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: 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.

Glaucoma is a group of optic nerve disorders characterized by progressive optic nerve atrophy and visual field defects, which can lead to irreversible blindness. Early diagnosis of glaucoma is essential for preventing visual loss. However, due to the absence of obvious early symptoms, the diagnosis of glaucoma remains challenging. Visual electrophysiological examinations, an objective approach for evaluating visual function, have the potential to be used in the early diagnosis of glaucoma. This review integrates the latest publications to introduce visual electrophysiological examination techniques, including electroretinography(ERG)and visual evoked potential(VEP). It also explores the mechanisms underlying these techniques and their application value in the early diagnosis of glaucoma. In addition, this review summarizes the advantages, limitations, and applicable scenarios of different visual electrophysiological techniques. Finally, the review provides an outlook on the development prospects of visual electrophysiological techniques in the early diagnosis of glaucoma. The findings of this review can assist clinicians in selecting appropriate diagnostic methods, promote the innovation and development of early visual electrophysiological diagnostic techniques for glaucoma, and contribute to reducing the risk of blindness caused by glaucoma.

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.

Optical imaging is highly valued for its superior temporal and spatial resolution. This is particularly important in near-infrared II (NIR-II, 1 000-3 000 nm) imaging, which offers advantages such as reduced tissue absorption, minimal scattering, and low autofluorescence. These characteristics make NIR-II imaging especially suitable for deep tissue visualization, where high contrast and minimal background interference are critical for accurate diagnosis and monitoring. Currently, inorganic fluorescent probes—such as carbon nanotubes, rare earth nanoparticles, and quantum dots—offer high brightness and stability. However, they are hindered by ambiguous structures, larger sizes, and potential accumulation toxicity in vivo. In contrast, organic fluorescent probes, including small molecules and polymers, demonstrate higher biocompatibility but are limited by shorter emission wavelengths, lower quantum yields, and reduced stability. Recently, gold clusters have emerged as a promising class of nanomaterials with potential applications in biocatalysis, fluorescence sensing, biological imaging, and more. Water-soluble gold clusters are particularly attractive as fluorescent probes due to their remarkable optical properties, including strong photoluminescence, large Stokes shifts, and excellent photostability. Furthermore, their outstanding biocompatibility—attributed to good aqueous stability, ultra-small hydrodynamic size, and high renal clearance efficiency—makes them especially suitable for biomedical applications. Gold clusters hold significant potential for NIR-II fluorescence imaging. Atomic-precision gold clusters, typically composed of tens to hundreds of gold atoms and measuring only a few nanometers in diameter, possess well-defined three-dimensional structures and clear spatial coordination. This atomic-level precision enables fine-tuned structural regulation, further enhancing their fluorescence properties. Variations in cluster size, surface ligands, and alloying elements can result in distinct physicochemical characteristics. The incorporation of different atoms can modulate the atomic and electronic structures of gold clusters, while diverse ligands can influence surface polarity and steric hindrance. As such, strategies like alloying and ligand engineering are effective in enhancing both fluorescence and catalytic performance, thereby meeting a broader range of clinical needs. In recent years, gold clusters have attracted growing attention in the biomedical field. Their application in NIR-II imaging has led to significant progress in vascular, organ, and tumor imaging. The resulting high-resolution, high signal-to-noise imaging provides powerful tools for clinical diagnostics. Moreover, biologically active gold clusters can aid in drug delivery and disease diagnosis and treatment, offering new opportunities for clinical therapeutics. Despite the notable achievements in fundamental research and clinical translation, further studies are required to address challenges related to the standardized synthesis and complex metabolic behavior of gold clusters. Resolving these issues will help accelerate their clinical adoption and broaden their biomedical applications.

Aim To explore the effects of Huannao Yicong decoction(HYD)on the learning and memory ability and brain iron metabolism in APP/PS1 mice and the correlation of HAMP knockout mice and APP/PS1 double transgenic model mice.Methods The ex-periment was divided into five groups,namely,HAMP-/-group(6-month HAMP gene knockout mice),APP/PS1 group(6-month APP/PS1-double-transgenic mice),HAMP-/-+HYD,APP/PS1+HYD,and negative control group(6-month C57BL/6J mice),with six mice in each group.The dose was ad-ministered(13.68 g·kg-1 weight),and the other groups received distilled water for gavage once a day for two months.After the administration of the drug,the mice in each group were tested for learning and memory in the Morris water maze;Biochemical detec-tion was performed to detect iron ion content in each mouse brain;Western blot and RT-qPCR were carried out to analyze hippocampal transferrin(TF),transfer-rin receptor1(TFR1),membrane iron transporter1(FPN1)divalent metal ion transporter 1(DMT1)and β-amyloid protein(Aβ)protein and mRNA expression levels in each group.Results Compared with the normal group,both HAMP-/-mice and APP/PS1 mice had reduced the learning and memory capacity,in-creased iron content in brain tissue,Aβ protein ex-pression increased in hippocampus of HAMP-/-group and APP/PS1 group mice(P＜0.01),the protein and mRNA expression of TF,TFR1 and DMT1 increased in hippocampal tissues of HAMP-/-and APP/PS1 groups(P＜0.01),and the FPN1 protein and mRNA expres-sion decreased(P＜0.01).Compared with the HAMP-and APP/PS1 groups,respectively,HAMP-/-+HYD group and APP/PS1+HYD group had improved learning and memory ability,decreased iron content,decreased Aβ protein expression(P＜0.01),decreased TF,TFR1,DMT1 protein and mR-NA expression(P＜0.01),and increased expression of FPN1 protein and mRNA(P＜0.01).Conclusions There is some association between HAMP-/-mice and APP/PS1 mice,HYD can improve the learning and memory ability of HAMP-/-and APP/PS1 mice and reduce the Aβ deposition.The mechanism may be related to the regulation of TF,TFR1,DMT1,FPN1 expression and improving brain iron overload.

Objective To explore the effect of Kümmell's disease with kyphosis on the sagittal morphology of the spine-pelvis.Methods A retrospective analysis of 34 patients of Kümmell's disease with kyphosis(Kümmell group)admitted from August 2015 to September 2022,including 10 males and 24 females with an average age of(71.1±8.5)years old.A control group of 37 asymptomatic population aged(69.3±6.7)years old was matched.Spinal-pelvic sagittal parameters were measured on the anterior-posterior and lateral X-rays of the whole spine in the standing position,including segmental kyphosis(SK)or thoracolumbar kyphosis(TLK),thoracic kyphosis(TK),lumbar lordosis(LL),pelvic incidence(PI),pelvic tilt(PT),sacral slope(SS),sagittal vertical axis(SVA),T1 pelvic angle(TPA)and PI-LL.Vertebral wedge angle(WA)in Kümmell was mea-sured and differences in parameters among groups were analyzed and the relationship between spino-pelvic parameters and WA,SK were also investigated.Results TK,SK,PT,SVA,TPA and PI-LL in Kümmell group were significantly larger than those in control group(P＜0.05),LL and SS in Kümmell group were significantly decreased than those in control group(P＜0.05),and there was no significant difference in PI between two groups(P＞0.05).In Kümmell group,WA(30.8±5.9)° showed a positive correlation with SK and TK(r=0.366,0.597,P＜0.05),and SK was significantly correlated with LL and SS(r=0.539,-0.591,P＜0.05).Strong positive correlation between LL and PI,SS,SVA,TPA,PI-LL were also confirmed in patients with Kümmell with kyphosis(r=0.559,0.741,-0.273,-0.356,-0.882,P＜0.05).Conclusion Patients with Kümmell with kyphosis not only have segmental kyphosis,but also changes the overall spinal-pelvic sagittal parameters,including loss of lumbar lordosis,pelvic retrorotation,trunk forward tilt.The surgical treatment of Kümmell disease should not only pay attention to the recovery of the height of the collapsed vertebra,but also focus on the overall balance of the spine-pelvic sagittal plane for patients with kyphosis.

Objective To explore the risk factors related to afterload mismatch(AM)after transcatheter mitral valve repair(MitraClip).Methods This was a retrospective cohort study.48 patients hospitalized in the Department of Cardiovascular Medicine,the First Affiliated Hospital of Sun Yat-sen University from December 2021 to December 2023,who underwent MitraClip due to severe mitral regurgitation(MR)were included.Preoperative clinical data,laboratory tests,and preoperative and postoperative color Doppler echocardiographic examination results of surgical patients were collected.AM was defined as the left ventricular ejection fraction(LVEF)decreased by 15％or more after surgery compared with the one before(dLVEF≤-15％).Patients were divided into AM group and non-AM group according to whether afterload mismatch occurred.Univariate and multivariate logistic regression were used to analyze the risk factors of postoperative AM.Results Among 48 patients who underwent MitraClip,14 of them(29.2％)developed afterload-mismatched.For those without AM,their overall LVEF was improved after the operation;for patients in both AM group and non-AM group,their overall left ventricular end-diastolic diameter(LVEDd),left ventricular end-diastolic diameter volume index(LVEDVi)was reduced compared with the preoperative ones.Univariate regression analysis showed that C-reactive protein levels(OR 1.98,95％CI 1.02-3.83),platelets(OR 2.22,95％CI 1.08-4.53),systemic immune inflammation index(OR 1.96,95％CI 1.03-3.71)were associated with an increased risk of AM in patients undergoing MitraClip(all P＜0.05),while those with larger right atrial diameter(OR 0.35,95％CI 0.13-0.93)or moderate to severe tricuspid regurgitation(OR 0.19,95％CI 0.05-0.81)were less likely to develop into AM(both P＜0.05),which is still satisfied after adjustment.Conclusions For patients who underwent MitraClip,C-reactive protein levels,platelets and systemic immune inflammation index(SII)are associated with an increased risk of afterload mismatched,while those with larger right atrial diameter or moderate to severe tricuspid regurgitation were less likely to develop into AM.