Objective To screen ferroptosis genes related to the prognosis of cervical cancer and to construct a prognosis model. Methods Ferroptosis genes were obtained from FerrDb database, and cervical cancer related data were obtained from The Genome-Wide Association Study Catalog database and The Cancer Genome Atlas database. Transcriptome-Wide Association Study, colocalization analysis and differential expression analysis were conducted to screen out candidate ferroptosis genes; Gene Ontology functional and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis were conducted on candidate genes. Univariate Cox regression analysis was used to further screen out genes related to the prognosis of cervical cancer. Kaplan-Meier method was used to analyze the relationship between genes and the overall survival of patients. The expression levels of genes in pan-cancer were analyzed through the TIMER database. Two prognostic models were conducted, Model 1 included age and tumor stage, while Model 2 incorporated age, tumor stage, and prognostic genes. The predictive capabilities of the two models were compared. Results A total of 91 candidate genes related to ferroptosis were obtained. Univariate Cox regression analysis showed that 15 genes were associated with the prognosis of cervical cancer. CA9, SCD, TFRC, QSOX1 and CDO1 were risk factors affecting the prognosis of cervical cancer patients (P＜0.05), while PTPN6, ALOXE3, HELLS, IFNG, MIOX, ALOX12B, DUOX1, ALOX15, AQP3 and IDO1 were protective factors (P＜0.05). The mRNA expression levels of the 15 genes showed significant upregulation or downregulation in at least 7 types of cancers, among which TFRC was associated with the largest number of cancer types. Kaplan-Meier analysis showed that HELLS, DUOX1 and ALOXE3 were associated with poor prognosis in cervical cancer. The AUC of the model 1 for predicting 1-year and 3-year overall survival rates of cervical cancer patients was 0.455 and 0.478, and the AUC of Model 2 was 0.854 and 0.595. Model 2 (C-index = 0.727) had better predictive ability than Model 1 (C-index = 0.502). Conclusion The prognostic model composed of 15 prognostic-related genes selected based on bioinformatics has better predictive performance for the survival outcomes of cervical cancer patients, providing important reference value for the prognostic assessment of cervical cancer patients.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective: To investigate the prevalence and associated factors of spinal curvature abnormalities among primary and secondary school students in the Guangxi Zhuang Autonomous Region, so as to provide a scientific basis for the prevention and control of such abnormalities. Methods: From September to November 2023, adopting a stratified cluster random sampling method, spinal curvature screenings and questionnaire surveys were conducted among 168 931 students from grade 4 of primary school to grade 12 of high school in 111 districts and counties across 14 cities in Guangxi. Chi square tests and binary Logistic regression analysis were used to analyze influencing factors of spinal curvature abnormalities. Results: In 2023, the detection rate of poor posture among students above grade 4 in Guangxi was 4.24% , and the detection rate of spinal curvature abnormalities was 2.13%. The detection rate was higher among urban students (2.84%) than rural students (1.66%), boarding students (2.61%) than non-boarding students (1.60%), and high school students (3.16%) than junior high (2.45%) and primary school students (1.15%), and the differences were statistically significant ( χ 2=269.85, 221.44, 565.10, P <0.01). A trend of increasing detection rates with higher grade levels was observed ( χ 2 trend =617.63, P <0.01). Binary Logistic regression analysis indicated that students without boarding at school ( OR =0.82, 95% CI =0.75-0.90), engaging in high-intensity physical activity for over 60 min per day ≥5 days per week ( OR =0.90, 95% CI =0.82-0.98), and adequate sleep ( OR =0.87, 95% CI =0.81-0.94) had lower risks of detecting spinal curvature abnormalities ( P <0.05). Conclusions The prevalence of spinal curvature abnormalities increases with grade level among primary and secondary school students in Guangxi. Regular moderate-to-vigorous physical activity demonstrates protective effects against spinal abnormalities.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

Objective: Machine learning (ML) has been reported to have better predictive capability than traditional statistical techniques. The aim of this study was to assess the efficacy of ML algorithms and logistic regression (LR) for predicting depressive symptoms during the COVID-19 pandemic. Methods: Analyses were carried out in a national cross-sectional study involving 21,916 participants. The ML algorithms in this study included random forest (RF), support vector machine (SVM), neural network (NN), and gradient boosting machine (GBM) methods. The performance indices were sensitivity, specificity, accuracy, precision, F1-score, and area under the receiver operating characteristic curve (AUC). Results: LR and NN had the best performance in terms of AUCs. The risk of overfitting was found to be negligible for most ML models except for RF, and GBM obtained the highest sensitivity, specificity, accuracy, precision, and F1-score. Therefore, LR, NN, and GBM models ranked among the best models. Conclusion Compared with ML models, LR model performed comparably to ML models in predicting depressive symptoms and identifying potential risk factors while also exhibiting a lower risk of overfitting.

ObjectiveTo investigate the diagnostic accuracy and grading capability of ultrasound-derived fat fraction （UDFF）， controlled attenuation parameter （CAP）， and hepatic/renal ratio （HRR） in assessing hepatic steatosis in metabolic associated fatty liver disease （MAFLD） with magnetic resonance imaging-proton density fat fraction （MRI-PDFF） as the gold standard. MethodsA total of 150 patients with MAFLD who attended The First Hospital of Hebei Medical University from January 2023 to December 2024 were enrolled， and 148 healthy volunteers were recruited. All subjects underwent MRI-PDFF， UDFF， CAP， and HRR examinations. Hepatic steatosis was graded based on MRI-PDFF （S0：148 cases； S1：92 cases； S2：21 cases； S3：37 cases）， and the MAFLD patients with different grades of hepatic steatosis were compared in terms of UDFF， CAP， HRR， and clinical features. A one-way analysis of variance was used for comparison of normally distributed continuous data between multiple groups and the Tukey HSD test was used for further comparision between two groups； the Kruskal-Wallis H test was used for comparison of non-normally distributed continuous data between multiple groups， and the Mann-Whitney U test was used for further comparison between two groups； the chi-square test was used for comparison of categorical data between groups. The Spearman correlation analysis was used to investigate the correlation between UDFF， CAP， HRR， and MRI-PDFF in different grades of MAFLD； the receiver operating characteristic （ROC） curve was used to investigate the efficacy of UDFF， CAP， and HRR in the diagnosis of different degrees of hepatic steatosis in MAFLD； the Bland-Altman difference plot was used to analyze the consistency between UDFF and MRI-PDFF in different degrees of hepatic steatosis in MAFLD. ResultsUDFF measurement gradually increased with the increase in the grade of fatty liver （H=201.52，P0.001）. The Spearman correlation analysis showed that there was a strong correlation between any two indicators of UDFF， CAP， HRR， and MRI-PDFF in S1， S2， and S3 MAFLD （all P0.001）， with the strongest correlation between UDFF and MRI-PDFF （r_s1=0.884，r_s2=0.962，r_s3=0.929， all P0.001）. The ROC curve analysis showed that UDFF had a larger area under the ROC curve （AUC） than CAP and HRR in the graded diagnosis of S1 and S3 （all P0.05）， while in the diagnosis of S2 MAFLD， UDFF had a significantly larger AUC than HRR （P0.05） and a similar AUC to CAP （P0.05）. The Bland-Altman difference plot showed good consistency between UDFF and MRI-PDFF in different degrees of hepatic steatosis in MAFLD. ConclusionCompared with CAP and HRR， UDFF can accurately measure liver fat content and has good efficacy in identifying varying degrees of hepatic steatosis in MAFLD.

OBJECTIVE To investigate the effects of the peroxisome proliferator-activated receptors δ （PPARδ） agonist GW501516 on the injury of pulmonary artery endothelial cells （PAECs） induced by hypoxia and its mechanism. METHODS The cytotoxic effects of GW501516 were observed by detecting the relative survival rate of PAECs； the protein expression of PPARδ was determined by Western blot assay. The cellular model of PAECs injury was established under hypoxic conditions； using antioxidant N-acetylcysteine （NAC） as positive control， the effects of GW501516 on cell injury and reactive oxygen species （ROS） production were investigated by detecting cell apoptotic rate， cell viability， lactate dehydrogenase （LDH） activity and ROS levels. Using nuclear factor erythroid 2-related factor 2（Nrf2） activator dimethyl fumarate （DMF） as positive control， PAECs were incubated with GW501516 and/or Nrf2 inhibitor ML385 under hypoxic conditions； the mechanism of GW501516 on PAECs injury induced by hypoxia was investigated by detecting cell injury （cell apoptosis， cell viability， LDH activity）， the levels of superoxide dismutase （SOD）， glutathione peroxidase （GPx）， catalase （CAT）， malondialdehyde （MDA） and ROS， the expressions of Nrf2， heme oxygenase-1 （HO-1） and cleaved-caspase-3 （C-caspase-3） protein. RESULTS The results demonstrated that hypoxia inhibited the protein expression of PPARδ （P＜0.05）， while GW501516 promoted the protein expression of PPARδ in hypoxia- exposed PAECs without obvious cytotoxic effects. GW501516 inhibited the apoptosis of PAECs， improved cell viability， and reduced LDH activity and ROS levels. GW501516 could up-regulate the protein expression of HO-1 in PAECs and the levels of SOD， GPx and CAT， while down-regulated the levels of MDA and ROS by activating the Nrf2 pathway （P＜0.05）； but Nrf2 inhibitor ML385 could reverse the above effects of GW501516 （P＜0.05）. GW501516 exerted similar effects to Nrf2 activator DMF in down-regulating the expression of C-caspase-3 and inhibiting the injury of PAECs under conditions of hypoxia （P＜0.05）. Moreover， Nrf2 inhibitor ML385 reversed the 163.com inhibition effects of GW501516 on PAECs injury （P＜0.05）. CONCLUSIONS GW501516 can relieve the hypoxia-induced injury of PAECs via the inhibition of oxidative stress， the mechanism of which may be associated with activating Nrf2.

Objective To analyze the literature on artificial intelligence in forensic research from 2012 to 2022 in the Web of Science Core Collection Database,to explore research hotspots and developmen-tal trends.Methods A total of 736 articles on artificial intelligence in forensic medicine in the Web of Science Core Collection Database from 2012 to 2022 were visualized and analyzed through the litera-ture measuring tool CiteSpace.The authors,institution,country(region),title,journal,keywords,cited references and other information of relevant literatures were analyzed.Results A total of 736 articles published in 220 journals by 355 authors from 289 institutions in 69 countries(regions)were identi-fied,with the number of articles published showing an increasing trend year by year.Among them,the United States had the highest number of publications and China ranked the second.Academy of Forensic Science had the highest number of publications among the institutions.Forensic Science Inter-national,Journal of Forensic Sciences,International Journal of Legal Medicine ranked high in publica-tion and citation frequency.Through the analysis of keywords,it was found that the research hotspots of artificial intelligence in the forensic field mainly focused on the use of artificial intelligence technol-ogy for sex and age estimation,cause of death analysis,postmortem interval estimation,individual identification and so on.Conclusion It is necessary to pay attention to international and institutional cooperation and to strengthen the cross-disciplinary research.Exploring the combination of advanced ar-tificial intelligence technologies with forensic research will be a hotspot and direction for future re-search.