As a distinctive resource of Chinese civilization， traditional Chinese medicine （TCM） technology transfer faces significant opportunities under the background of digital and intelligent transformation， while also being constrained by unique challenges such as the complexity of its theoretical system， lengthy industrial chains， and multidimensional policy restrictions， resulting in a "high-value-high-threshold" paradox. At present， TCM technology transfer is deeply trapped in a "threefold reluctance" dilemma， i.e.， unwillingness to transfer， inability to transfer， and lack of capacity to transfer. Specifically， the disconnection between scientific research evaluation systems and market demand leads to low conversion rates of research achievements， unclear ownership and compliance risks suppress innovation incentives， and the absence of professional services intensifies supply-demand mismatches. This article systematically analyzes the specific characteristics of TCM technology transfer and proposes a breakthrough pathway centered on full-chain digital and intelligent transformation. By integrating technologies such as intelligent sorting systems， blockchain-based traceability， and AI diagnostic models， the TCM ecosystem spanning "cultivation-production-service" can be reconstructed. In terms of standardization， promoting the progression from "experience-based data conversion" to "data standardization" and further to "intelligent standardization" is advocated to resolve quality control challenges. For example， a "three-no-one-full" certification system can strengthen quality trust. Policy coordination should focus on optimizing mechanisms for the transformation of scientific and technological achievements， while exploring intellectual property securitization and risk-sharing models to stimulate research momentum. In terms of internationalization， reliance on the Belt and Road Initiative platform to promote the export of geo-authentic medicinal material brands and standards is recommended to build a dual-driven model of "technology plus culture". Looking ahead， through the construction of national-level databases， the cultivation of interdisciplinary talent， and the mutual recognition of international standards， a new paradigm of "scientific intelligent manufacturing" can be formed， providing systematic solutions for the modernization of TCM and global health governance.

ObjectiveThe essence of syndrome manifestation recognition in traditional Chinese medicine (TCM) is to infer the body’s latent pathogenesis state from clinical observational information, rather than to perform simple label matching. However, previous studies have largely modeled this task as syndrome pattern classification within a fixed label space, which does not adequately reflect the cognition process of TCM syndrome differentiation centered on pathogenesis reasoning, and is also insufficient to capture the openness, semantic variability, and cross-disease reusability of syndrome manifestation expression. This study aimed to investigate whether introducing pathogenesis reasoning chain-of-thought (PR-CoT) supervision into large language models (LLMs) could improve the quality and cognitive consistency of syndrome manifestation recognition and support cross-disease transfer. MethodsSyndrome manifestation recognition was formulated as a conditional generation task under the framework of clinical observational information (X)→pathogenesis structure (Z)→syndrome pattern output (Y), where Z serves as an explicit intermediate structural variable linking the clinical evidence and syndrome judgment. Within this framework, a PR-CoT-supervised dataset for syndrome manifestation recognition was constructed based on medical case records of spleen-stomach disorders. After preprocessing, information extraction, manual proofreading, and data cleaning, the dataset comprised 4 800 training cases, 400 development cases, and 400 test cases. Each sample was annotated with a structured PR-CoT consisting of three progressive levels: clinical information summarization, comprehensive pathogenesis analysis, and syndrome pattern output. Supervised fine-tuning was conducted on open-source LLMs, with an end-to-end model serving as the baseline. Qwen3-32B was used as the primary experimental model, and Qwen3-14B as the scale comparison model. A progressive multidimensional evaluation framework was further established, comprising a structural parsing level, a semantic similarity level, and an expert blind review level. At the structural parsing level, syndrome pattern expressions were decomposed into structural elements and evaluated using Precision, Recall, F1 score, and Jaccard similarity. At the semantic similarity level, independent LLMs scored the theoretical proximity between predicted and reference syndrome patterns. At the expert blind review level, three TCM experts independently evaluated model outputs on two dimensions: syndrome differentiation consistency and terminology standardization of syndrome patterns. In addition, zero-shot cross-disease transfer evaluation was conducted on gynecological and heart-system disorder test sets. ResultsAt the structural parsing level, PR-CoT supervision did not lead to a stable improvement in the element-wise overlap of syndrome pattern structural components. Compared with the corresponding baselines, neither Qwen3-32B nor Qwen3-14B showed consistent advantages in structural matching metrics after the introduction of PR-CoT supervision. In contrast, at the semantic similarity level, PR-CoT supervision produced stable positive gains across different model scales and evaluation systems. The average semantic score of Qwen3-32B increased from 6.425 8 in the baseline model to 6.585 0 after PR-CoT supervision, and that of Qwen3-14B increased from 5.870 0 to 5.964 2. At the expert blind review level, the overall score of Qwen3-32B (PR-CoT) was 7.026 0±0.107 7, higher than 6.416 3±0.288 9 for its baseline. In zero-shot cross-disease testing, the PR-CoT model still showed advantages in semantic evaluation and expert evaluation on both gynecological and heart-system disorder test sets, indicating a certain degree of transferability. ConclusionThe benefits of PR-CoT supervision are mainly reflected in TCM semantic consistency and clinical plausibility, rather than in improved hard matching of structural elements. These findings support understanding syndrome manifestation recognition as a process of generating and expressing latent pathogenesis structures, rather than as a classification task within a traditional fixed label space. By introducing pathogenesis reasoning as an explicit intermediate structure into the modeling process and combining it with a progressive multidimensional evaluation framework, this study provides a methodological pathway for intelligent TCM syndrome differentiation that integrates theoretical alignment, interpretability, and multi-level evaluation.

ObjectiveThe essence of syndrome manifestation recognition in traditional Chinese medicine (TCM) is to infer the body’s latent pathogenesis state from clinical observational information, rather than to perform simple label matching. However, previous studies have largely modeled this task as syndrome pattern classification within a fixed label space, which does not adequately reflect the cognition process of TCM syndrome differentiation centered on pathogenesis reasoning, and is also insufficient to capture the openness, semantic variability, and cross-disease reusability of syndrome manifestation expression. This study aimed to investigate whether introducing pathogenesis reasoning chain-of-thought (PR-CoT) supervision into large language models (LLMs) could improve the quality and cognitive consistency of syndrome manifestation recognition and support cross-disease transfer. MethodsSyndrome manifestation recognition was formulated as a conditional generation task under the framework of clinical observational information (X)→pathogenesis structure (Z)→syndrome pattern output (Y), where Z serves as an explicit intermediate structural variable linking the clinical evidence and syndrome judgment. Within this framework, a PR-CoT-supervised dataset for syndrome manifestation recognition was constructed based on medical case records of spleen-stomach disorders. After preprocessing, information extraction, manual proofreading, and data cleaning, the dataset comprised 4 800 training cases, 400 development cases, and 400 test cases. Each sample was annotated with a structured PR-CoT consisting of three progressive levels: clinical information summarization, comprehensive pathogenesis analysis, and syndrome pattern output. Supervised fine-tuning was conducted on open-source LLMs, with an end-to-end model serving as the baseline. Qwen3-32B was used as the primary experimental model, and Qwen3-14B as the scale comparison model. A progressive multidimensional evaluation framework was further established, comprising a structural parsing level, a semantic similarity level, and an expert blind review level. At the structural parsing level, syndrome pattern expressions were decomposed into structural elements and evaluated using Precision, Recall, F1 score, and Jaccard similarity. At the semantic similarity level, independent LLMs scored the theoretical proximity between predicted and reference syndrome patterns. At the expert blind review level, three TCM experts independently evaluated model outputs on two dimensions: syndrome differentiation consistency and terminology standardization of syndrome patterns. In addition, zero-shot cross-disease transfer evaluation was conducted on gynecological and heart-system disorder test sets. ResultsAt the structural parsing level, PR-CoT supervision did not lead to a stable improvement in the element-wise overlap of syndrome pattern structural components. Compared with the corresponding baselines, neither Qwen3-32B nor Qwen3-14B showed consistent advantages in structural matching metrics after the introduction of PR-CoT supervision. In contrast, at the semantic similarity level, PR-CoT supervision produced stable positive gains across different model scales and evaluation systems. The average semantic score of Qwen3-32B increased from 6.425 8 in the baseline model to 6.585 0 after PR-CoT supervision, and that of Qwen3-14B increased from 5.870 0 to 5.964 2. At the expert blind review level, the overall score of Qwen3-32B (PR-CoT) was 7.026 0±0.107 7, higher than 6.416 3±0.288 9 for its baseline. In zero-shot cross-disease testing, the PR-CoT model still showed advantages in semantic evaluation and expert evaluation on both gynecological and heart-system disorder test sets, indicating a certain degree of transferability. ConclusionThe benefits of PR-CoT supervision are mainly reflected in TCM semantic consistency and clinical plausibility, rather than in improved hard matching of structural elements. These findings support understanding syndrome manifestation recognition as a process of generating and expressing latent pathogenesis structures, rather than as a classification task within a traditional fixed label space. By introducing pathogenesis reasoning as an explicit intermediate structure into the modeling process and combining it with a progressive multidimensional evaluation framework, this study provides a methodological pathway for intelligent TCM syndrome differentiation that integrates theoretical alignment, interpretability, and multi-level evaluation.

Objectives:To investigate association of the non-high-density lipoprotein cholesterol to high-density lipoprotein cholesterol ratio (NHHR) with the stroke severity and short-term outcome in patients with acute ischemic stroke (AIS), and to evaluate the predictive value of NHHR for outcome.Methods:Patients with the first-ever AIS admitted to the Affiliated Hospital of Qingdao University from June 2018 to June 2024 whose etiological types were large artery atherosclerosis (LAA), small vessel occlusion (SVO) and cardiac embolism (CE) were included retrospectively. According to the National Institutes of Health Stroke Scale (NIHSS) score at admission, the patients were divided into mild stroke group (≤8) and moderate to severe stroke group (>8). According to the modified Rankin Scale score at discharge, they were divided into good outcome group (≤2) and poor outcome group (>2). Multivariate logistic regression analysis was use to determine the independent correlation between NHHR and stroke severity and short-term outcome in patients with AIS. Receiver operating characteristic (ROC) curve was used to evaluate the predictive value of NHHR for short-term poor outcome in overall patients with AIS and different etiological subtypes. Results:A total of 2 865 patients with AIS were enrolled, including 1 925 males (67.2%), aged (61.00 ± 10.17) years. 2 483 patients (86.67%) had mild stroke and 382 (13.33%) had moderate to severe stroke; 2 161 (75.43%) had good short-term outcome, while 704 (24.57%) had poor short-term outcome. Multivariate logistic regression analysis showed that NHHR was significantly and independently associated with moderate to severe stroke (odds ratio [ OR] 2.251, 95% confidence interval [ CI] 1.895-2.675; P<0.001) and poor short-term outcome ( OR 3.454, 95% CI 2.936-4.063; P<0.001). ROC curve analysis showed that NHHR had a high predictive value for short-term poor outcome in patients with AIS (the area under the curve [AUC] 0.764, 95% CI 0.745-0.784), and it also demonstrated high predictive value in patients with various etiological types such as LAA (AUC=0.755, 95% CI 0.730-0.781), SVO (AUC=0.801, 95% CI 0.777-0.824) and CE (AUC=0.797, 95% CI 0.774-0.820). Conclusion:NHHR is significantly correlated with the severity of stroke and poor short-term outcome in patients with AIS, and has a high predictive value for poor short-term outcome.

Objective:To construct a clinical decision-making ability indicator system based on evidence-based practice for pediatric nursing interns, and to provide a scientific basis for clinical teaching and evaluation.Methods:A method combining literature analysis, Delphi expert consultation, and empirical research was used. Firstly, a systematic search of Chinese and English databases (2018-2023) was conducted. Literature was screened based on the PICO framework and evidence-based data were extracted, resulting in a preliminary system consisting of 4 primary indicators, 12 secondary indicators, and 39 tertiary indicators. Subsequently, the indicators were revised through two rounds of Delphi expert consultation (25 experts with 19-27 years of work experience). The expert authority coefficients (Cr) were 0.898-0.907 and the Kendall's concordance coefficients were 0.351-0.420 ( P<0.001). Finally, the analytic hierarchy process was used to determine the weights, and the reliability and validity were verified through a questionnaire survey (sample size: 30 participants in preliminary survey and 58 participants in formal survey). Results:The constructed indicator system included 4 primary indicators (weights), 13 secondary indicators, and 42 tertiary indicators. The weights of the primary indicators were as follows: knowledge integration ability (0.300), evidence-based practice ability (0.250), clinical judgment ability (0.280), and ethical decision-making ability (0.170). The importance scores of all items exceeded 4.0 points (out of 5 points), and the coefficients of variation were less than 0.20. The reliability and validity tests showed that the Cronbach's α of the overall scale was 0.89, and the intraclass correlation coefficient was 0.88. The cumulative variance contribution rate of exploratory factor analysis was 69.30%. The confirmatory factor analysis demonstrated a good model fit with a comparative fit index of 0.93 and a root mean square error of approximation of 0.05. Conclusions:This indicator system has high scientificity and practicality, and can provide a reference for the standardized cultivation and evaluation of clinical decision-making ability of pediatric nursing interns. In the future, it is necessary to strengthen advanced evidence-based skills training and long-term application effectiveness tracking.

OBJECTIVE To compare the predictive accuracy and clinical applicability of four vancomycin individualized dosing tools （SmartDose， ClinCalc， Gulou， Pharmado） and provide a basis for rational clinical medication use. METHODS A retrospective cohort study was conducted， enrolling 479 adult patients who received vancomycin therapy and underwent steady-state trough concentration monitoring in Zhongshan Hospital， Fudan University （Xiamen Branch） from January 1， 2022， to June 30， 2024. The predictive accuracy of each tool was evaluated using indicators， such as mean error （ME）, mean absolute error （MAE）， mean percentage error （MPE）， mean absolute percentage error （MAPE）， the proportion of patients with an absolute percentage error （APE） of less than 30%， the 95% limits of agreement， and the overall relative percentage difference between predicted and measured values. Using indicators such as accessibility， patient management， and recommendation of multiple treatment options， the clinical panxso@163.com applicability of the tools for all patients was evaluated； using the discrepancy in accuracy between the predicted and actual measured blood drug concentrations as an indicator， the clinical applicability was assessed for patients in different renal function subgroups （hyperfunction， normal， mild impairment， moderate impairment， and severe impairment）. RESULTS In terms of accuracy， SmartDose demonstrated the best overall performance with an MAPE of 46.40% and a proportion of APE ＜30% （46.56%）. Bland-Altman analysis indicated that SmartDose had the smallest overall relative percentage difference （－7.25%）， although the 95% limits of agreement were broad for all tools， with differences between the upper and lower limits exceeding 200%. In terms of applicability， all four dosing tools were freely accessible and demonstrated good availability； SmartDose and Pharmado provided the most comprehensive solutions， offering features such as patient management， multiple regimen recommendations， and drug concentration-time curve plotting. Stratified analysis based on renal function revealed that Pharmado showed optimal prediction for hyperfiltration patients （mean difference： 0.11 mg/L）. SmartDose and ClinCalc showed relatively better performance in normal and mild renal impaiment （mean difference： 0.37， 0.51 mg/L and －1.13， －1.33 mg/L，respectively）. SmartDose performed best in moderate renal impairment （mean difference： －2.60 mg/L）. Pharmado and Gulou had smaller prediction biases in severe renal impairment （mean differences： 1.52 mg/L and －0.23 mg/L， respectively）. CONCLUSIONS The four individualized dosing tools demonstrated limited accuracy in the initial prediction of vancomycin concentrations. Among them， SmartDose demonstrates the highest overall prediction accuracy and possesses comprehensive clinical management features. It is recommended that Pharmado be preferred for patients with renal hyperfiltration； SmartDose or ClinCalc can be used for patients with normal or mildly impaired renal function； SmartDose is recommended for patients with moderately impaired renal function； Pharmado or Gulou may be considered for patients with severely impaired renal function.

Artificial intelligence (AI) researchers and cheminformatics specialists strive to identify effective drug precursors while optimizing costs and accelerating development processes. Digital molecular representation plays a crucial role in achieving this objective by making molecules machine-readable, thereby enhancing the accuracy of molecular prediction tasks and facilitating evidence-based decision making. This study presents a comprehensive review of small molecular representations and AI-driven drug discovery downstream tasks utilizing these representations. The research methodology begins with the compilation of small molecule databases, followed by an analysis of fundamental molecular representations and the models that learn these representations from initial forms, capturing patterns and salient features across extensive chemical spaces. The study then examines various drug discovery downstream tasks, including drug-target interaction (DTI) prediction, drug-target affinity (DTA) prediction, drug property (DP) prediction, and drug generation, all based on learned representations. The analysis concludes by highlighting challenges and opportunities associated with machine learning (ML) methods for molecular representation and improving downstream task performance. Additionally, the representation of small molecules and AI-based downstream tasks demonstrates significant potential in identifying traditional Chinese medicine (TCM) medicinal substances and facilitating TCM target discovery.
Artificial Intelligence ; Drug Discovery/methods* ; Humans ; Machine Learning ; Medicine, Chinese Traditional ; Small Molecule Libraries/chemistry*

Objective:To evaluate the utility of the Rapid Axial Roll Test （RART）, Supine Roll Test （SRT）, and Lying-Down Test （LDT） in determining the affected semicircular canal in cases of horizontal semicircular canal benign paroxysmal positional vertigo （HSC-BPPV）. Methods:A total of 330 patients diagnosed with HSCBPPV from September 2022 to September 2023 were collected and divided into three groups based on the different positional tests received: ①SRT Group, ②LDT+SRT Group, ③RART+SRT Group. The trial was divided into two stages: LDT/RART for patients in the first stage, and SRT for patients in the second stage. The elicitation rate of nystagmus among the three groups was compared to evaluate the accuracy in determining the affected semicircular canal in HSCBPPV. Results:Nystagmus was elicited in 84.55% （279/330） of the patients by positional tests. The elicitation rate of nystagmus in the RART+SRT/LDT group was 94.55% （104/110）, in the LDT+SRT group it was 84.11% （90/107）, and in the SRT group it was 69.91% （79/113）. The differences among the three groups were statistically significant （χ²= 23.88, P<0.001）. In the ② and ③ groups, there was a statistically significant difference in the elicitation rate of nystagmus between stage Ⅰ （patients with LDT or RART） （χ²=43.842, P<0.001）. SRT was performed in the stage Ⅱ, and there was a statistically significant difference in nystagmus extraction rate between the two groups （χ² =4.690, P=0.030）. The difference in the proportion of agreement between stage Ⅰ（LDT or RART） and stageⅡ （SRT） in determining the affected side of the semicircular canal was also statistically significant （χ² =40.502, P<0.001）. For patients with a consistent diagnosis of the affected semicircular canal, the difference in cure rate was not significant （P=0.149）. The Kappa statistic indicated substantial agreement between RART and SRT in terms of eliciting nystagmus （agreement 96.36%, Kappa = 0.730, P<0.001）. Conclusion:RART and SRT show a high degree of agreement regarding the elicitation rate of nystagmus. RART is simple and safe, and it can effectively induce the characteristic nystagmus of HSC-BPPV, accurately identify the responsible semicircular canal and provide a more optimized examination protocol for clinical practice in HSCBPPV.
Humans ; Semicircular Canals/physiopathology* ; Benign Paroxysmal Positional Vertigo/diagnosis* ; Female ; Male ; Middle Aged ; Nystagmus, Pathologic/diagnosis* ; Vestibular Function Tests/methods* ; Aged ; Vertigo/diagnosis* ; Adult

Background::The increasing incidence of inflammatory bowel disease (IBD) presents significant medical and societal challenges. A well-designed IBD database is crucial for both epidemiological studies and clinical management. However, inconsistencies between regional databases hinder cross-institutional and international research, especially between Eastern and Western societies.Methods::We developed a new IBD database, the 301 IBD database, integrating the IBD clinical characteristics from the Penn IBD database (USA) and the latest IBD guidelines and consensus and clinical practices of the Chinese PLA General Hospital (PLAGH). We applied this database to analyze clinical data of IBD inpatients at PLAGH from 2008 to 2023.Results::The 301 IBD database contains 490 items in 6 sections including demographic characteristics, personal history, clinical phenotype, disease activity, laboratory tests and examinations, and treatment. Features of the 301 IBD database include inpatient focus, biochemical indicators and opportunistic infection focus, and more about ulcerative colitis (UC)-associated complications. Single-center analysis revealed an increasing hospitalization trend, from 2.35% in 2008 to 3.94% in 2023. We found that the clinical characteristics of our UC inpatients are predominantly male (62.5%), extensive lesions (55.1%), low usage of biologics (4.1%), and a high incidence of UC-CRC (3.0%). The clinical characteristics of CD inpatients included male predominance (68.39%), early onset age (35.43 ± 14.75-year-old), and high rate of surgery (25.81%).Conclusion::The 301 IBD database, integrating Eastern and Western clinical data, provides a valuable tool for IBD clinical research. Future international, multicenter collaborations are expected to further enhance its utility.

Hypertrophic scar is a fibrous hyperplastic disorder that arises from skin injuries. The current therapeutic modalities are constrained by the dense and rigid scar tissue which impedes effective drug delivery. Additionally, insufficient autophagic activity in fibroblasts hinders their apoptosis, leading to excessive matrix deposition. Here, we developed an active microneedle (MN) system to overcome these challenges by integrating micromotor-driven drug delivery with autophagy regulation to remodel the scar microenvironment. Specifically, sodium bicarbonate and citric acid were introduced into the MNs as a built-in engine to generate CO₂ bubbles, thereby enabling enhanced lateral and vertical drug diffusion into dense scar tissue. The system concurrently encapsulated curcumin (Cur), an autophagy activator, and triamcinolone acetonide (TA), synergistically inducing fibroblast apoptosis by upregulating autophagic activity. In vitro studies demonstrated that active MNs achieved efficient drug penetration within isolated scar tissue. The rabbit hypertrophic scar model revealed that TA-Cur MNs significantly reduced the scar elevation index, suppressed collagen I and transforming growth factor-β1 (TGF-β1) expression, and elevated LC3 protein levels. These findings highlight the potential of the active MN system as an efficacious platform for autonomous augmented drug delivery and autophagy-targeted therapy in fibrotic disorder treatments.