ObjectiveTo develop a predictive model and application for spontaneous passage of common bile duct stones using automated machine learning algorithms given the complexity of treatment decision-making for patients with common bile duct stones， and to reduce unnecessary endoscopic retrograde cholangiopancreatography （ERCP） procedures. MethodsA retrospective analysis was performed for the data of 835 patients who were scheduled for ERCP after a confirmed diagnosis of common bile duct stones based on imaging techniques in Changshu First People’s Hospital （dataset 1） and Changshu Traditional Chinese Medicine Hospital （dataset 2）. The dataset 1 was used for the training and internal validation of the machine learning model and the development of an application， and the dataset 2 was used for external testing. A total of 22 potential predictive variables were included for the establishment and internal validation of the LASSO regression model and various automated machine learning models. The area under the receiver operating characteristic curve （AUC）， sensitivity， specificity， and accuracy were used to assess the performance of models and identify the best model. Feature importance plots， force plots， and SHAP plots were used to interpret the model. The Python Dash library and the best model were used to develop a web application， and external testing was conducted using the dataset 2. The Kolmogorov-Smirnov test was used to examine whether the data were normally distributed， and the Mann-Whitney U test was used for comparison between two groups， while the chi-square test or the Fisher’s exact test was used for comparison of categorical data between groups. ResultsAmong the 835 patients included in the study， 152 （18.20%） experienced spontaneous stone passage. The LASSO model achieved an AUC of 0.875 in the training set （n=588） and 0.864 in the validation set （n=171）， and the top five predictive factors in terms of importance were solitary common bile duct stones， non-dilated common bile duct， diameter of common bile duct stones， a reduction in serum alkaline phosphatase （ALP）， and a reduction in gamma-glutamyl transpeptidase （GGT）. A total of 55 models were established using automated machine learning， among which the gradient boosting machine （GBM） model had the best performance， with an AUC of 0.891 （95% confidence interval： 0.859‍ ‍—‍ ‍0.927）， outperforming the extreme randomized tree mode， the deep learning model， the generalized linear model， and the distributed random forest model. The GBM model had an accuracy of 0.855， a sensitivity of 0.846， and a specificity of 0.857 in the test set （n=76）. The variable importance analysis showed that five factors had important influence on the prediction of spontaneous stone passage， i.e.， were solitary common bile duct stones， non-dilated common bile duct， a stone diameter of <8 mm， a reduction in serum ALP， and a reduction in GGT. The SHAP analysis of the GBM model showed a significant increase in the probability of spontaneous stone passage in patients with solitary common bile duct stones， non-dilated common bile duct， a stone diameter of <8 mm， and a reduction in serum ALP or GGT. ConclusionThe GBM model and application developed using automated machine learning algorithms exhibit excellent predictive performance and user-friendliness in predicting spontaneous stone passage in patients with common bile duct stones. This application can help avoid unnecessary ERCP procedures， thereby reducing surgical risks and healthcare costs.

Objective To explore the association of working hours and shift work with occupational stress, fatigue accumulation, and depressive symptoms among primary community medical workers. Methods A total of 516 medical workers from five community medical service centers in Pudong New Area, Shanghai City, were selected as the research subjects using the convenience sampling method. The Core Scale of Occupational Stress Measurement, the Workers' Fatigue Accumulation Self-diagnosis Questionnaire, and the Patient Health Questionnaire were used to assess research subjects' occupational stress, fatigue accumulation, and depressive symptoms, respectively. Results Long working hours (>40 hours/week) were reported by 50.4% of workers among the research subjects, while shift works were reported by 16.9% of the workers. The detection rates of occupational stress, fatigue accumulation, and depressive symptoms were 26.6%, 41.7%, and 30.8%, respectively. Multivariate logistic regression analysis result revealed that, after adjusting for confounders such as age, sex, and education level, longer working hours were associated with higher risks of occupational stress, fatigue accumulation, and depressive symptoms (all P<0.05). Shift workers in community medical centers had higher risks of occupational stress, fatigue accumulation, and depressive symptoms compared with non-shift workers (all P<0.05). Conclusion Long working hours and shift work could increase the risks of occupational stress, fatigue accumulation, and depressive symptoms among community medical workers.

Objective To investigate the regulatory role of the programmed cell death protein 1 (PD-1) and its ligand programmed cell death protein ligand 1 (PD-L1) signaling on the subtypes and functions of natural killer (NK) cells at the maternal-fetal interface during the second trimester in mice following Toxoplasma gondii infection during the first trimester. Methods Twelve 6- to 8-week-old female mice of the C57BL/6J strain were divided into a control group and an infection group, of 6 mice in each group. On the 6.5th day of pregnancy (Gd6.5), each pregnant mouse in the infection group was intraperitoneally injected with 150 tachyzoites of the Toxoplasma gondii PRU strain, while mice in the control group were injected with an equal volume of physiological saline. On the 12.5th day of pregnancy (Gd12.5), uterus and placenta tissues were sampled from pregnant mice for pathological observations, and the mRNA expression levels of PD-1, PD-L1, and tumor necrosis factor-α (TNF-α) were quantified in uterus and placenta tissues. The PD-1 and DX5 expression was measured on NK cells at the maternal-fetal interface using flow cytometry. In addition, the in vitro JEG-3 trophoblast cells and NK-92MI cells co-culture system was established as the control group, and the addition of T. gondii tachyzoites in the co-culture system served as the infection group. The PD-1, PD-L1, and DX5 mRNA expression was quantified in cells using real-time fluorescence quantitative reverse transcription PCR (RT-qPCR) assay, and the TNF-α concentration was measured in the cell culture supernatant using enzyme-linked immunosorbent assay (ELISA). Results On Gd12.5, clear and intact cellular structures of placental decidual tissues were seen in pregnant mice in the control group, with no remarkable abnormal changes found in the uterine columnar epithelial cells, and inflammatory cell infiltration and blood stasis at varying degrees were found in uterine and placental tissues from pregnant mice in the infection group. The relative PD-1, PD-L1, and TNF-α mRNA expression was (1.004 ± 0.004), (1.001 ± 0.001), and (1.001 ± 0.001) in uterine tissues from pregnant mice in the control group and (2.480 ± 0.720), (3.355 ± 0.920), and (2.391 ± 0.073) in the infection group, respectively. The relative PD-1, PD-L1, and TNF-α mRNA expression was (1.007 ± 0.010), (1.006 ± 0.006), and (1.001 ± 0.001) in the uterine tissues in the control group and (6.948 ± 1.918), (3.225 ± 1.034), and (1.536 ± 0.150) in the infection group, respectively. The relative PD-1, PD-L1, and TNF-α mRNA expression was higher in both the uterine (t = 3.55, 4.43 and 33.02, all P values < 0.05) and placental tissues (t = 5.36, 3.72 and 6.18, all P values < 0.05) in the infection group than in the control group. Flow cytometry showed that the proportions of PD-1⁺ NK cells, PD-1⁺ DX5⁺ NK cells, and DX5⁺ NK cells were (12.200 ± 1.082)%, (9.373 ± 7.728)%, and (44.000 ± 4.095)% in uterine tissues from pregnant mice in the control group, and (21.733 ± 1.630)%, (18.767 ± 1.242)%, and (73.367 ± 0.611)% in the infection group, respectively. The proportions of PD-1⁺ NK cells, PD-1⁺ DX5⁺ NK cells, and DX5⁺ NK cells were (1.100 ± 0.510)%, (2.277 ± 1.337)%, and (96.167 ± 2.831)% in placental tissues from mice in the control group, and (26.867 ± 9.722)%, (23.433 ± 6.983)%, and (82.467 ± 2.248)% in the infection group, respectively. The proportions of PD-1⁺ NK cells (t = 8.45, P < 0.05) and DX5⁺ NK cells (t = 12.29, P < 0.05) were higher in uterine tissues from pregnant mice in the infection group than in the control group, and no significant difference was seen in the proportion of PD-1⁺ DX5⁺ NK cells (Z = -1.09, P > 0.05). The proportions of PD-1⁺ NK cells (t = 4.58, P < 0.05) and PD-1⁺ DX5⁺ NK cells (t = 5.15, P < 0.05) were higher in placental tissues from pregnant mice in the infection group than in the control group, while the proportion of DX5⁺ NK cells was lower in the infection group than in the control group (t = -6.56, P < 0.05). RT-qPCR assay revealed that the relative PD-1, PD-L1, and DX5 mRNA expression was (1.010 ± 0.005), (1.002 ± 0.003), and (1.001 ± 0.001) in the JEG-3 cells and NK92MI cells co-culture system and (3.638 ± 1.258), (0.397 ± 0.158), and (4.267 ± 1.750) in the control group, and ELISA measured that the TNF-α concentration was higher in the cell culture supernatant in the infection group [(22.056 ± 3.205) pg/mL] than in the control group [(12.441 ± 0.001) pg/mL] (t = 5.20, P < 0.05). The PD-1(t = 3.62, P < 0.05) and DX5 mRNA expression (t = 3.23, P < 0.05) was higher in the infection group than in the control group, and the PD-L1 mRNA expression was lower in the infection group than in the control group (t = -6.63, P < 0.05). Conclusions Following T. gondii infection, both PD-L1 expression and PD-1 expression on DX5⁺ NK cells at the maternal-fetal interface are upregulated in mice during the second trimester; however, the proportion of DX5⁺ NK cells decreases. These findings suggest that PD-1/PD-L1 signaling may suppress NK cell functions by modulating DX5⁺ NK cell subsets.

Delirium is a common cause and complication of hospitalization in the elderly and is associated with higher risk of future dementia and progression of existing dementia, of which 70% is Alzheimer's disease (AD). AD and delirium, which are known to be aggravated by one another, represent significant societal challenges, especially in light of the absence of effective treatments. The intricate biological mechanisms have led to numerous clinical trial setbacks and likely contribute to the limited efficacy of existing therapeutics. Artificial intelligence (AI) presents a promising avenue for overcoming these hurdles by deploying algorithms to uncover hidden patterns across diverse data types. This review explores the pivotal role of AI in revolutionizing drug discovery for AD and delirium from target identification to the development of small molecule and protein-based therapies. Recent advances in deep learning, particularly in accurate protein structure prediction, are facilitating novel approaches to drug design and expediting the discovery pipeline for biological and small molecule therapeutics. This review concludes with an appraisal of current achievements and limitations, and touches on prospects for the use of AI in advancing drug discovery in AD and delirium, emphasizing its transformative potential in addressing these two and possibly other neurodegenerative conditions.

Objective To prepare an ophthalmic solution of dihydromyricetin(DMY)based on dipotassium glycyr-rhizinate(DG)nanomicelle solubilization(DG-DMY)and evaluate its effect on dry eyes of mice.Methods DG-DMY was prepared using the thin-film hydration method,and its micelle size,potential,encapsulation efficiency and storage sta-bility at room temperature were tested.The ocular safety of DG-DMY was tested on mice.Dry eye models were built in mice,which were divided into normal control group(normal mice without intervention),PBS control group(dry eye mouse models,intervened by PBS),HA treatment group[dry eye mouse models,intervened by 1 g·L-1 hyaluronic acid(HA)]and DG-DMY treatment group(dry eye mouse models,intervened by DG-DMY),with 10 mice in each group.The fluorescein sodium staining of corneal epithelium and surface tear secretion were recorded after 10 days of intervention.Morphological changes in corneal epithelium,corneal stroma and endothelial cells were monitored by hematoxylin & eosin staining.The enzyme-linked immunosorbent assay(ELISA)was adopted to measure the expression levels of interleukin-6(IL-6)and interleukin-1β(IL-1β).Results DG-DMY is a light yellow,transparent solution with a nanomicelle size of(208.8±3.9)nm,polydispersity index of 0.277,Zeta potential of-(17.6±1.42)mV,encapsulation efficiency of 76.72％,and drug loading efficiency of 10.21％.It is stable at room temperature(25℃)and storage temperature(4 ℃).The mouse studies showed that DG-DMY displayed good in vivo tolerance in mice eyes.The therapeutic results showed that mice in the PBS treatment group still had extensive corneal staining,mice in the HA treatment group had reduced corneal staining,and mice in the DG-DMY treatment group had almost no corneal staining.The tear secretion of mice in the normal control group,PBS control group,HA treatment group and DG-DMY treatment group was(5.15±0.47)mm,(2.26±0.41)mm,(4.02±0.53)mm,and(4.11±0.54)mm.The histopathological results showed that the corneal epithelium,loose collagen structure and basal layer were damaged in the PBS control group;the corneal histopathological injury of mice in the HA treatment group and DG-DMY treatment group were mitigated,with normal corneal epithelium,corneal stroma and endothelial tissues.ELISA results showed that the expression level of IL-6 in the normal control group,PBS control group,HA treatment group and DG-DMY treatment group was(22.98±0.69)ng·g-1,(108.1±6.06)ng·g-1,(56.79±4.87)ng·g-1 and(44.01±0.99)ng·g-1,respectively,and the expression level of IL-1β was(27.97±2.74)ng·g-1,(115.70±5.16)ng·g-1,(50.36±1.56)ng·g-1 and(42.21±1.46)ng·g-1,respectively.Compared with the HA treatment group,the expression levels of IL-6 and IL-1β in the cornea of mice in the DG-DMY treatment group were lower,and the differences were statistically significant(both P＜0.05).Conclusion DG-DMY nano-preparation successfully prepared in this study is verified to act on benzalkonium chloride-induced dry eye effectively and control the inflammatory response of dry eye mouse models by inhibiting the expressions of IL-6 and IL-1β with high safety.

Objective To develop deep learning models for colonoscopy quality control using various deep learning architectures,and to delve into the decision-making mechanisms.Methods The colonoscopy images were selected from two datasets separately constructed by the HyperKvasir and Changshu Hospital Affiliated to Soochow University,encompassing intestines of varying degrees of cleanliness,polyps,and cecums.After image preprocessing and enhancement,transfer learning was carried out using the pre-trained models based on convolutional neural network(CNN)and Transformer.The model training adopted cross-entropy loss functions and Adam optimizer,and simultaneously implemented learning rate scheduling.To enhance model transparency,a thorough interpretability analysis was conducted using Grad-CAM,Guided Grad-CAM,and SHAP.The final model was converted to ONNX format and deployed on various equipment terminals to achieve real-time colonoscopy quality control.Results In a dataset of 3 831 colonoscopy images,EfficientNet model outperformed the other models on the test set,achieving an accuracy of 0.992 which was higher than those of the other models based on CNN(DenseNet121,ResNet50,VGG19)and Transformer(ViT,Swin,CvT),with a precision,recall rate,and F1 score of 0.991,0.989,and 0.990.On an external test set of 358 images,EfficientNet model had an average AUC,precision,and recall rate of 0.996,0.948,and 0.952,respectively.Although EfficientNet model is high-performing,some misjudgments still occurred.Interpretability analysis highlighted key image areas affecting decision-making.In addition,EfficientNet model was successfully converted to ONNX format and deployed on multiple platforms and devices,and it ensured real-time colonoscopy quality control with an inference speed of over 60 frames per second.Conclusion Among the 7 models developed for colonoscopy quality control based on CNN and Transformer,EfficientNet demonstrated exemplary performance across all categories and is deployed for real-time predictions on multiple terminals,aiming to provide patients with better medical care.

Objective:To compare with video-fluoroscopic swallowing function test(VFSS),so as to evaluate the application value of ultrasonography in the test of dysphagia after stroke.Methods:A total of seventy-two patients with dysphagia after stroke who admitted to The Second Affiliated Hospital of Hainan Medical University from January 2022 to July 2023 were selected as cases group,and a total of 45 healthy aged with normal swallowing function were selected as healthy control group at the same time.All of them underwent video X-ray fluoroscopic examination of swallowing function and quantitative assessment of ultrasound.Quantitative assessment of ultrasound was performed by twice to compare the internal consistency of ultrasound test.Meanwhile,the correlation between ultrasound and VFSS results was tested to verify the validity of ultrasound quantitative assessment.The VFSS was used as gold standard to analyze the sensitivity and specificity of ultrasound assessment.The differences of abnormal grade of Geniohyoid muscle;movement time,movement distance and the change of tongue muscle thickness of semi quantitative description of ultrasound between two groups were compared.Results:The intra group correlation coefficient(ICC),movement time,movement distance and movement speed of grade description of muscle abnormalities of cases group were respectively 0.90,(1.743±0.235)s,(6.323±0.823)mm and(3.826±0.778)mm/s,which intra reliabilities were high correlation(ICC=0.90,0.82,0.87,0.85,P＜0.01),which inter reliabilities of these indicators were high correlation(ICC=0.86,0.85,0.88,0.87,P＜0.01),respectively.The positive results of ultrasound test highly correlated with the results of VFSS examination(r=0.91,P＜0.01).The movement distance,the average movement speed and the changes of the thickness of tongue muscle in patients of cases group were significantly smaller than those of healthy control group,but the movement time was larger than that of healthy control group(t=9.03,30.49,-7.02,22.69,P＜0.05),respectively.The results of the ultrasound on the muscles of all patients in cases group existed abnormality.Conclusion:Ultrasound technique can quantitatively assess dysphagia after stroke.Compared with VFSS technique,ultrasonography can measure and determine the related data of the movement of swallowing muscle,and dynamically record movement parameters of geniohyoid muscle and the changes of the thickness of tongue muscle.At the same time,it can detect the grade of muscle abnormalities of patients with dysphagia,which will contribute to further explore the potential pathological mechanism about muscle,and promote the healthy management for patients with dysphagia.

Objective:To investigate the relationship between key molecules of interleukin-6/signal transducer and activator of transcription 3 (IL-6/STAT3) signaling pathway and breast cancer susceptibility.Methods:A case-control study design was adopted. 136 patients with breast cancer admitted to Department of Breast Surgery, Yantai Yantaishan Hospital from Mar. 2021 to Mar. 2023 were selected as the case group, and 136 healthy subjects of the same age were matched as the control group in a 1:1 ratio. The clinical data and key molecules of IL-6/STAT3 signaling pathway were compared between the two groups. The risk factors of breast cancer susceptibility were evaluated by conditional Logistic regression and addition analysis, and the diagnostic efficiency of receiver operating characteristic curve (ROC) and area under curve (AUC) were plotted.Results:The body mass index (BMI) (22.21±0.68 vs. 21.30±0.70 kg/m 2), age at menarche (13.50±1.24 years vs. 14.83±1.42 years), IL-6 (3.50±1.05 vs. 2.41±0.74), Janus protein kinase 1 (JAK1) mRNA (1.23±0.36 vs. 0.88±0.26), STAT3 mRNA (1.68±0.50 vs. 1.17±0.35), and menopause (30.15% vs. 55.88%), breastfeeding (82.35% vs. 92.65%), physical exercise (44.12% vs. 58.09%) were significantly different between the case group and the control group ( t=10.87, 8.23, 9.89, 9.19, 9.75, χ2=18.37, 6.59, 5.31, P<0.05). Breastfeeding [odd ratio ( OR) : 0.550], interleukin-6 (IL-6) ( OR: 4.409), janus protein kinase 1 (JAK1) ( OR: 5.370) and signal transducer and activator of transcription 3 (STAT3) ( OR: 4.386) mRNAs were risk factors for breast cancer susceptibility ( P<0.05). ROC curve showed that the combined diagnostic efficacy of IL-6, JAK1, STAT3 mRNA in breast cancer was significantly better than that of a single diagnostic efficacy. The incidence rate of breast cancer in those who were exposed to IL-6 mRNA and breast feeding was 3.508 times higher than that in those who were not exposed. The incidence rate of breast cancer in those who were exposed to JAK1 mRNA and breast feeding at the same time was 4.136 times higher than that in those who were not exposed. The incidence rate of breast cancer in those who were exposed to STAT3 mRNA and breast feeding was 3.537 times higher than that in those who were not exposed. Conclusion:The key molecules of IL-6/STAT3 signaling pathway and breast-feeding are predisposing factors of breast cancer, and they have additive interaction, thus increasing the susceptibility of breast cancer, which has significant practical significance for the differential diagnosis and treatment of this disease.

Objective·To compare the performance of various deep learning methods in the segmentation and classification of colorectal polyp endoscopic images,and identify the most effective approach.Methods·Four colorectal polyp datasets were collected from three hospitals,encompassing 1 534 static images and 15 videos.All samples were pathologically validated and categorized into two types:serrated lesions and adenomatous polyps.Polygonal annotations were performed by using the LabelMe tool,and the annotated results were converted into integer mask formats.These data were utilized to train various architectures of deep neural networks,including convolutional neural network(CNN),Transformers,and their fusion,aiming to develop an effective semantic segmentation model.Multiple performance indicators for automatic diagnosis of colon polyps by different architecture models were compared,including mIoU,aAcc,mAcc,mDice,mFscore,mPrecision and mRecall.Results·Four different architectures of semantic segmentation models were developed,including two deep CNN architectures(Fast-SCNN and DeepLabV3plus),one Transformer architecture(Segformer),and one hybrid architecture(KNet).In a comprehensive performance evaluation of 291 test images,KNet achieved the highest mIoU of 84.59％,significantly surpassing Fast-SCNN(75.32％),DeepLabV3plus(78.63％),and Segformer(80.17％).Across the categories of"background","serrated lesions"and"adenomatous polyps",KNet's intersection over union(IoU)were 98.91％,74.12％,and 80.73％,respectively,all exceeding other models.Additionally,KNet performed excellently in key performance metrics,with aAcc,mAcc,mDice,mFscore,and mRecall reaching 98.59％,91.24％,91.31％,91.31％,and 91.24％,respectively,all superior to other models.Although its mPrecision of 91.46％was not the most outstanding,KNet's overall performance remained leading.In inference testing on 80 external test images,KNet maintained an mIoU of 81.53％,demonstrating strong generalization capabilities.Conclusion·The semantic segmentation model of colorectal polyp endoscopic images constructed by deep neural network based on KNet hybrid architecture,exhibits superior predictive performance,demonstrating its potential as an efficient tool for detecting colorectal polyps.

Objective To invesigate the feasibility of low radiation dose and low contrast dosage in triple-rule-out computed tomo-graphy angiography(TRO-CTA)on the 320-row detector CT.Methods A total of 120 patients who underwent CTA were prospec-tively selected.All patients were divided into control group(n=90)and experimental group(n=30).The control group employed standard-doses protocol of pulmonary CTA(120 kV tube voltage,45 mL contrast dosage),coronary CTA(120 kV,50-60 mL),and aortic CTA(120 kV,75 mL),while the experimental group received TRO-CTA with 100 kV and 70-80 mL.The peak time of contrast dosage at the pulmonary artery and aorta was measured by low-dose detection method in the experimental group,and the contrast examination was performed sequentially in the control group.Subjective scores and objective image quality of the pulmonary artery,coronary artery,and aorta in the experimental group and the control group were measured and compared,respectively.The effective dose(ED)between the two groups were recorded and compared.Independent samples t-test and Fisher exact probability were used to analyze the statistical differences between the above measures.Results There were no significant differences in CT values,con-trast-to-noise ratio(CNR),signal-to-noise ratio(SNR)of pulmonary artery,coronary artery and aorta between the two groups(P＞0.05).The mean subjective scores of pulmonary artery,coronary artery and aorta segments in the two groups were not less than 3 points,meeting the requirement of clinical diagnosis.There was no statistical difference in subjective scores between the two groups(P＞0.05).There was statistically significant difference in ED between the two groups(P＜0.05).The ED of pulmonary artery,coronary artery,and aorta in the experimental group were 11.49％,13.33％,and 11.46％significantly lower than those in the control group,respec-tively.Conclusion It is feasible to obtain TRO-CTA images used by the low radiation dose and low contrast dosage on the 320-row detector CT,and radiation dose and contrast dosage can be reduced reasonably without alterations of TRO-CTA images quality in clinical practice.