Objective:To develop a deep learning-based segmentation model for anterior segment ultrasound biomicroscopy (UBM) images to automatically segment the anterior segment tissues of patients with primary angle-closure glaucoma (PACG).Methods:A single-center retrospective case series was conducted.A small-scale dataset comprised 468 UBM images of the anterior chamber angle closure from 156 patients with PACG who underwent the UBM examination at Tianjin Medical University Eye Hospital between July 12, 2022, and February 20, 2023.The UBM images were randomly split into a training dataset of 228 images and a testing dataset of 152 images using a random seed method in a ratio of 6∶4.The models were trained using the PSPNet model with MobileNet V2 and ResNet50 as backbones, the DeepLab v3+ model with MobileNet V2 and Xception as backbones, and the SegFormer model with MiT-B0 and MiT-B2 as backbones.The testing dataset was used for result prediction and to achieve segmentation of four regions: the cornea and sclera, iris, ciliary body, and anterior lens surface.To evaluate the performance of the models in segmenting the anterior segment structures, multiple metrics were assessed, including the mean intersection over union (mIoU), Dice coefficient, precision, recall, false negative rate, and specificity.A comparative analysis of the test results across the different models was subsequently performed.This study adhered to the Declaration of Helsinki.The study protocol was approved by the Ethics Committee of Tianjin Medical University Eye Hospital (No.2023KY-05).Results:The two models with the best segmentation performance were PSPNet and DeepLab v3+ .The PSPNet model with ResNet50 as the backbone achieved the mIoU of 85.11%, Dice coefficient of 91.38%, precision of 91.83%, recall of 90.94%, false negative rate of 9.06%, and specificity of 98.89%.The DeepLab v3+ model with MobileNet V2 as the backbone achieved an mIoU of 85.84%, Dice coefficient of 92.01%, precision of 92.67%, recall of 91.36%, false negative rate of 8.64%, and specificity of 98.90%.Among the five key metrics, mIoU, Dice coefficient, recall, false negative rate, and specificity, DeepLab v3+ exhibited the best segmentation performance.In addition, the DeepLab v3+ model with Xception as the backbone had the highest precision among all models, reaching 92.77%.Conclusions:The deep learning-based DeepLab v3+ model achieves precise segmentation of anterior segment tissue structures in PACG anterior segment UBM image segmentation, providing auxiliary support for clinical diagnosis.

Objective:To develop a deep learning-based segmentation model for anterior segment ultrasound biomicroscopy (UBM) images to automatically segment the anterior segment tissues of patients with primary angle-closure glaucoma (PACG).Methods:A single-center retrospective case series was conducted.A small-scale dataset comprised 468 UBM images of the anterior chamber angle closure from 156 patients with PACG who underwent the UBM examination at Tianjin Medical University Eye Hospital between July 12, 2022, and February 20, 2023.The UBM images were randomly split into a training dataset of 228 images and a testing dataset of 152 images using a random seed method in a ratio of 6∶4.The models were trained using the PSPNet model with MobileNet V2 and ResNet50 as backbones, the DeepLab v3+ model with MobileNet V2 and Xception as backbones, and the SegFormer model with MiT-B0 and MiT-B2 as backbones.The testing dataset was used for result prediction and to achieve segmentation of four regions: the cornea and sclera, iris, ciliary body, and anterior lens surface.To evaluate the performance of the models in segmenting the anterior segment structures, multiple metrics were assessed, including the mean intersection over union (mIoU), Dice coefficient, precision, recall, false negative rate, and specificity.A comparative analysis of the test results across the different models was subsequently performed.This study adhered to the Declaration of Helsinki.The study protocol was approved by the Ethics Committee of Tianjin Medical University Eye Hospital (No.2023KY-05).Results:The two models with the best segmentation performance were PSPNet and DeepLab v3+ .The PSPNet model with ResNet50 as the backbone achieved the mIoU of 85.11%, Dice coefficient of 91.38%, precision of 91.83%, recall of 90.94%, false negative rate of 9.06%, and specificity of 98.89%.The DeepLab v3+ model with MobileNet V2 as the backbone achieved an mIoU of 85.84%, Dice coefficient of 92.01%, precision of 92.67%, recall of 91.36%, false negative rate of 8.64%, and specificity of 98.90%.Among the five key metrics, mIoU, Dice coefficient, recall, false negative rate, and specificity, DeepLab v3+ exhibited the best segmentation performance.In addition, the DeepLab v3+ model with Xception as the backbone had the highest precision among all models, reaching 92.77%.Conclusions:The deep learning-based DeepLab v3+ model achieves precise segmentation of anterior segment tissue structures in PACG anterior segment UBM image segmentation, providing auxiliary support for clinical diagnosis.

Objective:To examine the impact of Developmental Human Resource Management Practices(D-HRMPs)on innova-tive performance of health professionals from the public hospitals.Methods:A questionnaire survey was conducted to collect data from health professionals in the public hospitals in Beijing,including D-HRMPs scale,innovation performance scale,inclusive lead-ership scale and team innovation atmosphere scale.Descriptive statistics and structural equation model were used for the analysis.Results:D-HRMPs had a direct effect on the innovation performance of public hospital health professionals.Inclusive leadership can positive-ly affect the implementation of D-HRMPs and the formation of team innovation atmosphere.Meanwhile,team innovation atmo-sphere plays a mediating effect,that is,D-HRMPs can actively promote the formation of team innovation climate,and further indi-rectly affect the innovation performance of employees.Conclusion:The directors of public hospitals can further explore the im-plementation of D-HRMPs,and pay attention to adopting inclusive leadership to deepen the reform of personnel management system,and promote high-quality development of public hospitals.

Objective To investigate the clinical efficacy of arthroscopic simple repair technique and en-hanced repair technique in the treatment of chronic lateral ankle instability(CLAI).Methods Forty-one cases of CLAI treated in the General Hospital of Northern Theater Command of PLA from January 2018 to January 2019 were selected and conducted arthroscopic lateral ankle ligament repair or enhanced repair treatment.The follow up data in 39 cases were complete.Among them,18 cases conducted the anterior talofibular ligament repair with wire anchor under arthroscopy and were included in the repair group,while 21 cases conducted the arthroscopic enhanced repair with linear anchors and knots free anchors and were included in the enhanced re-pair group.The postoperative follow up lasted for 12 months.The VAS score and AOFAS ankle-hind foot score before operation and in postoperative,3,6,12 months were compared between the two groups.The effi-cacies of treating CLAI by the two operation modes were evaluated.Results The preoperative VAS score and AOFAS ankle-hind foot score had no statistical differences between the two groups(P＞0.05).The VAS score at postoperative time points had no statistical difference between the enhanced repair group and repair group(P＞0.05).The AOFAS ankle-hind foot score in the enhanced repair group was higher than that in the repair group,and the difference was statistically significant(P＜0.05).No serious complications occurred in the grouped patients during the follow up period.Conclusion In terms of pain and ankle function,the effect of arthroscopic enhanced repair in treating CLAI is better than that of simple repair.

Objective:To explore the role of 17β-estradiol(17β-E2) derived from the hippocampus and gonad in the cognitive memory during the reconsolidation period of fear memory in rats with post-traumatic stress disorder(PTSD).Methods:The 8-week-old clean grade female SD rats were used for this study. Single prolonged stress combined with context fear conditioning was used to prepare the rat PTSD model.(1) Gonad estradiol experiment: 50 female rats were randomly divided into blank control group, model group, sham ovariectomy group, ovariectomy group, and ovariectomy+ estradiol group by random number table method, with 10 rats in each group. The rats in model group, sham ovariectomy group, ovariectomy group, and ovariectomy+ estradiol group were established PTSD model. Rats in ovariectomy group and ovariectomy+ estradiol group underwent ovariectomy, while rats in sham ovariectomy group only were removed adipose tissue of the same mass around the ovaries. And the rats in ovariectomy+ estradiol group were injected with 17β-E2(1 mg/kg, once a day for 14 days) 7 days after ovariectomy.(2) Hippocampal estradiol experiment: 40 female rats were randomly divided into blank control group, model group, hippocampal solvent control group and hippocampal estradiol inhibitor group by random number table method, with 10 rats in each group.The rats in model group, hippocampal solvent control group and hippocampal estradiol inhibitor group were established PTSD model.Rats in hippocampal estradiol inhibitor group were given letrozole once(bilateral, 0.5 μL per side) during the fear memory consolidation period) and rats in the hippocampus solvent control group were injected once with dimethyl sulfoxide in the hippocampus(bilateral, 0.5 μL per side) . The open field test and elevated cross maze test were used to evaluate the anxiety level and autonomous exploration ability, the freeze test was used to evaluate the fear memory, the new object recognition test was used to evaluate the non-spatial memory, and ELISA was used to detect serum 17β-E2 level. SPSS 25.0 software was used for data analysis, and one-way ANOVA was used for inter group comparisons, LSD test was used for further pairwise comparisons.Results:(1)In the gonad estradiol experiment, there were statistical differences in various indicators among the five groups of rats in the open field test, elevated cross maze test, freeze test, and new object recognition test( F=20.200, 12.702, 7.514, 10.094, 7.899, 13.211, all P<0.05). The number of upright times, central area activity time, distance and frequency of entering the open arm, and cognitive index of the model group rats were all lower than those of the blank control group(all P<0.05), and the freezing time was higher than that of the blank control group( P<0.05). The number of upright positions((11.20±1.55) times), central area activity time((11.33±1.80) s), distance((1.49±0.26) m) and times((10.00±1.50) times) entering the open arm, the freezing time((92.20±6.07) s) and cognitive index((60.40±3.71)%) in ovariectomy+ estradiol group were all higher than those of ovariectomy group((4.90±0.65) times, (4.31±1.07) s, (0.49±0.06) m, (3.10±0.62) times, (60.30±5.28) s, (32.60±8.08)%)(all P<0.05). (2) In the hippocampal estradiol experiment, there were statistical differences in various indicators among the four groups in the open field test, elevated cross maze test, freeze test, and new object recognition test( F=40.831, 5.553, 9.087, 5.848, 7.657, 9.191, all P<0.05). The numbers of upright positions, distance and frequency of entering open arms of the model group rats were lower than those of the blank control group(all P<0.05). The number of upright positions((3.00 ± 0.39) times), distance of entering open arm((1.17±0.37) m), freezing time((46.70±3.57)s), and cognition index((29.60±2.70)%) in the hippocampal estradiol inhibitor group were all lower than those in the hippocampal solvent control group((10.10±1.40) times, (4.02±0.79) m, (93.70±9.73) s, (54.20±5.08)%)(all P<0.05). Conclusion:17β-E2 derived from the hippocampus and gonad both had ameliorative effect on cognitive memory and anxiety-like behavior in the reconsolidation period of fear memory in PTSD model rats.

Background The converter stations of high-voltage direct current (HVDC) transmission lines generate special total electric fields. At present, few investigations have been conducted on total electric fields in the workplace of converter stations from an perspective of occupational health. Objective To understand the current situation of total electric field strength in the workplace of converter stations. Methods Using purposive sampling, a calibrated HDEM-1 direct current (DC) total electric field strength measurement system was used to measure the total electric fields of 12 converter stations serving 6 DC lines in Southeast and Southwest China according to the Measurement method for total electric field strength and ion current density of the converter stations and DC transmission lines (DL/T 1089—2008). The results were evaluated according to occupational exposure limits recommended by The limits of electromagnetic environment at ±800 kV UHV DC converter station (DL/T 275—2012), the American Conference of Governmental Industrial Hygienists (ACGIH), and the International Commission on Non-Ionizing Radiation Protection (ICNIRP). Results A total of 615 check points were planned, the total electric field strength was 0.05-37.05 kV·m⁻¹, and the median was 10.45 kV·m⁻¹. The total electric field strength of 39 check points (6.3%) exceeded 25 kV·m⁻¹ (the limits of ACGIH and ICNIRP), and the total electric field strength of 12 check points (2.0%) exceeded 30 kV·m⁻¹ (the limit of DL/T 275—2012). There were statistically significant differences in the total electric field strength values and the proportions of exceeding 25 kV·m⁻¹ between the neutral regions and the positive regions and between the neutral regions and the negative regions (P < 0.01). The proportion of total electric field strength exceeding 30 kV·m⁻¹ in the negative regions was higher than that in the positive regions (P < 0.01). There were no significant differences in the total electric field strength of converter stations at different voltage levels and different altitudes (P > 0.05). There were no significant differences in the proportions of total electric field exceeding 25 kV·m⁻¹ and exceeding 30 kV·m⁻¹ in converter stations at different voltage levels and different altitudes (P > 0.05). Conclusion The total electric field in some workplace of converter stations exceeds selected limits. Converter station operators may be exposed to high-strength total electric field for a short time.

Background: Thyroid diseases are highly prevalent worldwide, but their diagnosis remains a challenge. We established reference intervals (RIs) for thyroid-associated hormones and evaluated the prevalence of thyroid diseases in China. Methods: After excluding outliers based on the results of ultrasound screening, thyroid antibody tests, and the Tukey method, the medical records of 20,303 euthyroid adults, who visited the Department of Health Care at Peking Union Medical College Hospital from January 2014 to December 2018, were analyzed. Thyroid-associated hormones were measured by the Siemens Advia Centaur XP analyzer. The RIs for thyroid-associated hormones were calculated according to the CLSI C28-A3 guidelines, and were compared with the RIs provided by Siemens. The prevalence of thyroid diseases over the five years was evaluated and compared using the chi-square test. Results: The RIs for thyroid stimulating hormone (TSH), free thyroxine (FT4), free triiodothyronine (FT3), total thyroxine (TT4), and total triiodothyronine (TT3) were 0.71–4.92 mIU/L, 12.2–20.1 pmol/L, 3.9–6.0 pmol/L, 65.6–135.1 nmol/L, and 1.2–2.2 nmol/L, respectively. The RIs of all hormones except TT4 differed significantly between males and females. The RIs of TSH increased with increasing age. The prevalence of overt hypothyroidism, overt hyperthyroidism, subclinical hypothyroidism, and subclinical hyperthyroidism was 0.5% and 0.8%, 0.2% and 0.6%, 3.8% and 6.1%, and 3.3% and 4.7% in males and females, respectively, which differed from those provided by Siemens. Conclusions Sex-specific RIs were established for thyroid-associated hormones, and the prevalence of thyroid diseases was determined in the Chinese population.

Objective:To analyze the association between low-frequency variants of ARID1A gene and primary liver cancer using latent category model.Methods:The low-frequency variants of ARID1A gene was combined according to different functional areas, and the combined variables were analyzed by using the latent class model to obtain the latent variables. Then the logistic regression was used to analyze the association between low-frequency variants of ARID1A gene and primary liver cancer.Results:The low-frequency variants of ARID1A gene were divided into three categories by the latent class model. The class 1 was mainly unmutated population, the proportion was 94.2% (2 454/2 603). The class 2 was mainly transcriptional regulatory domain mutation, take 4.8% (124/2 603). The class 3 was dominantly exon mutation, about 1.0% (27/2 603). Using class 1 as a reference, it was found that mutations in the transcriptional regulatory domain could reduce the risk of liver cancer ( OR=0.601, 95% CI=0.364-0.992, P=0.046). Conclusion:The latent class model can identify low-frequency variants of gene associated with liver cancer and can be extended to more genetic association studies of low-frequency variants related to complex diseases.

Objective:To analyze the association between low-frequency variants of ARID1A gene and primary liver cancer using latent category model.Methods:The low-frequency variants of ARID1A gene was combined according to different functional areas, and the combined variables were analyzed by using the latent class model to obtain the latent variables. Then the logistic regression was used to analyze the association between low-frequency variants of ARID1A gene and primary liver cancer.Results:The low-frequency variants of ARID1A gene were divided into three categories by the latent class model. The class 1 was mainly unmutated population, the proportion was 94.2% (2 454/2 603). The class 2 was mainly transcriptional regulatory domain mutation, take 4.8% (124/2 603). The class 3 was dominantly exon mutation, about 1.0% (27/2 603). Using class 1 as a reference, it was found that mutations in the transcriptional regulatory domain could reduce the risk of liver cancer ( OR=0.601, 95% CI=0.364-0.992, P=0.046). Conclusion:The latent class model can identify low-frequency variants of gene associated with liver cancer and can be extended to more genetic association studies of low-frequency variants related to complex diseases.

OBJECTIVE: To investigate the prevalence and risk factors of multi-site work-related musculoskeletal disorders(WMSDs) among workers in the industry of electronic equipment manufacturing. METHODS: A total of 815 workers in three factories of electronic equipment manufacturing in Guangdong Province were selected as study subjects by convenience sampling. The prevalence of multi-site WMSDs in the past year was investigated using Musculoskeletal Disorders Investigating Questionnaire and the influencing factors were analyzed. RESULTS: The total prevalence of WMSDs was 69.4%(566/815). The prevalence of multi-site WMSDs was 54.5%(444/815), and the prevalence of one-site WMSDs was 15.0%(122/815). Multiple logistic regression showed that female workers had higher prevalence of multi-site WMSDs than males [odds radio(OR) and 95% confidence interval(CI): 1.59(1.12-2.26), P<0.05]. The prevalence of multi-site WMSDs in left-handed workers was lower than that of right-handed workers [OR(95% CI): 0.42(0.19-0.91), P<0.05]. The longer service of current position and the more neck forward movement, the higher prevalence of multi-site WMSDs [OR(95% CI) were 1.33(1.09-1.63) and 1.62(1.23-2.15), P<0.01]. The workers who had long-time sitting at work, adopted uncomfortable working posture, could decide when to work on their own, kept head down for a long time, or often bending wrists up/down had higher prevalence of multi-site WMSDs [OR(95% CI) were 1.41(1.16-1.73), 1.82(1.40-2.38), 1.79(1.16-2.75), 1.92(1.38-2.69) and 1.60(1.14-2.24), respectively, P<0.01]. The workers who could take turns with colleagues to finish work or had enough rest time had lower prevalence of multi-site WMSDs [OR(95% CI): 0.57(0.41-0.78) and 0.67(0.48-0.92), P<0.05]. The workers who worked >10 h per day had lower prevalence of multi-site WMSDs than those who worked ≤8 h per day [OR(95% CI): 0.57(0.37-0.87), P<0.05]. CONCLUSION: Multi-site WMSDs were more common than one-site WMSDs among workers in the industry of electronic equipment manufacturing, and the prevalence of multi-site WMSDs was high. The risk factors include personal factors, work organization and adverse ergonomic factors.