Large language models (LLMs) offer a modern approach to help inherit traditional Chinese medicine (TCM). This article discussed the progress of LLM applications in TCM and proposed future development directions by reviewing the existing research. We have found that LLMs and related technologies have excellent applications and performance in the management of TCM knowledge and data. They are often applied in information extraction, knowledge graph construction, and data standardization processing. However, data quality and security issues need to be given more attention. In clinical diagnosis and treatment, LLMs can imitate the thinking of TCM by disassembling and reconstructing its diagnostic process and can achieve functions such as prescription recommendation and question and answer (Q&A). However, this approach involves LLMs making inferences and predictions based on existing corpora and thus may not flexibly handle complex environments and tasks. Moreover, the current evaluation criteria for TCM LLMs can be summarized into 3 categories: general evaluation metrics, technical framework evaluation, and evaluation criteria for the characteristics of TCM (such as consistency rates of prescriptions and diagnostic suggestions). However, the lack of a unified and standardized evaluation system hinders the clinical application of TCM LLMs. The future progress of TCM LLMs should focus on the 3 aforementioned critical aspects to achieve technological breakthroughs. In addition, we are promoting the research on vertical TCM LLMs and application terminals. We believe this will bring new ideas to the research on TCM LLMs.

Intelligent prescription recommendation has become an important research direction in traditional Chinese medicine (TCM), offering new opportunities to support clinical decision-making and promote the modernization of TCM practice. With the rapid development of artificial intelligence (AI), a variety of computational approaches have been proposed to learn prescription patterns from clinical data and generate personalized treatment recommendations. However, despite increasing research activity, systematic and comprehensive reviews of AI-driven methods for TCM prescription recommendation remain limited. In this study, we present a comprehensive review of computational approaches for herbal prescription recommendation (HPR) in TCM. Existing methods are systematically categorized into several major paradigms, including traditional machine learning methods, topic model methods, sequential generative methods, deep learning and graph-based methods, and large language model–based frameworks. In addition to summarizing methodological developments, we also review commonly used public datasets and evaluation metrics in this field. Furthermore, representative models with publicly available implementations are experimentally evaluated on multiple benchmark datasets to provide a comparative analysis of their performance on the HPR task. Finally, we discuss the key challenges that hinder the practical deployment of intelligent prescription recommendation systems, including data heterogeneity, limited interpretability, and insufficient integration of TCM domain knowledge. Future research directions are outlined to facilitate the development of more reliable, interpretable, and clinically applicable AI-assisted HPR systems for TCM.

Driven by policy initiatives promoting the integration of digital-intelligent technologies with primary healthcare and the digital transformation of traditional Chinese medicine (TCM), Smart TCM has emerged as a pivotal strategy for enhancing primary healthcare services for major chronic diseases. This paper reviews the current status, challenges, and feasible pathways of Smart TCM in community-level management of cardiovascular–cerebrovascular diseases and tumors, which represent major chronic disease burdens. Our findings indicate that Smart TCM demonstrates emerging potential in primary healthcare for chronic diseases across the entire continuum of “prevention-screening-diagnosis-treatment-management.” However, several significant challenges persist, including data silos and security vulnerabilities, limited applicability of existing models to real-world clinical needs, and insufficient digital literacy among primary healthcare physicians and elderly patients. To address these constraints, this paper proposes a multidimensional strategy encompassing the development of secure and interoperable regional data platforms, lightweight intelligent devices and support services aligned with primary care capacity, unified technical and data standards with corresponding quality control systems, adaptive and dynamically updated artificial intelligent models, interdisciplinary workforce training and patient education programs, and enhanced policy and health insurance support. Overall, Smart TCM shows great promise for improving the efficiency of primary healthcare delivery and establishing innovative TCM-based chronic disease management paradigms.

The digitization of traditional Chinese medicine (TCM) has generated vast amounts of data. However, these data are characterized by significant heterogeneity and complex semantic structures, posing substantial challenges for systematic integration and intelligent analysis, and limiting its potential for modern clinical and computational research. To address the challenges posed by the high heterogeneity and complex structure in TCM data, we designed and developed the TCM Data Hub platform, which is powered by the YiYuan large language models (LLMs). This platform aims to enhance intelligent data processing capabilities and unlock the potential for clinical application of TCM data through systematic integration and efficient utilization, thereby bridging the gap between traditional knowledge and modern computational research. This study first analyzed the heterogeneity and complexity of TCM information with respect to data types, structures, and semantics. A standardized data framework was constructed to enhance data integration and interoperability. Based on the TCM Intelligent Computing Platform of the China Academy of Chinese Medical Sciences, we trained the YiYuan LLMs to acquire domain-specific semantic understanding of TCM, thereby improving the platform’s comprehension of specialized terminology and knowledge systems. Leveraging the natural language processing capabilities of the LLMs, we developed a human-in-the-loop data processing system to enable efficient extraction, cleansing, and structured organization of TCM data. In addition, utilizing Vue and Java technologies, we developed multiple LLM-powered intelligent agents and systems, including a human-in-the-loop data processing system, as well as automated prescription mining and network pharmacology analysis agents. Task-specific agents tailored to TCM data processing were developed to enhance the model’s effectiveness in clinical knowledge discovery. System functionality and platform infrastructure were implemented using Java and Vue technologies.The TCM Data Hub platform has completed system construction and core functionality implementation. It supported integrated management and efficient access to 8 key types of TCM data: prescriptions, materia medica, ingredients, targets, diseases (Western medicine), diseases (TCM), syndromes, and therapeutic methods. The human-in-the-loop data processing system achieved an accuracy of 95.34% in structuring TCM data and supported annotation for data requiring manual labeling. The intelligent agent-driven big-data analytics module enabled 1-click, end-to-end workflows for TCM prescription mining, herb-syndrome association analysis, network pharmacology, and molecular biology research, completing a full data mining task in approximately 30 minutes. Users can interact with and manipulate data through a visual front-end interface. The system demonstrated stable performance, strong scalability, and a user-friendly experience. Empowered by the YiYuan LLMs, the TCM Data Hub platform significantly improves the accessibility, usability, and intelligence of TCM data. It effectively bridges traditional TCM knowledge with modern intelligent technologies, providing robust data support and intelligent tools for TCM research and clinical applications.

Background: Diabetic peripheral neuropathy (DPN) is a common chronic complication of diabetes mellitus that significantly impairs patients’ quality of life. Traditional Chinese medicine (TCM), as a major component of complementary and alternative medicine, has accumulated numerous effective formulas for the clinical management of DPN. However, systematic approaches for optimizing TCM formulas remain limited. Objective: To establish a pathway-oriented approach for TCM formula optimization and to evaluate the efficacy and mechanisms of the optimized formula in DPN. Methods: We developed 2 formula optimization algorithms that defined pathway-oriented herbal correlation (HC) and herbal contribution (HO) to screen and construct a new herbal formula, Qihongtongbi (QHTB), for treating Qi deficiency and blood stasis syndrome complicated with DPN. An animal experiment was subsequently conducted to evaluate the therapeutic effectiveness of QHTB. A total of 32 specific-pathogen-free male ob/ob mice (18–20 g) were randomly divided into 4 groups: model, Mudan Granules (MDG), QHTB low- and high-dose groups (n = 8). Ten male C57BL/6J mice (18–20 g) served as the control group. Metabolomics analysis was further employed to elucidate the biological mechanisms underlying the effects of QHTB. Results: Based on our previous study on TCM medication patterns for treating DPN, 22 candidate herbs were selected for formula optimization. The top 5 candidate herbs (total HC = 25.24 × 10 ) exhibited HC values comparable to those of MDG (total HC = 26.1 × 10 ). These herbs were Carthamus tinctorius L. (HC = 5.40 × 10 ), Astragalus mongholicus Bunge (HC = 5.23 × 10 ), Salvia miltiorrhiza Bunge (HC = 5.15 × 10 ), Commiphora myrrha (T. Nees) Engl. (HC = 5.02 × 10 ), and Glycyrrhiza glabra L. (HC = 4.44 × 10 ). HO analysis showed that the cumulative HO of these 5 herbs exceeded 50%. In vivo experiments demonstrated that, compared with the model group, QHTB significantly lowered blood glucose (P < 0.05), enhanced nerve conduction velocity (P < 0.01), improved nociceptive hypersensitivity (P < 0.01), and ameliorated sciatic nerve morphology, with efficacy comparable to MDG. Serum and fecal metabolomics further revealed that QHTB exerted multipathway regulatory effects, among which nicotinate and nicotinamide metabolism represented a key mechanism. Conclusion: In summary, this study provides a methodological reference for the systematic optimization of TCM formulas.

Background: Xiaochaihu capsules are a widely used traditional Chinese patent medicine. The extraction process, a critical step in their production, necessitates online monitoring to determine the levels of key indicators, thereby informing subsequent processing stages. Objective: The objective of this study was to establish online quantitative models for 6 key indicators (solid content, liquiritin, baicalin, wogonoside, glycyrrhizic acid, and saikosaponin B ) during the extraction process of Xiaochaihu capsules. This approach aims to provide a technical foundation for enhanced process control and to serve as a methodological reference for the application of multisource information fusion in monitoring the extraction of multi-herb traditional Chinese medicine formulations. Methods: Online quantitative models were developed based on multiple sensors, ultraviolet-visible spectroscopy (UV-Vis), and near-infrared spectroscopy (NIR) to monitor 6 indicators during the extraction process of Xiaochaihu capsules. Different preprocessing methods for spectroscopic data were optimized, and data fusion strategies were integrated to improve the predictive performance of the online quantitative models. Results: The optimal modeling strategies for the 6 indicators were as follows: for solid content, a second-order polynomial model was established using the measurement results from 2 sensors (pH and conductivity), with a predictive coefficient of determination (R ) of 0.9821. For baicalin, the optimal model was developed using UV-Vis spectra pretreated with the first derivative (1stD), achieving a predictive R of 0.9841. For wogonoside, modeling was also conducted after 1stD preprocessing of the UV-Vis spectra, yielding a predictive R of 0.9929. For liquiritin, a high-level fusion strategy combining UV-based feature extraction via the random frog (RF) algorithm and NIR-based feature extraction via competitive adaptive reweighted sampling (CARS) was used, achieving an R of 0.9406 in the validation set. For glycyrrhizic acid, the optimal approach involved a combination of UV-CARS feature extraction and NIR-RF feature extraction, coupled with high-level data fusion, resulting in an R of 0.9859 in the validation set. For saikosaponin B , the optimal scheme utilized UV-RF feature extraction combined with NIR-CARS feature extraction and high-level fusion (R = 0.7702 in the validation set). Conclusion: The integration of process analytical technologies and data fusion strategies enabled the successful establishment of online quantitative models for 6 indicators during the extraction process of Xiaochaihu capsules. Overall, the established online quantitative models effectively captured and reflected the dynamic changes of these indicators throughout the extraction process.

Rheumatoid arthritis belongs to arthralgia in the theory of traditional Chinese medicine, with cold dampness obstruction pattern (CDO) and damp heat obstruction pattern (DHO) as the main pattern types. Fine therapeutic effects have been obtained in clinical practice following the differentiation of CDO and DHO. However, suitable animal models are not available currently. For adapting to the clinical diagnosis, as well as carrying out basic research of integrated Chinese and Western medicine and preclinical study on new Chinese medicine in a better way, the “Guidelines for Establishing Animal Models of Rheumatoid Arthritis with Cold Dampness Obstruction Pattern and Damp Heat Obstruction Pattern” (referred to as “Guidelines”) were compiled by our research group in cooperation with renowned experts in clinical, pharmaceutical, zoological, and methodological research fields. According to the theory of disease and syndrome integration, experts standardized the establishment methods for animal models of rheumatoid arthritis with CDO and DHO. The Guidelines were compiled via the nominal group method according to the principle of evidence coming in main place, consensus as the auxiliary, and experience as the references. Contents such as animal species, arthritis induction methods, external stimulation conditions, and assessment indicators were specified in the Guidelines based on the comprehensive evaluation of pathogenesis homology, behavioral phenotypic consistency, and drug therapy predictability between animal models and human diseases. The Guidelines will be applicable to the research on syndromes in rheumatoid arthritis, elucidation of acting mechanism, and development of new medicine, and provide reference for standardizing other types of animal models with the integration of disease and syndrome, to promote the modernization of Chinese medicine research.

Background: Advances in intensive care have markedly improved survival after severe brain injury, leading to a growing population of patients with prolonged disorders of consciousness (pDOC). Current management of pDOC remains largely supportive, and evidence-based neuromodulatory interventions are limited; moreover, existing guidelines provide insufficiently explicit recommendations regarding mechanisms of action and objective biomarkers of treatment response. Transcutaneous auricular vagus nerve stimulation (taVNS) has emerged as a potential noninvasive intervention; however, its modulatory effects on brain function in pDOC are not yet well characterized, and the paucity of integrative mechanistic evidence has constrained its translation into routine clinical practice. Objectives: Within a multimodal assessment framework, this study aims to systematically elucidate the neurobiological mechanisms by which taVNS modulates brain function and autonomic activity in patients with pDOC, and to evaluate its clinical potential to enhance levels of consciousness. Methods: In this randomized controlled trial, 60 patients with vegetative state/minimally conscious state will be enrolled and randomly allocated to a taVNS group, a transcutaneous nonauricular vagus nerve stimulation group (sham), or a control group (n = 20 per group) for a 4-week intervention. The primary outcome will be changes in the Coma Recovery Scale-Revised scores from baseline to weeks 1, 2, and 4 of treatment. Secondary outcomes will include functional brain activity assessed by electroencephalography and functional near-infrared spectroscopy, as well as autonomic modulation indexed by heart rate variability. Functional prognosis will be evaluated using the Glasgow Outcome Scale-Extended at the end of treatment and at a 6-month follow-up. Safety will be assessed by continuous monitoring and documentation of adverse events throughout the study period. Results and discussion: By integrating electroencephalography–functional near-infrared spectroscopy with heart rate variability, this study will characterize the effects of taVNS on functional brain networks and consciousness recovery in pDOC across complementary behavioral, electrophysiological, hemodynamic, and autonomic domains, while interrogating potential sources of clinical and neurobiological heterogeneity. The findings are expected to provide a mechanistic and evidence-based foundation for the mechanism-driven clinical implementation of taVNS and the optimization of stimulation protocols in pDOC. Clinical trial registration: International Traditional Medicine Clinical Trial Registry, ITMCTR20250021041, https://itmctr.ccebtcm.org.cn.