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.

Endoscopic transforaminal lumbar interbody fusion (TLIF) offers substantial advantages in the management of degenerative spinal diseases, including accelerated postoperative recovery. However, its technical complexity and steep learning curve pose risks for nerve root injury. Optimizing nerve root protection in full-endoscopic facet-sparing TLIF (FE fs-TLIF) and full-endoscopic facet-resecting TLIF (FE fr-TLIF) is essential for enhancing surgical safety. This study aimed to improve the nerve root protection in FE fs-TLIF and FE fr-TLIF by optimizing cage glider selection and insertion techniques based on the specific cage shape—banana-shaped or bullet-shaped. The goal was to ensure safe cage positioning and mitigate nerve root injury during discectomy, endplate preparation, and cage insertion. These strategies were validated through cadaveric simulations and clinical implementation. In FE fr-TLIF utilizing bullet-shaped (straight) cages, one-tip and two-tip cage gliders effectively protected the traversing nerve root by facilitating medial cage entry, thereby minimizing irritation of the exiting nerve root. Conversely, in FE fr-TLIF with banana-shaped cages, the lateral tilt of the cage holder during implantation required the use of a two-tip cage glider to protect the traversing and exiting nerve roots, thereby mitigating the potential risk of nerve irritation. In FE fs-TLIF, a one-tip cage glider is preferred for safeguarding the exiting nerve root, while the traversing root is inherently protected by the medial wall of the facet joint. The use of a two-tip cage glider in FE fs-TLIF can cause injury to the nerve root during glider insertion. In addition to the selection of cage gliders, improper cage insertion steps can also contribute to postoperative neurapraxia. The appropriate selection of cage gliders with corresponding insertion techniques is critical for nerve root protection in endoscopic TLIF. Tailoring these choices to the specific approach (FE fs-TLIF or FE fr-TLIF) and cage type (banana or bullet) enhances surgical safety and clinical outcomes.

Objective To explore the teaching model of the "LungSmart" smart healthcare system in clinical pulmonology teaching and its effectiveness in enhancing the clinical reasoning skills of postgraduate students. Methods A single-center, single-group pretest-posttest educational intervention study was conducted among 30 postgraduate students who participated in respiratory medicine-related teaching activities and enrolled in the "LungSmart" smart healthcare course at Shanghai Pulmonary Hospital from 2024 to 2025 academic year. The course was structured around three core components, namely an AI case repository, dynamic simulation, and immediate feedback, and was delivered over 16 weeks with a total of 64 class hours. Teaching effectiveness was assessed using pretest and posttest clinical reasoning ability scores, while students’ acceptance of the course was evaluated using a 5-point Likert questionnaire. Results All 30 students completed the teaching activities and were included in the final analysis. The pretest score was (78.83±6.25) points, and the posttest score increased to (93.50±4.18) points, with a mean improvement of (14.67±7.06) points (95%CI, 12.03 to 17.30), indicating a statistically significant improvement after the intervention (t=11.37, P<0.001). A total of 30 valid questionnaires were collected at the end of the course, with a response rate of 100.0%. The overall satisfaction score was (4.67±0.15) points, and the mean scores for content satisfaction, practical value, interest stimulation, and professional competence were (4.72±0.23) points, (4.53±0.35) points, (4.72±0.39) points, and (4.75±0.43) points, respectively. Conclusion These findings suggest that the clinical teaching model based on the "LungSmart" smart healthcare system is feasible and well accepted, and may help improve postgraduate students’ clinical reasoning ability.

Kynurenine pathway (KP) is a major catabolic route of tryptophan, generating a series of bioactive metabolites including kynurenine, kynurenic acid (KA), quinolinic acid, and 3-hydroxykynurenine. Beyond its fundamental role in amino acid metabolism, KP exerts critical regulatory functions in neurodevelopment, synaptic plasticity, and immune modulation. Autism spectrum disorder (ASD) is a heterogeneous neurodevelopmental condition with onset in early childhood, characterized by persistent deficits in social communication and restricted, repetitive behaviors. Despite extensive research, the etiological mechanisms underlying ASD remain highly complex and incompletely understood, involving genetic, environmental, and immunological factors. Emerging clinical evidence has consistently revealed a significant imbalance in KP metabolites in individuals with ASD, often accompanied by altered ratios of neuroprotective versus neurotoxic byproducts. However, the primary drivers of this metabolic dysregulation and its causative contribution to ASD pathogenesis are not yet fully elucidated. In this review, we systematically review the enzymatic steps of KP and the principal physiological functions of its major metabolites, with particular emphasis on their dual roles in neuroprotection and neurotoxicity. We then analyze the potential pathogenic mechanisms through which KP dysfunction may contribute to ASD, focusing on two interconnected etiological dimensions. First, KP imbalance can promote oxidative stress, which in turn triggers chronic inflammatory responses via microglial activation and release of pro-inflammatory cytokines, thereby disrupting neuronal homeostasis. Second, aberrant KP metabolism affects neurotransmitter systems, particularly glutamatergic and dopaminergic signaling, leading to impaired neural circuit development and synaptic pruning. By integrating current findings on the KP-ASD association, this review offers a comprehensive etiological framework and a clinically relevant paradigm. Furthermore, we highlight that specific KP metabolites, such as the KA/quinolinic acid ratio, hold promise as peripheral biomarkers for early diagnosis and disease stratification. Finally, we discuss the therapeutic potential of targeting KP enzymes or receptors for personalized intervention strategies, while acknowledging the challenges in translating these findings into clinical practice.

The construction of a medical safety system based on the medical core quality and safty systems is the foundation of the hospital. Multi-campus operation coordinated development is key to the high-quality development of public hospitals and the balanced distribution of high-quality medical resources. Building a medical safety system that comfoms to the specialized layout and operation of branch districts is very important the construction of multi-campus hospitals. In January 2023, Xidan Campus of Peking Union Medical College Hospital established a medical safety system that was compatible with its development, based on the construction of the medical core quality and safty system. The system covered four dimensions: early identification, early assessment, early intervention, and fast response. It included high-risk surgical evaluation and filing, early warning of nursing rooms, periodic medical safety rounds, and rapid response teams and cross hospital transportation of critical care rapid response teams. As of June 2024, the hospital had recorded 570 high-risk surgeries with no unplanned secondary surgeries or unplanned readmissions; Reported nursing warnings 68 times, initiated 93 emergency treatments and cross hospital transfers. All emergency patients received early warning assessments and completed graded and classified transfers and management, effectively ensuring patient safety. This practice could provide references for other multi-campus hospitals to promote the construction and development of medical safety systems.

Objective:To observe the effect of moxibustion on angiogenesis-related indicators in knee joint synovial tissue of adjuvant arthritis model rats,and to explore the mechanism of moxibustion in inhibiting vascular endothelial growth factor(VEGF)expression in synovial tissue and further limiting the activation of Rho family proteins Rac1 and Cdc42,thereby inhibiting angiogenesis during rheumatoid arthritis(RA)treatment.Methods:Forty-eight male Sprague-Dawley rats were equally divided into a normal group,a model group,a moxibustion group,and a moxibustion+VEGF agonist group according to the random principle.The complete Freund's adjuvant method was used for modeling.On the 12th day after modeling,the moxibustion group and the moxibustion+VEGF agonist group were subjected to suspended moxibustion at bilateral Zusanli(ST36),Guanyuan(CV4),and Ashi points for 20 min each time,once a day,for a total of 15 times.The moxibustion+VEGF agonist group received VEGF agonist(tirofiban hydrochloride hydrate)injection in the knee joint cavity at the same time.Hematoxylin-eosin staining was used to evaluate the pathological changes of rat synovial tissue in each group.Immunohistochemistry was used to observe the CD31 expression level in rat synovial tissue.Western blotting was used to detect the levels of VEGF,Rac1,and Cdc42 protein in rat synovial tissue,and polymerase chain reaction(PCR)was used to detect the VEGF mRNA expression.Results:Compared to the normal group,the expression levels of CD31 protein and VEGF mRNA and protein in rat synovial tissue in the model group increased significantly(P<0.01),and the expression levels of phospho-Rac1 and phospho-Cdc42 proteins also increased significantly(P<0.01).After moxibustion intervention,the expression levels of CD31 protein and VEGF mRNA and protein in the moxibustion group were significantly lower than those in the model group(P<0.01),while the differences in each indicator between the moxibustion+VEGF agonist group and the model group were not statistically significant(P>0.05).Compared to the moxibustion group,the expression levels of CD31 protein,VEGF mRNA and protein,phospho-Cdc42,and phospho-Rac1 in the moxibustion+VEGF agonist group increased significantly(P<0.01).Conclusion:Moxibustion improved synovial inflammation in RA by inhibiting angiogenesis.The mechanism may be to regulate angiogenesis-related VEGF,restrict the activation of Rac1 and Cdc42,and inhibit pseudopodia formation in vascular endothelial cells,thereby reducing angiogenesis.

As a new generation of time-of-flight mass spectrometry,multiple-reflection time-of-flight mass spectrometry(MR-TOF-MS)has been increasingly applied in the fields such as nuclear physics,chemistry,and biology due to its ultra-high resolution and rapid analysis capabilities.However,the analytical performance of MR-TOF-MS largely depends on the ion bunch state entering the mass analyzer.In this study,a rectangular ion trap(RIT)was developed,designed and processed using printed circuit board technology,as an ion accumulating and focusing device for MR-TOF mass analyzer.Compared to traditional ion traps composed of two sets of planar electrodes,this RIT had higher voltage utilization efficiency,resulting in more efficient ion collection and focusing.The ions were cooled to a sufficiently small bunch for precise mass measurement with MR-TOF-MS mass spectrometry in only 1 ms of cooling time in the RIT,then orthogonally ejected to the MR-TOF mass spectrometer for mass analysis.Experimental results indicated that the working cycle,ion flux,and ion focusing state of the RIT fully met the requirements of the MR-TOF mass analyzer.When coupled with the MR-TOF mass analyzer,the RIT enabled MR-TOF-MS to achieve a mass resolution of 1.5×105.

Detection of poor-water-soluble substances typically requires an organic solvent containing solution,and therefore,the biosensors that can operate in such conditions are highly promising for practical applications.However,most existing biosensors have been developed in aqueous solution and it is hard to work effectively in organic solvent systems.In this work,we uncovered that the G-quadruplex(G4)/hemin DNAzyme(G4HD)could be strongly reactivated by NH4 in organic solvents at concentration up to 30％,and the activity was significantly higher than that in aqueous solutions containing K+.Based on this discovery,a biosensing system was developed for small molecule in 40％methanol by coupling G4HD with our previously identified RNA-cleaving DNAzyme(E3).This system comprised an ATP-activated aptazyme that was made of E3 and a G4HD that was activated by the cleavage product of the aptazyme.By using a manmade pipette tip filter to separate the reaction products,the subsequent colorimetric reaction of the G4HD could be triggered.This biosensing system enabled the visualization of target molecules in organic cosolvents without any other instruments.The activable DNAzyme-cascade sensing system had potential in point-of-care detection of poor-water-soluble pollutants,thereby enhancing the practical values of functional nucleic acid-based biosensors in real-world detection conditions.

Objective To propose a multiple regression-variational auto-encoders(MR-VAE)model to realize precise and non-invasive prediction of intra-abdominal pressure(IAP)in critically ill patients.Methods At first,a dataset was constructed by retrospectively analysing baseline characteristics and clinical indicators of 100 critically ill patients admitted to the Intensive Care Unit of Daping Hospital of Army Medical University between 30 August 2019 and 30 March 2021.Then,a MR-VAE prediction model was developed by integrating a feedforward neural network for supervised regression onto a variational autoencoder(VAE)framework and incorporating multiple regression strategies to mitigate feature interference.Finally,the MR-VAE model had its performance evaluated by its comparison with five classical models including support vector machines(SVM),convolutional neural networks(CNN),Scikit-learn integrated model(SIM),multi-layer perceptron(MLP)and K-nearest neighbors(KNN),and its prediction accuracy verified by testing the data of 10 randomly selected patients.Results The MR-VAE model behaved the best when compared with the five classical models,with a mean squared error(MSE)of 0.207,a root mean square error(RMSE)of 0.454,a mean absolute error(MAE)of 0.361,a median absolute deviation(MAD)of 0.243,an explained variance score(EVS)of 0.814 and a R2of 0.823,which also outperformed the five models in fitting performance,convergence and final loss.In random sample testing,the MR-VAE model exhibited high consistency between predicted and actual values.Conclusion The MR-VAE model proposed can accurately predict IAP,which has great potential in reducing the repeated measurements of IAP in critically ill patients and providing new ideas for the early diagnosis and treatment of IAH.