ObjectiveTo construct a large language model （LLM）-based intelligent pre-consultation system for traditional Chinese medicine （TCM） to improve efficacy of clinical practice. MethodsA TCM large language model was fine-tuned using DeepSpeed ZeRO-3 distributed training strategy based on YAYI 2-30B. A weighted undirected graph network was designed and an agent-based syndrome differentiation model was established based on relationship data extracted from TCM literature and clinical records. An agent collaboration framework was developed to integrate the TCM LLM with the syndrome differentiation model. Model performance was comprehensively evaluated by Loss function， BLEU-4， and ROUGE-L metrics， through which training convergence， text generation quality， and language understanding capability were assessed. Professional knowledge test sets were developed to evaluate system proficiency in TCM physician licensure content， TCM pharmacist licensure content， TCM symptom terminology recognition， and meridian identification. Clinical tests were conducted to compare the system with attending physicians in terms of diagnostic accuracy， consultation rounds， and consultation duration. ResultsAfter 100 000 iterations， the training loss value was gradually stabilized at about 0.7±0.08， indicating that the TCM-LLM has been trained and has good generalization ability. The TCM-LLM scored 0.38 in BLEU-4 and 0.62 in ROUGE-L， suggesting that its natural language processing ability meets the standard. We obtained 2715 symptom terms， 505 relationships between diseases and syndromes， 1011 relationships between diseases and main symptoms， and 1 303 600 relationships among different symptoms， and constructed the Agent of syndrome differentiation model. The accuracy rates in the simulated tests for TCM practitioners， licensed pharmacists of Chinese materia medica， recognition of TCM symptom terminology， and meridian recognition were 94.09%， 78.00%， 87.50%， and 68.80%， respectively. In clinical tests， the syndrome differentiation accuracy of the system reached 88.33%， with fewer consultation rounds and shorter consultation time compared to the attending physicians （P＜0.01）， suggesting that the system has a certain pre- consultation ability. ConclusionThe LLM-based intelligent TCM pre-diagnosis system could simulate diagnostic thinking of TCM physicians to a certain extent. After understanding the patients' natural language， it collects all the patient's symptom through guided questioning， thereby enhancing the diagnostic and treatment efficiency of physicians as well as the consultation experience of the patients.

Objective: To investigate the objective characteristics of tongue manifestation in different stages of damp-heat syndrome in diabetic kidney disease (DKD). Methods: A cross-sectional study enrolled 134 patients with DKD G3-5 stages who met the diagnostic criteria for damp-heat syndrome in DKD. The patients were treated at Dongzhimen Hospital, Beijing University of Chinese Medicine, from May 2023 to January 2024. The patients were divided into three groups: DKD G3, DKD G4, and DKD G5 stage, with 53, 33, and 48 patients in each group, respectively. Clinical general data (gender, age, and body mass index) and damp-heat syndrome scores were collected from the patients. The YZAI-02 traditional Chinese medicine (TCM) AI Tongue Image Acquisition Device was used to capture tongue images from these patients. The accompanying AI Open Platform for TCM Tongue Diagnosis of the device was used to analyze and extract tongue manifestation features, including objective data on tongue color, tongue quality, coating color, and coating texture. Clinical data and objective tongue manifestation characteristics were compared among patients with DKD G3-5 based on their DKD damp-heat syndrome status. Results: No statistically significant difference in gender or body mass index was observed among the three patient groups. The DKD G3 stage group had the highest age (P<0.05). The DKD G3 stage group had a lower score for symptoms of poor appetite and anorexia(P<0.05) than the DKD G5 group. No statistically significant difference was observed in damp-heat syndrome scores among the three groups. Compared with the DKD G5 stage group, the DKD G3 stage group showed a decreased proportion of pale color at the tip and edges of the tongue (P<0.05). The DKD G4 stage group exhibited an increased proportion of crimson at the root of the tongue, a decreased proportion of thick white tongue coating at the root, a decreased proportion of pale color at the tip and edges of the tongue, an increased hue value (indicating color tone) of the tongue color in the middle, an increased brightness value (indicating color lightness) of the tongue coating color in the middle, and an increased thickness of the tongue coating (P<0.05). No statistically significant difference was observed in other tongue color proportions, color chroma values, body characteristics, coating color proportions, coating color chroma values, and coating texture characteristics among the three groups. Conclusion Tongue features differ in different stages of DKD damp-heat syndrome in multiple dimensions, enabling the inference that during the DKD G5 stage, the degree of qi and blood deficiency in the kidneys, heart, lungs, liver, gallbladder, spleen, and stomach is prominent. Dampness is more likely to accumulate in the lower jiao, particularly in the kidneys, whereas heat evil in the spleen and stomach is the most severe. These insights provide novel ideas for the clinical treatment of DKD.

Objective To predict the radiation impact of discharging wastewater containing uranium within the specified limit generated during the normal operation of a new production line at a nuclear fuel plant on the receiving water body and its downstream, and to provide a reference for the management of radioactive liquid effluent discharge from nuclear facilities. Methods Based on the technical guidelines for environmental impact assessment, literature on radiation environmental impact assessment, and data collected from on-site investigations, appropriate hydrological parameters and prediction models were selected to analyze and predict the variation pattern of radioactive nuclide uranium along the receiving water body and the radiation exposure of nearby residents. Results The maximum increase in uranium concentration in the receiving water body and its downstream caused by the discharge of uranium-containing wastewater was 1.14 μg/L. The maximum predicted concentration was 2.75 μg/L after adding the background data of the water body. The resulting maximum individual annual effective dose for the public was 1.49 × 10⁻⁴ mSv/a. Conclusion The maximum predicted uranium concentration in the receiving water body and its downstream is lower than the uranium concentration limit of 30 μg/L specified in the Standards for Drinking Water Quality (GB 5749-2022). The maximum individual annual effective dose for the public is much lower than the control value of 0.2 mSv/a specified in the Radiation Protection Regulations for Uranium Processing and Fuel and Fuel Manufacturing Facilities (EJ 1056-2018). The radiation impact is acceptable.

Objective To compare the performance and efficacy of prognostic models constructed by different machine learning algorithms in predicting the survival period of lung transplantation (LTx) recipients. Methods Data from 483 recipients who underwent LTx were retrospectively collected. All recipients were divided into a training set and a validation set at a ratio of 7:3. The 24 collected variables were screened based on variable importance (VIMP). Prognostic models were constructed using random survival forest (RSF) and extreme gradient boosting tree (XGBoost). The performance of the models was evaluated using the integrated area under the curve (iAUC) and time-dependent area under the curve (tAUC). Results There were no significant statistical differences in the variables between the training set and the validation set. The top 15 variables ranked by VIMP were used for modeling and the length of stay in the intensive care unit (ICU) was determined as the most important factor. Compared with the XGBoost model, the RSF model demonstrated better performance in predicting the survival period of recipients (iAUC 0.773 vs. 0.723). The RSF model also showed better performance in predicting the 6-month survival period (tAUC _{6 months} 0.884 vs. 0.809, P = 0.009) and 1-year survival period (tAUC _{1 year} 0.896 vs. 0.825, P = 0.013) of recipients. Based on the prediction cut-off values of the two algorithms, LTx recipients were divided into high-risk and low-risk groups. The survival analysis results of both models showed that the survival rate of recipients in the high-risk group was significantly lower than that in the low-risk group (P<0.001). Conclusions Compared with XGBoost, the machine learning prognostic model developed based on the RSF algorithm may preferably predict the survival period of LTx recipients.

Objective To construct large medical model named by "Huaxi HongYi"and explore its application effectiveness in assisting medical record generation. Methods By the way of a full-chain medical large model construction paradigm of "data annotation - model training - scenario incubation", through strategies such as multimodal data fusion, domain adaptation training, and localization of hardware adaptation, "Huaxi HongYi" with 72 billion parameters was constructed. Combined with technologies such as speech recognition, knowledge graphs, and reinforcement learning, an application system for assisting in the generation of medical records was developed. Results Taking the assisted generation of discharge records as an example, in the pilot department, after using the application system, the average completion times of writing a medical records shortened (21 min vs. 5 min) with efficiency increased by 3.2 time, the accuracy rate of the model output reached 92.4%. Conclusion It is feasible for medical institutions to build independently controllable medical large models and incubate various applications based on these models, providing a reference pathway for artificial intelligence development in similar institutions.

Objective To study the chemical constituents of Fallopia aubertii L.Henry Holub. Methods The chemical constituents of Fallopia aubertii L.Henry Holub. were separated and purified by online two-dimensional preparative liquid chromatography and identified by physical and chemical constants and spectral analysis. The inhibitory activities on xanthine oxidase were determined by ultraviolet spectrophotometry. Results Ten compounds were isolated from the extract of Fallopia aubertii L.Henry Holub, including isotachioside（1）, 3,4,5-trimethoxyphenyl-（6'-O-galloyl）-O-β-D-Glucopyranoside（2）, 1-hydroxy-,4,5-1-O-[6'-O-（4''-carboxy-1'',3'',5'trihydrotrimethoxyphenylxy）-phenyl]-β-D-glucopyranoside（3）, myricetrin（4）, myricetin（5）, rutin（6）, quercetin-3-O-β-D-galactoside（7）, quercetin-3-O-β-D-glucopyranoside（8）, lyciumideA（9）, and N-trans-Feruloyltyramine（10）. The inhibitory activity test results showed that the IC₅₀ of compound 5 was 15.92 μmol/L, and the IC₅₀ of compound 6 was 87.36 μmol/L. Conclusion Compounds 1,2,3,4 and 8 were isolated from Medicago polymorpha for the first time. Compounds 5 and 6 had xanthine oxidase inhibitory activity.

Objective:To evaluate the predictive effect of the loss of signal (LOS) on the recurrent laryngeal nerve (RLN) injury.Methods:The literatures on PubMed, Web of Science, CNKI, and Wanfang Medical Network database were published before 30 April 2023. English search terms included "thyroid gland surgery" "thyroidectomy" "intraoperative neuromonitoring" "intraoperative nerve monitoring" and "recurrent laryngeal nerve". Chinese search terms included "thyroidectomy", "thyroid surgery" "recurrent laryngeal nerve" "intraoperative nerve monitoring". Two evaluators screened the literature, extracted the materials and evaluated the risk of bias of the study independently. If there were different opinions, researchers should resolve which through consultation and ask the third-party researcher when necessary. The Meta-analysis was performed with the Review Manager 5.4 software.Results:A total of thirty-three studies were included and were all analyzed for primary outcome measures while only twenty-six of which were analyzed for secondary outcome measures. Meta-analysis showed that the positive predictive value of LOS in intraoperative nerve monitoring (IONM) was 65% [ OR=1.88, 95% CI: 1.36-2.60]. Then these thirty-three articles included in IONM were divided into I-IONM, C-IONM and mixed groups. Subgroup analysis showed that the positive predictive value of LOS in I-IONM, C-IONM and mixed groups were 62% [ OR=1.63, 95% CI: 1.05-2.52], 75% [ OR=2.93, 95% CI: 1.64-5.22] and 70% [ OR=2.38, 95% CI: 1.77-3.12] respectively. When these thirty-three included articles were divided into Asian, European, North American and Oceania, subgroup analysis showed that the positive predictive value of LOS was 50% [ OR=1.01, 95% CI: 0.44-2.31], 70% [ OR=2.29, 95% CI: 1.60-3.28], 82% [ OR=4.68, 95% CI: 3.79-5.78] and 83% [ OR=4.81, 95% CI: 3.10-7.46] respectively. Meta-analysis of secondary outcome measures in twenty-six articles showed that the negative predictive value of LOS was 99.6% [ RD=1.51, 95% CI: 1.48-1.53], with the sensitivity of 89% [ RD=1.24, 95% CI: 1.11-1.37] and specificity of 98%[ RD=1.43, 95% CI: 1.40-1.47]. Conclusion:The occurrence of LOS during IONM in thyroidectomy has a positive prediction effect and a higher negative prediction effect, sensitivity and specificity on RLN injury.

Aim To explore the effect and mechanism of the artesunate(ART)on hepatocellular carcinoma(HCC).Methods The cell lines MHCC-97H and HCC-LM3 were used to be detected.MTT and clone formation were used to determine the cell proliferation;Wound healing was used to detect the cell migration;Transwell was used to test the cell invasion.Flow-cy-tometry was used to detect cell apoptosis and cell cy-cle.RNA-seq and qRT-PCR was used to detect the genes expression.Results The proliferation,migra-tion and invasion of treated cells were obviously inhibi-ted(P＜0.01).Moreover,the apoptosis rate in-creased significantly,so did the proportion of G2/M cells.Transcriptomic analysis identified GADD45A as a potential target of ART through RNA-sequencing da-ta,and suggested that ART might induce apoptosis and cell cycle arrest through regulating the expression of GADD45A.In addition,the results of mechanism studies and signaling analysis suggested that GADD45A had interaction with its upstream gene NACC1(nucle-us accumbens associated 1).Moreover,after ART treatment,the expressions of GADD45A and NACC1 were changed significantly.Conclusion ART may be a potential drug to resist HCC by affecting the expres-sion of GADD45A and its upstream gene NACC1,which provides a new drug,a new direction and a new method for the clinical treatment of HCC.

Aim To establish the fingerprint of raw bupleurum and vinegar bupleurum,investigate the difference in their anti-liver fibrosis effects,and ex-plore the relationship between the chemical composition of raw bupleurum and vinegar bupleurum and their an-ti-liver fibrosis efficacy.Methods The fingerprints of 10 batches of raw bupleuri and 10 batches of bupleuri were established by UPLC method.The liver fibrosis cell model in vitro was established by TGF-β induced LX-2 hepatic stellate cells.The liver fibrosis cell mod-el was analyzed with collagen type Ⅰ(col1a1)and α-smoothmuscleactin.The expression of α-SMA protein was used as the pharmacodynamic index.MetaboAna-lyst5.0 was used to screen the difference markers af-fecting the quality of raw bupledges and vinegar bu-pledges with VIP value＞1 as the criterion.Orthogo-nal partial least squares discriminant analysis(OPLS-DA)was used to screen the main components of raw bupleurum and vinegar bupleurum against liver fibro-sis.Results There were 18 peaks in the UPLC fin-gerprints of raw bupleurum and vinegar bupleurum,and the analysis results showed that there were nine main differences between raw bupleurum and vinegar bupleurum,among which peaks 9,7 and 6 could be considered as bupleurin d,bupleurin a and bupleurin f.The results of spectral effect relationship showed that the main components of bupleurum anti-liver fibrosis were peaks 11,12,14,15 and 18.Conclusions The established fingerprint method of raw bupleurum and vinegar bupleurum is simple and feasible,and the important components of anti-liver fibrosis activity are screened through the spectrum effect relationship,which provides a basis for clarifying the material basis of anti-liver fibrosis effect of raw bupleurum and vine-gar bupleurum.

Aim To observe the effect of Astragalus membranaceus and Angelica sinensis on the apoptosis of chondrocytes,and to investigate the effect of Astrag-alus membranaceus and Angelica sinensis on the sur-vival of fresh ostecartilage allograft.Methods Forty-eight 4-month-old New Zealand white rabbits,half male and half female,were randomly divided into sham operation group,model group,positive group and As-tragalus and Angelica 5∶1 group.In addition to the sham operation group,the other groups were both male and female donors and recipients for knee joint osteo-cartilage cross transplantation modeling.After 8 weeks of drug intervention,samples were taken for general observation,HE staining,saffrane-O staining,immu-nohistochemical staining,qPCR and Western blot de-tection.Results Compared with model group,As-tragalus and Angelica 5∶1 group and positive group,the repair site healed better,the morphology of osteo-chondrocytes tended to be normal,and the division and proliferation were obvious.Proteoglycan deposition in-creased and type Ⅱ collagen content was higher,the differences were statistically significant(P＜0.05).qPCR and Western blot results showed that compared with model group,the mRNA and protein expressions of Bax,caspase-3 and caspase-9 in other groups were significantly decreased(P＜0.05).Conclusion As-tragalus and Angelica can promote the survival of fresh osteochondral allograft,and its mechanism may be re-lated to promoting collagen production,promoting chondrocyte proliferation and inhibiting chondrocyte apoptosis.