Based on the preprocessing of the spectral data,the contents of the four components were predicted using a partial least squares regression(PLSR)model,and the optimal preprocessing was determined.Subsequently,the characteristic bands were extracted using the random forest(RF)algorithm.Finally,the performances of PLSR,convolutional neural network(CNN)and RACNN models were compared.The results showed that for GA,the RACNN model worked best with a validation set coefficient of determination(R2)of 0.946 and a root mean square error of the prediction set(RMSEP)of 1.173;for GC,the RACNN model works best with a validation set R2 of 0.928 and RMSEP of 6.081;for EGC,the RACNN model works best with a validation set R2 of 0.891 and a RMSEP of 15.197;for ECG,the RACNN model worked best with a validation set R2 of 0.878 and a RMSEP of 7.837.The RACNN model established by Vis-NIR spectroscopy combined with chemometrics could realize the accurate detection of the contents of components in fresh tea.

Digital polymerase chain reaction(dPCR)enables absolute quantitative detection of nucleic acid samples.Since the quantitative upper and lower limits of detectable samples mainly depend on the volume and number of single droplets,the abundance of the sample to be tested,the volume of single droplets,and the number of droplets need to be adapted.For samples with unknown abundance,repeated adjustment of droplet size is not allowed.In this study,a one-step double-volume droplet generation method was proposed,and a double-volume droplet microfluidic chip was developed to verify the quantitative detection capability of the chip using a duck-derived kit.The results showed that the droplets with different volumes had different quantitative capabilities.Large-volume droplets had higher reliability for low-abundance sample concentrations,while small-volume droplets had advantages in detecting high-abundance sample concentrations.The double-volume droplets produced by the double-volume droplet microfluidic chip proposed in this study greatly improved the reliability of quantitative capabilities,and had broad application prospects in detection of precious nucleic acid samples with unknown abundance in the field of microfluidic PCR.

The method of decocting and taking medicine can directly influence the efficacy of the Chinese herbal medicine,and is the key to enhancing efficacy and reducing toxicity.As the originator of classic books for traditional Chinese medicine(TCM)prescriptions,Treatise on Exogenous Febrile Diseases has recorded various specific methods of decocting and taking medicine in details.This paper summarized and sorted out various methods of decocting the same Chinese herbal medicine,re-decocting method with the removal of dregs(for concentrating medicinal solution),method of decocting pills,dosage of medicine,time for taking medicine,and notices and healthcare after medication recorded in Treatise on Exogenous Febrile Diseases.Moreover,the dialectical view in the method of decocting medicine recorded in Treatise on Exogenous Febrile Diseases.was explored.The special method of decocting and taking medicine in Treatise on Exogenous Febrile Diseases included the TCM dialectical view of using the same Chinese herbal medicine for the treatment of different diseases,consideration of both Chinese herbal medicine and syndromes,drastic purgatives for chasing long-term efficacy.The method of taking medicine contained the TCM dialectical view of modification of the medicine dosage according to syndrome differentiation,adapting to the general trend,and suspension after medicine starting an effect.It is believed that the method of decocting and taking medicine for the prescriptions in Treatise on Exogenous Febrile Diseases is established according to syndrome differentiation,and the utilization of various methods of decocting and taking medicine as well as notices and healthcare after medication in accordance with the characteristics of diseases and syndromes ensures the prescriptions meeting the pathogenesis,and then enhance the clinical efficacy.

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 assess the effectiveness of the evaluation of military physical function(EMPF)system in predicting the occurrence of military training injuries among new recruits to provide scientific guidance and methodological choice for military training.Methods A total of 527 new recruits from 5 grassroots units from July 2016 to February 2018 were selected for the study.The recruits underwent EMPF testing,and their military training injuries were monitored over a 2-year follow-up period.Those who sustained injuries during training were divided into injury group(n=163),while the remaining recruits were placed in healthy group(n=364).The predictive ability of the total EMPF score for training injuries was assessed using the receiver operating characteristic curve(ROC),and the correlation between the total EMPF score,individual test scores,and military training injuries were analyzed using binary logistic regression.Results The total EMPF score of new recruits in injury group(19.52±1.97)was significantly lower than that of healthy group(24.31±1.54)(P<0.001),which also demonstrated a high diagnostic value in predicting the risk of military training injuries,with an area under the curve(AUC)of ROC of 0.971(P<0.001).A cut-off value of 22 scores was found to have the highest accuracy in predicting future training injuries,with an odds ratio(OR)of 25.63,sensitivity of 0.939,specificity of 0.879,positive likelihood ratio of 7.76,and a post-test probability of 0.67.Binary logistic regression analysis revealed that 6 EMPF tests,including holding the ball over and leaning back,bending forward and touching the ground with the ball,lunge squat and twist,swallow balance with holding the ball afterward,vertical jump,and respiratory pattern assessment,were negatively associated with the risk of military training injuries(P<0.0001).Conclusion The EMPF system can effectively predict the risk of military training injuries,with military personnel whose total EMPF score is less than 22 being at higher risk of sustaining such injuries.