Objective By analyzing the main factors affecting the average length of stay through deep neural networks,redundant factors are eliminated to improve the management effectiveness of the average length of stay.Methods Based on prin-cipal component analysis in machine learning,the multi factor features of average length of hospital stay are reduced and the main factors are extracted.Then,deep neural networks are used to learn the weight relationships between the main factors and predict the actual average length of hospital stay.The data used in this article comes from the homepage of 131 740 inpatient medical re-cords in the HIS system of a tertiary hospital in 2021.Results The main influencing factors of the average length of stay in the hospital in 2021 are preoperative average length of stay,transfusion reactions,admission year,age,admission route,and medical payment method.The corresponding absolute weight values are 2.58,1.89,1.77,0.96,0.76,and 0.75,respectively;The t-test compared and analyzed the predicted average length of stay with the actual average length of stay,and the results showed that there was no significant difference between the two(P＞0.05).Conclusion The DNN model based on the main factors can ef-fectively predict the actual average length of stay,and the hospital's classification of the main influencing factors of the average length of stay obtained in this article can effectively improve the management efficiency of the average length of stay.

Objective: To explore the association between negative life events and smartphone addiction among middle school students, so as to provide theoretical support and practical guidance for prevention and intervention of smartphone addiction among middle school students. Methods: Using cluster sampling, 8 890 students were selected to survey from 27 junior high schools and 3 senior high schools in a district of Shenzhen in 2022 (baseline) and 2023 (followup). Data were collected through selfresigned questionnaires on basic information, the Smartphone Addiction Scale-Short Version, and the Adolescent Selfrating Life Events Checklist. Mixedeffects models were employed to analyze the association. Results: Compared to 2022, the punishment scores of middle school students in 2023 [1.00 (0.00, 6.00) and 1.00 (0.00, 6.00)] decreased (Z=4.27), while the scores of interpersonal stress, learning stress and adaptation [4.00(0.00, 8.00), 4.00(0.00, 8.00); 4.00(1.00, 8.00), 5.00(2.00, 9.00); 2.00 (0.00, 6.00), 3.00 (0.00, 7.00)] increased (Z=-3.04, -8.36, -6.80) (P<0.01). Mixedeffects models revealed a positive doseresponse relationship between negative life events and smartphone addiction (OR=1.08-1.17, P<0.01). Stepwise regression showed independent positive effects of interpersonal stress (OR=1.05), academic stress (OR=1.03), and adaptation stress (OR=1.11) on smartphone addiction (P<0.01). Subgroup analysis of nonaddicted students in 2022 confirmed persistent associations for academic stress (OR=1.03) and adaptation (OR=1.07) (P<0.01). Conclusion Negative life events exhibit a positive doseresponse relationship with smartphone addiction, particularly interpersonal stress, academic stress, and adaptationrelated events.

Objective: To explore the potential causal association between adolescent compulsive behaviour and smartphone addiction based on longitudinal data, so as to provide reference for the establishment of adolescent smartphone addiction interventions. Methods: A preliminary survey and follow-up were conducted on 8 907 middle and high school students in a district of Shenzhen in 2022 and 2023, respectively. Compulsive behaviours were measured by using the Mental Health Inventory for Middle School Students-60 Items (MMHI-60), smartphone addiction was assessed by using the Smartphone Addiction Scale-Short Version ( SAS- SV), and the associations between compulsive behaviours and smartphone addiction were analysed by using multilevel mixed-effects models and subgroup analyses. Results: Smartphone addiction detection rates among middle school students were significantly associated with genders, father s education level, mother s education level, study load subgroups, and whether or not they were single-parent families, and there were statistical differences ( χ 2=17.21-175.34, P <0.05). Students with compulsive behaviours were 2.98 times more likely to develop smartphone addiction than those without compulsive behaviours ( OR=2.98, 95%CI=2.77-3.22, P <0.05). Subgroup analysis of middle school students without smartphone addiction in the first year found that compulsive behaviours significantly predicted smartphone addiction ( OR= 1.76 , 95%CI=1.54-2.01, P <0.05). Conclusion There is a potential causal association between obsessive-compulsive behaviours and smartphone addiction in middle school students, and obsessive-compulsive behaviours in middle school students could significantly predicted the occurrence of smartphone addiction.

Objective:To explore the capability of deep learning image reconstruction (DLIR) compared to adaptive statistical iterative reconstruction (ASiR-V) in improving the image quality and myocardial edge sharpness of dynamic stress myocardial CT perfusion imaging (CTP).Methods:Thirty subjects who underwent dynamic stress myocardial CTP at Guangdong Provincial People′s Hospital from September 2023 to February 2024 were recruited. Image data of all enrolled patients were reconstructed using ASiR-V 50%, ASiR-V 80%, medium-intensity DLIR(DLIR-M), and high-intensity DLIR(DLIR-H), respectively. Regions of interest were selected in the left ventricular cavity, interventricular septum, and left ventricular lateral wall for measurement of CT values and standard deviations (SD), and calculation of signal to noise ratio (SNR) and contrast to noise ratio (CNR). Matlab was utilized to obtain the differences (d) and slopes (s) of CT value changes at four left ventricular myocardial edges for objective edge sharpness evaluation. Two radiologists subjectively scored the images for noise, natural appearance, and edge sharpness. In case of disagreement between the two radiologists, a third senior radiologist′s score was decisive. Left ventricular myocardial blood flow (MBF) of ASiR-V and DLIR images with lower SD, higher SNR and CNR were calculated, respectively. When the normal distribution was satisfied, the independent sample t test was used for comparison between two groups, and the random block design ANOVA was used for comparison between multiple groups. And analysis was conducted using Friedman test for non-normally distributed data, and Bonferroni correction for pairwise comparisons. Results:There were statistically significant differences in SD, SNR, and CNR among the four images in the interventricular septum and left ventricular lateral wall (all P<0.05), with ASiR-V 80% and DLIR-H demonstrating the lowest SD, highest SNR and CNR, and the subjective image noise score. Statistically significant differences were observed in d and s for the four left ventricular myocardial edges (all P<0.05), with DLIR-M and DLIR-H exhibiting the best objective edge sharpness [5 (5, 5)], and ASiR-V 80% the worst [3.5 (3, 4)]. In the subjective scores for natural appearance, DLIR-M and DLIR-H received the highest scores [5 (5, 5)], while ASiR-V 80% received the lowest scores [3 (3, 4)], with statistically significant differences (all P<0.05). There was no statistically significant difference in MBF values calculated from ASiR-V 80% and DLIR-H images (all P>0.05). Conclusions:The SD value, SNR and CNR of dynamic stress myocardial CTP images reconstructed by DLIR-H are equivalent to ASiR-V 80%, and using DLIR-H can improve the edge sharpness of left ventricular myocardium without affecting the calculation of MBF.

Objective By analyzing the main factors affecting the average length of stay through deep neural networks,redundant factors are eliminated to improve the management effectiveness of the average length of stay.Methods Based on prin-cipal component analysis in machine learning,the multi factor features of average length of hospital stay are reduced and the main factors are extracted.Then,deep neural networks are used to learn the weight relationships between the main factors and predict the actual average length of hospital stay.The data used in this article comes from the homepage of 131 740 inpatient medical re-cords in the HIS system of a tertiary hospital in 2021.Results The main influencing factors of the average length of stay in the hospital in 2021 are preoperative average length of stay,transfusion reactions,admission year,age,admission route,and medical payment method.The corresponding absolute weight values are 2.58,1.89,1.77,0.96,0.76,and 0.75,respectively;The t-test compared and analyzed the predicted average length of stay with the actual average length of stay,and the results showed that there was no significant difference between the two(P＞0.05).Conclusion The DNN model based on the main factors can ef-fectively predict the actual average length of stay,and the hospital's classification of the main influencing factors of the average length of stay obtained in this article can effectively improve the management efficiency of the average length of stay.

Objective:To explore the capability of deep learning image reconstruction (DLIR) compared to adaptive statistical iterative reconstruction (ASiR-V) in improving the image quality and myocardial edge sharpness of dynamic stress myocardial CT perfusion imaging (CTP).Methods:Thirty subjects who underwent dynamic stress myocardial CTP at Guangdong Provincial People′s Hospital from September 2023 to February 2024 were recruited. Image data of all enrolled patients were reconstructed using ASiR-V 50%, ASiR-V 80%, medium-intensity DLIR(DLIR-M), and high-intensity DLIR(DLIR-H), respectively. Regions of interest were selected in the left ventricular cavity, interventricular septum, and left ventricular lateral wall for measurement of CT values and standard deviations (SD), and calculation of signal to noise ratio (SNR) and contrast to noise ratio (CNR). Matlab was utilized to obtain the differences (d) and slopes (s) of CT value changes at four left ventricular myocardial edges for objective edge sharpness evaluation. Two radiologists subjectively scored the images for noise, natural appearance, and edge sharpness. In case of disagreement between the two radiologists, a third senior radiologist′s score was decisive. Left ventricular myocardial blood flow (MBF) of ASiR-V and DLIR images with lower SD, higher SNR and CNR were calculated, respectively. When the normal distribution was satisfied, the independent sample t test was used for comparison between two groups, and the random block design ANOVA was used for comparison between multiple groups. And analysis was conducted using Friedman test for non-normally distributed data, and Bonferroni correction for pairwise comparisons. Results:There were statistically significant differences in SD, SNR, and CNR among the four images in the interventricular septum and left ventricular lateral wall (all P<0.05), with ASiR-V 80% and DLIR-H demonstrating the lowest SD, highest SNR and CNR, and the subjective image noise score. Statistically significant differences were observed in d and s for the four left ventricular myocardial edges (all P<0.05), with DLIR-M and DLIR-H exhibiting the best objective edge sharpness [5 (5, 5)], and ASiR-V 80% the worst [3.5 (3, 4)]. In the subjective scores for natural appearance, DLIR-M and DLIR-H received the highest scores [5 (5, 5)], while ASiR-V 80% received the lowest scores [3 (3, 4)], with statistically significant differences (all P<0.05). There was no statistically significant difference in MBF values calculated from ASiR-V 80% and DLIR-H images (all P>0.05). Conclusions:The SD value, SNR and CNR of dynamic stress myocardial CTP images reconstructed by DLIR-H are equivalent to ASiR-V 80%, and using DLIR-H can improve the edge sharpness of left ventricular myocardium without affecting the calculation of MBF.

Objective To compare the clinicopathological characteristics between primary and contralateral cancers in patients with metachronous bilateral breast cancer (MBBC) who carried a BRCA1/2 germline pathogenic variant. Methods A total of 496 BRCA1/2 carriers with primary unilateral breast cancer were included (196 with BRCA1 and 300 with BRCA2). Clinicopathological information of patients was collected, and the median follow-up for the entire cohort was 10.4 years (0.4-20.8 years). Results Among all patients, 31 (15.8%) of the 196 BRCA1 carriers and 49 (16.3%) of the 300 BRCA2 carriers had MBBC, respectively. Among the 31 BRCA1 carriers who developed MBBC, the proportion of triple-negative breast cancer (TNBC) in primary cancer and contralateral cancer was 61.3% and 67.7%, respectively. If the primary cancer of BRCA1-mutated MBBC was TNBC, the probability of the contralateral breast cancer with TNBC was 89.5% (17/19), which was significantly higher than that if the primary cancer was non-TNBC (33.3%, 4/12) (P=0.004). Among the 49 BRCA2 carriers who developed MBBC, the predominant molecular phenotype of the primary and contralateral cancers was HR+ & HER2- (77.6% and 67.3%, respectively; P=0.53). Conclusion Approximately 60% of BRCA1 carriers exhibit TNBC. If a BRCA1 carrier with a TNBC primary breast cancer had an MBBC, the probability of the contralateral breast cancer being TNBC phenotype is almost 89.5%.

Objective:To explore the prescription and medication law of Traditional Chinese Medicine (TCM) compounds in the treatment of vascular dementia (VD) based on patent database.Methods:TCM compounds with patents about VD were retrieved from Chinese patent announcement website of the State Intellectual Property Office and CNKI. The retrieval time was from the establishment to the databases to 31 st, March 2022. The frequency, clusteringand association analysis were carried out with the help of TCM inheritance auxiliary platform (V2.5). The medication law was analyzed. Results:154 TCM compound patents for the treatment of vascular dementia were screened, involving 227 kinds of Chinese materia medica. Among them, Acori Tatarinowii Rhizoma (44 times, 28.57%) was used more frequently, and the common medicinal pair was Salviea Miltiorrhizae Radix et Rhizoma- Acori Tatarinowii Rhizoma (17 times, 11.03%). The medicinal property was mainly warm, the taste was mainly sweet, and the meridian was mainly liver meridian. Those with high confidence based on association rules were " Corni Fructus -Acori Tatarinowii Rhizoma" (0.90), " Corni Fructus -Rehmannize Radix et Praeparata" (0.90). Based on the complex network, it was concluded that the core drugs were 14 groups such as " Rehmannize Radix et Praeparata- Cistanches Herba- Corni Fructus". The new prescriptions extracted by entropy cluster analysis included 7 groups such as " Rehmannize Radix et Praeparata, Cistanches Herba, Corni Fructus and Asparagi Radix". Conclusion:The treatment of VD by TCM compounds with national patents is mainly based on tonifying deficiency, promoting blood circulation and removing blood stasis, eliminating phlegm and dampness, expelling wind and dredging collaterals, opening orifices and resuscitation, which can provide reference for clinical practice and new drug research and development.

Objective To analyze the main active components and potential molecular mechanism of Sophora flavescens against breast cancer based on network pharmacology and molecular docking. Methods The chemical constituents were collected and screened by TCMSP, ETCM database and literature review. The targets of active ingredients were predicted by Swiss Target Prediction database. Breast cancer-related targets were collected by GeneCards, TTD, Drugbank and OMIM. The anti-breast cancer targets of Sophora flavescens were screened by Venny 2.1.0 software. Cytoscape software was used to construct the network diagram of Sophora flavescens-key active ingredients-targets. STRING database was used to analyze the common targets, and PPI network diagram was constructed. GO function enrichment analysis and KEGG pathway enrichment analysis of key target proteins were performed by DAVID database and Hiplot online platform. Schrodinger software was used to calculate the molecular docking between the active ingredients and targets. Molecular biological methods were used to verify the key targets. Results A total of 36 active components with clear structures were screened from Sophora flavescens. 70 anti-breast cancer targets of Sophora flavescens were screened out. 12 core targets including EGFR, AKT1, ESR1, SRC, CYP19A1, AR and ABCB1 participate in endocrine resistance, EGFR tyrosine kinase inhibitors and estrogen signaling pathways in breast cancer. Moreover, the docking score between the core component and the key target AR is the highest. In vitro experiments showed that the extract of Sophora flavescens can inhibit the proliferation of breast cancer cells, induce cell apoptosis and up-regulate AR protein expression. Conclusion It was revealed that Sophora flavescens plays an anti-breast cancer role by regulating complex biological processes through multiple components acting on multiple targets and signaling pathways. The upregulation of AR protein by Sophora flavescens may become a new therapeutic strategy for the treatment of breast cancer.