Objective: To develop QingNangTCM, a specialized large language model (LLM) tailored for expert-level traditional Chinese medicine (TCM) question-answering and clinical reasoning, addressing the scarcity of domain-specific corpora and specialized alignment. Methods: We constructed QnTCM_Dataset, a corpus of 100 000 entries, by integrating data from ShenNong_TCM_Dataset and SymMap v2.0, and synthesizing additional samples via retrieval-augmented generation (RAG) and persona-driven generation. The dataset comprehensively covers diagnostic inquiries, prescriptions, and herbal knowledge. Utilizing P-Tuning v2, we fine-tuned the GLM-4-9B-Chat backbone to develop QingNangTCM. A multi-dimensional evaluation framework, assessing accuracy, coverage, consistency, safety, professionalism, and fluency, was established using metrics such as bilingual evaluation understudy (BLEU), recall-oriented understudy for gisting evaluation (ROUGE), metric for evaluation of translation with explicit ordering (METEOR), and LLM-as-a-Judge with expert review. Qualitative analysis was conducted across four simulated clinical scenarios: symptom analysis, disease treatment, herb inquiry, and failure cases. Baseline models included GLM-4-9B-Chat, DeepSeek-V2, HuatuoGPT-II (7B), and GLM-4-9B-Chat (freeze-tuning). Results: QingNangTCM achieved the highest scores in BLEU-1/2/3/4 (0.425/0.298/0.137/0.064), ROUGE-1/2 (0.368/0.157), and METEOR (0.218), demonstrating a balanced and superior normalized performance profile of 0.900 across the dimensions of accuracy, coverage, and consistency. Although its ROUGE-L score (0.299) was lower than that of HuatuoGPT-II (7B) (0.351), it significantly outperformed domain-specific models in expert-validated win rates for professionalism (86%) and safety (73%). Qualitative analysis confirmed that the model strictly adheres to the “symptom-syndrome-pathogenesis-treatment” reasoning chain, though occasional misclassifications and hallucinations persisted when dealing with rare medicinal materials and uncommon syndromes. Conclusion Combining domain-specific corpus construction with parameter-efficient prompt tuning enhances the reasoning behavior and domain adaptation of LLMs for TCM-related tasks. This work provides a technical framework for the digital organization and intelligent utilization of TCM knowledge, with potential value for supporting diagnostic reasoning and medical education.

Objective: To develop a clinical decision and prescription generation system (CDPGS) specifically for diarrhea in traditional Chinese medicine (TCM), utilizing a specialized large language model (LLM), Qwen-TCM-Dia, to standardize diagnostic processes and prescription generation. Methods: Two primary datasets were constructed: an evaluation benchmark and a fine-tuning dataset consisting of fundamental diarrhea knowledge, medical records, and chain-of-thought (CoT) reasoning datasets. After an initial evaluation of 16 open-source LLMs across inference time, accuracy, and output quality, Qwen2.5 was selected as the base model due to its superior overall performance. We then employed a two-stage low-rank adaptation (LoRA) fine-tuning strategy, integrating continued pre-training on domain-specific knowledge with instruction fine-tuning using CoT-enriched medical records. This approach was designed to embed the clinical logic (symptoms → pathogenesis → therapeutic principles → prescriptions) into the model’s reasoning capabilities. The resulting fine-tuned model, specialized for TCM diarrhea, was designated as Qwen-TCM-Dia. Model performance was evaluated for disease diagnosis and syndrome type differentiation using accuracy, precision, recall, and F1-score. Furthermore, the quality of the generated prescriptions was compared with that of established open-source TCM LLMs. Results: Qwen-TCM-Dia achieved peak performance compared to both the base Qwen2.5 model and five other open-source TCM LLMs. It achieved 97.05% accuracy and 91.48% F1-score in disease diagnosis, and 74.54% accuracy and 74.21% F1-score in syndrome type differentiation. Compared with existing open-source TCM LLMs (BianCang, HuangDi, LingDan, TCMLLM-PR, and ZhongJing), Qwen-TCM-Dia exhibited higher fidelity in reconstructing the “symptoms → pathogenesis → therapeutic principles → prescriptions” logic chain. It provided complete prescriptions, whereas other models often omitted dosages or generated mismatched prescriptions. Conclusion By integrating continued pre-training, CoT reasoning, and a two-stage fine-tuning strategy, this study establishes a CDPGS for diarrhea in TCM. The results demonstrate the synergistic effect of strengthening domain representation through pre-training and activating logical reasoning via CoT. This research not only provides critical technical support for the standardized diagnosis and treatment of diarrhea but also offers a scalable paradigm for the digital inheritance of expert TCM experience and the intelligent transformation of TCM.

Objective: To investigate methods for constructing a high-quality instructional dataset for traditional Chinese medicine (TCM) mental disorders and to validate its efficacy. Methods: We proposed the Fine-Med-Mental-T&P methodology for constructing high-quality instruction datasets in TCM mental disorders. This approach integrates theoretical knowledge and practical case studies through a dual-track strategy. (i) Theoretical track: textbooks and guidelines on TCM mental disorders were manually segmented. Initial responses were generated using DeepSeek-V3, followed by refinement by the Qwen3-32B model to align the expression with human preferences. A screening algorithm was then applied to select 16 000 high-quality instruction pairs. (ii) Practical track: starting from over 600 real clinical case seeds, diagnostic and therapeutic instruction pairs were generated using DeepSeek-V3 and subsequently screened through manual evaluation, resulting in 4 000 high-quality practice-oriented instruction pairs. The integration of both tracks yielded the Med-Mental-Instruct-T&P dataset, comprising a total of 20 000 instruction pairs. To validate the dataset’s effectiveness, three experimental evaluations (both manual and automated) were conducted: (i) comparative studies to compare the performance of models fine-tuned on different datasets; (ii) benchmarking to compare against mainstream TCM-specific large language models (LLMs); (iii) data ablation study to investigate the relationship between data volume and model performance. Results: Experimental results demonstrate the superior performance of T&P-model fine-tuned on the Med-Mental-Instruct-T&P dataset. In the comparative study, the T&P-model significantly outperformed the baseline models trained solely on self-generated or purely human-curated baseline data. This superiority was evident in both automated metrics (ROUGE-L > 0.55) and expert manual evaluations (scoring above 7/10 across accuracy). In benchmark comparisons, the T&P-model also excelled against existing mainstream TCM LLMs (e.g., HuatuoGPT and ZuoyiGPT). It showed particularly strong capabilities in handling diverse clinical presentations, including challenging disorders such as insomnia and coma, showcasing its robustness and versatility. Data ablation studies showed that T&P-model performance had an overall upward trend with minor fluctuations when training data increased from 10% to 50%; beyond 50%, performance improvement slowed significantly, with metrics plateauing and approaching a saturation point. Conclusion This study has successfully constructed the specialized Med-Mental-Instruct-T&P instruction dataset for TCM mental disorders proposed the systematic Fine-Med-Mental-T&P methodology for its development, effectively addressing the critical challenge of high-quality, domain-specific data scarcity in TCM, and providing essential data support for developing intelligent TCM diagnostic and therapeutic systems.

Objective: To systematically characterize the developmental trajectory and interdisciplinary integration of intelligent diagnosis in traditional Chinese medicine (TCM) through quantitative topic evolution analysis, we addressed the fragmentation of existing research and clarified the long-term research structure and evolutionary patterns of the field. Methods: A topic evolution analysis was performed on Chinese-language literature pertaining to intelligent diagnosis in TCM. Publications were retrieved from the China National Knowledge Infrastructure (CNKI), Wanfang Data, and China Science and Technology Journal Database (VIP), covering the period from database inception to July 3, 2025. A hybrid segmentation approach, based on cumulative publication growth trends and inflection point detection, was applied to divide the research timeline into distinct stages. Subsequently, the latent Dirichlet allocation (LDA) model was used to extract research topics, followed by alignment and evolutionary analysis of topics across different stages. Results: A total of 3 919 publications published between 2003 and 2025 were included, and the research trajectory was divided into five stages based on data-driven breakpoint detection. The field exhibited a clear evolutionary shift from early rule-based systems and tongue-pulse image and signal analysis (2006 – 2010), to machine-learning-based syndrome and prescription modeling (2011 – 2015), followed by deep-learning-driven pattern recognition and formula association (2016 – 2020). Since 2021, research has increasingly emphasized knowledge-graph construction, multimodal integration, and intelligent clinical decision-support systems, with recent studies (2024 – 2025) showing the emergence of large language models and agent-based diagnostic frameworks. Topic evolution analysis further revealed sustained cross-stage continuity in syndrome modeling and prescription association analysis, alongside the progressive consolidation of integrated intelligent diagnostic platforms. Conclusion By identifying key technological transitions and persistent core research themes, our findings offer a structured reference framework for the design of intelligent diagnostic systems, the construction of knowledge-driven clinical decision-support tools, and the alignment of AI models with TCM diagnostic logic. Importantly, the stage-based evolutionary insights derived from this analysis can inform future methodological choices, improve model interpretability and clinical applicability, and support the translation of intelligent TCM diagnosis from experimental research to real-world clinical practice.

Objective: To address the dual challenges of long-tail distribution and feature sparsity in traditional Chinese medicine (TCM) syndrome differentiation within real clinical settings, we propose a data-efficient learning framework enhanced by knowledge graphs. Methods: We developed Agent-GNN, a three-stage decoupled learning framework, and validated it on the Traditional Chinese Medicine Syndrome Diagnosis (TCM-SD) dataset containing 54 152 clinical records across 148 syndrome categories. First, we constructed a comprehensive medical knowledge graph encoding the complete TCM reasoning system. Second, we proposed a Functional Patient Profiling (FPP) method that utilizes large language models (LLMs) combined with Graph Retrieval-Augmented Generation (RAG) to extract structured symptom-etiology-pathogenesis subgraphs from medical records. Third, we employed heterogeneous graph neural networks to learn structured combination patterns explicitly. We compared our method against multiple baselines including BERT, ZY-BERT, ZY-BERT + Know, GAT, and GPT-4 Few-shot, using macro-F1 score as the primary evaluation metric. Additionally, ablation experiments were conducted to validate the contribution of each key component to model performance. Results: Agent-GNN achieved an overall macro-F1 score of 72.4%, representing an 8.7 percentage points improvement over ZY-BERT + Know (63.7%), the strongest baseline among traditional methods. For long-tail syndromes with fewer than 10 samples, Agent-GNN reached a macro-F1 score of 58.6%, compared with 39.3% for ZY-BERT + Know and 41.2% for GPT-4 Few-shot, representing relative improvements of 49.2% and 42.2%, respectively. Ablation experiments confirmed that the explicit modeling of etiology-pathogenesis nodes contributed 12.4 percentage points to this enhanced long-tail syndrome performance. Conclusion This study proposes Agent-GNN, a knowledge graph-enhanced framework that effectively addresses the long-tail distribution challenge in TCM syndrome differentiation. By explicitly modeling manifestation-mechanism-essence patterns through structured knowledge graphs, our approach achieves superior performance in data-scarce scenarios while providing interpretable reasoning paths for TCM intelligent diagnosis.

Objective: To develop a depression recognition model by integrating the spirit-expression diagnostic framework of traditional Chinese medicine (TCM) with machine learning algorithms. The proposed model seeks to establish a TCM-informed tool for early depression screening, thereby bridging traditional diagnostic principles with modern computational approaches. Methods: The study included patients with depression who visited the Shanghai Pudong New Area Mental Health Center from October 1, 2022 to October 1, 2023, as well as students and teachers from Shanghai University of Traditional Chinese Medicine during the same period as the healthy control group. Videos of 3 – 10 s were captured using a Xiaomi Pad 5, and the TCM spirit and expressions were determined by TCM experts (at least 3 out of 5 experts agreed to determine the category of TCM spirit and expressions). Basic information, facial images, and interview information were collected through a portable TCM intelligent analysis and diagnosis device, and facial diagnosis features were extracted using the Open CV computer vision library technology. Statistical analysis methods such as parametric and non-parametric tests were used to analyze the baseline data, TCM spirit and expression features, and facial diagnosis feature parameters of the two groups, to compare the differences in TCM spirit and expression and facial features. Five machine learning algorithms, including extreme gradient boosting (XGBoost), decision tree (DT), Bernoulli naive Bayes (BernoulliNB), support vector machine (SVM), and k-nearest neighbor (KNN) classification, were used to construct a depression recognition model based on the fusion of TCM spirit and expression features. The performance of the model was evaluated using metrics such as accuracy, precision, and the area under the receiver operating characteristic (ROC) curve (AUC). The model results were explained using the Shapley Additive exPlanations (SHAP). Results: A total of 93 depression patients and 87 healthy individuals were ultimately included in this study. There was no statistically significant difference in the baseline characteristics between the two groups (P > 0.05). The differences in the characteristics of the spirit and expressions in TCM and facial features between the two groups were shown as follows. (i) Quantispirit facial analysis revealed that depression patients exhibited significantly reduced facial spirit and luminance compared with healthy controls (P < 0.05), with characteristic features such as sad expressions, facial erythema, and changes in the lip color ranging from erythematous to cyanotic. (ii) Depressed patients exhibited significantly lower values in facial complexion L, lip L, and a values, and gloss index, but higher values in facial complexion a and b, lip b, low gloss index, and matte index (all P < 0.05). (iii) The results of multiple models show that the XGBoost-based depression recognition model, integrating the TCM “spirit-expression” diagnostic framework, achieved an accuracy of 98.61% and significantly outperformed four benchmark algorithms—DT, BernoulliNB, SVM, and KNN (P < 0.01). (iv) The SHAP visualization results show that in the recognition model constructed by the XGBoost algorithm, the complexion b value, categories of facial spirit, high gloss index, low gloss index, categories of facial expression and texture features have significant contribution to the model. Conclusion This study demonstrates that integrating TCM spirit-expression diagnostic features with machine learning enables the construction of a high-precision depression detection model, offering a novel paradigm for objective depression diagnosis.

Objective: To develop a dual-branch deep learning framework for accurate multi-label classification of fundus diseases, addressing the key limitations of insufficient complementary feature extraction and inadequate cross-modal feature fusion in existing automated diagnostic methods. Methods: The fundus multi-label classification dataset with 12 disease categories (FMLC-12) dataset was constructed by integrating complementary samples from Ocular Disease Intelligent Recognition (ODIR) and Retinal Fundus Multi-Disease Image Dataset (RFMiD), yielding 6 936 fundus images across 12 retinal pathology categories, and the framework was validated on both FMLC-12 and ODIR. Inspired by the holistic multi-regional assessment principle of the Five Wheels theory in traditional Chinese medicine (TCM) ophthalmology, the dual-branch multi-label network (DBMNet) was developed as a novel framework integrating complementary visual feature extraction with pathological correlation modeling. The architecture employed a TransNeXt backbone within a dual-branch design: one branch processed red-green-blue (RGB) images to capture color-dependent features, such as vascular patterns and lesion morphology, while the other processed grayscale-converted images to enhance subtle textural details and contrast variations. A feature interaction module (FIM) effectively integrated the multi-scale features from both branches. Comprehensive ablation studies were conducted to evaluate the contributions of the dual-branch architecture and the FIM. The performance of DBMNet was compared against four state-of-the-art methods, including EfficientNet Ensemble, transfer learning-based convolutional neural network (CNN), BFENet, and EyeDeep-Net, using mean average precision (mAP), F1-score, and Cohen's kappa coefficient. Results: The dual-branch architecture improved mAP by 15.44 percentage points over the single-branch TransNeXt baseline, increasing from 34.41% to 44.24%, and the addition of FIM further boosted mAP to 49.85%. On FMLC-12, DBMNet achieved an mAP of 49.85%, a Cohen’s kappa coefficient of 62.14%, and an F1-score of 70.21%. Compared with BFENet (mAP: 45.42%, kappa: 46.64%, F1-score: 71.34%), DBMNet outperformed it by 4.43 percentage points in mAP and 15.50 percentage points in kappa, while BFENet achieved a marginally higher F1-score. On ODIR, DBMNet achieved an F1-score of 85.50%, comparable to state-of-the-art methods. Conclusion DBMNet effectively integrates RGB and grayscale visual modalities through a dual-branch architecture, significantly improving multi-label fundus disease classification. The framework not only addresses the issue of insufficient feature fusion in existing methods but also demonstrates outstanding performance in balancing detection across both common and rare diseases, providing a promising and clinically applicable pathway for standardized, intelligent fundus disease classification.

Objective: Patients with atherosclerotic cardiovascular disease (ASCVD) following percutaneous coronary intervention (PCI) are classified as very-high-risk individuals in cardiovascular disease (CVD) risk stratification. The distribution pattern of traditional Chinese medicine (TCM) syndromes in this patient population, as well as its association with blood lipid profiles and clinical prognosis, remains unclear. The present prospective cohort study aims to investigate these correlations, thereby providing insights to enrich the research fields. Methods: We enrolled consecutive patients with ASCVD who underwent PCI at the Integrated Cardiology Unit of China-Japan Friendship Hospital between September 1, 2020 and December 31, 2022. Demographics and clinical characteristics, signs and symptoms defining each TCM syndrome, and fasting venous blood samples were collected at baseline and follow up or upon major adverse cardiovascular events (MACEs). We analyzed the correlation between TCM syndromes, blood lipid profiles, and MACEs, and developed a new joint prognostic model incorporating both TCM syndromes and blood lipids using logistic regression. The analyses were based on detailed baseline and one-year follow-up data. Results: A per-protocol analysis was performed on 586 patients with complete data ultimately. During the one-year follow-up, 174 patients (29.69%) experienced a MACE. We performed statistical analyses on comorbidities, medication, and biochemical indicators across groups defined by TCM syndrome differentiation. When comparing different TCM syndromes, no significant differences were found in age, body mass index (BMI), history of revascularization, comorbidities, family history of CVD, smoking or drinking, or statin intensity (P > 0.05). Patients with intertwined phlegm and blood stasis syndrome exhibited significantly higher levels of total cholesterol (TC, 5.27 ± 1.18 mmol/L, P < 0.001), triglyceride (TG, 1.96 ± 1.33 mmol/L, P = 0.008), low-density lipoprotein cholesterol (LDL-C, 3.35 ± 0.79 mmol/L, P < 0.001), and high-density lipoprotein cholesterol (HDL-C, 1.24 ± 0.81 mmol/L, P < 0.001) compared with those with other TCM syndromes combined. A multivariable logistic regression model was constructed to predict MACEs. The model included TCM syndrome type [with intertwined phlegm and blood stasis as a predictor, adjusted odds ratio (OR) = 1.413, 95% confidence interval (CI): 0.517 – 3.864, P = 0.501], age (adjusted OR = 0.97, 95% CI: 0.955 – 1.001, P = 0.057), male gender (adjusted OR = 0.698, 95% CI: 0.416 – 1.170, P = 0.173), TC (adjusted OR = 1.004, 95% CI: 0.513 – 1.965, P = 0.990), and LDL-C (adjusted OR = 5.825, 95% CI: 2.214 – 15.326, P < 0.001). This model demonstrated good discriminatory ability for MACEs in post-PCI ASCVD patients [the area under the receiver operating characteristic (ROC) curve (AUC) = 0.865, 95% CI: 0.816 – 0.914]. Conclusion The intertwined phlegm and blood stasis TCM syndrome is associated with a distinct atherogenic lipid profile characterized by elevated levels of TC and LDL-C. The prognostic model that incorporates this TCM syndrome type along with conventional lipid parameters (TC and LDL-C) shows good discriminatory ability for predicting MACEs in ASCVD patients after PCI, underscoring the potential clinical utility of integrating TCM syndrome differentiation into CVD risk assessment.

Objective: To investigate whether Tuina alleviates fibrotic symptoms in myofascial trigger points (MTrPs) by regulating transforming growth factor (TGF)-β1/Smad3 signaling pathway, thereby deactivating these points. Methods: This study comprised two experimental phases. In phase 1, 27 specific pathogen-free (SPF) grade female Sprague-Dawley (SD) rats were randomized into three groups: control 1, model 1, and Tuina 1 groups. Model 1 and Tuina 1 groups underwent an 8-week MTrPs modeling protocol involving blunt impact and eccentric exercise. After successful modeling, rats in Tuina 1 group received manual pressing on nodules or cord-like taut bands on the medial aspect of the left hindlimb. Pain sensitivity and tissue stiffness were evaluated via pressure pain threshold (PPT) and soft tissue tension (STT). Muscle histopathology and fibrosis were observed using hematoxylin and eosin (HE) and Masson staining. Inflammatory factors in muscle were measured by enzyme-linked immunosorbent assay (ELISA), while immunofluorescence (IF) and Western blot (WB) were used to detect the expression levels of α-smooth muscle actin (α-SMA), collagen Ⅲ, and TGF-β1. In phase 2, 45 SPF female SD rats were randomized into five groups: control 2, model 2, Tuina 2, TGF-β1 inhibitor (TI), and Tuina + TGF-β1 agonist (Tuina + TA) groups. All groups except control 2 underwent standardized MTrPs modeling. Rats in Tuina 2 group received consistent pressing manipulation. TI group received intraperitoneal injections of oxymatrine, while Tuina + TA group received intraperitoneal injections of SRI-011381 hydrochloride followed by the same pressing protocol as Tuina 2 group. WB was used to detect the expression of collagen I, collagen III, TGF-β1, and phosphorylated-Smad3 (p-Smad3)/Smad3. Results: In phase 1, Tuina significantly improved PPT and STT in MTrPs of rats (P < 0.01), reversed pathological damages including disorganized muscle fiber arrangement, abnormal myocyte morphology, and exacerbated fibrosis. In addition, in MTrPs of rats in model 1 group, expression levels of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), interleukin (IL)-1β, IL-6, tumor necrosis factor (TNF)-α, and fibrosis markers (α-SMA, collagen I, and collagen III) were upregulated, and all exhibited a significant downward trend after Tuina intervention (P < 0.05 or P < 0.01). This indicates that the therapeutic effects of Tuina are directly associated with reduced local inflammation and fibrosis in MTrPs. In phase 2, compared with model 2 group, rats in TI and Tuina 2 groups had decreased expression levels of TGF-β1 and p-Smad3/Smad3 in MTrPs, alongside reduced levels of inflammatory factors (IL-1β, IL-6, NF-κB, and TNF-α) and fibrosis markers (α-SMA, collagen I, and collagen III) (P < 0.05 or P < 0.01). When co-administered with TGF-β1 agonist, the therapeutic effects of Tuina were significantly attenuated, with rebounded TGF-β1 expression and p-Smad3/Smad3 in local MTrPs, and fibrosis and inflammatory responses were re-exacerbated (P < 0.05 or P < 0.01). Conclusion Tuina can effectively reduce inflammatory responses and fibrosis in MTrPs tissue, and its mechanism is closely related to the inhibition of the TGF-β1/Smad3 signaling pathway, which plays a critical role in Tuina-mediated regulation of MTrPs fibrosis.

Objective: To investigate the microbial mechanisms of Banxia Xiexin Decoction (半夏泻心汤, BXXXD) in the treatment of esophageal precancerous lesions. Methods: A total of 30 specific pathogen-free (SPF) grade female C57BL/6J mice were randomly assigned to a control group (n = 6) and a 4-nitroquinoline 1-oxide (4-NQO)-exposed group (n = 24). Esophageal precancerous lesions were induced by providing the 4-NQO-exposed group with 4-NQO in drinking water (100 μg/mL) for 17 consecutive weeks, whereas control group received sterile drinking water. After model establishment, the mice in 4-NQO-exposed group were further randomized into model group and three BXXXD-treated groups: low-dose (BXXXD-L, 3.7 g/kg), medium-dose (BXXXD-M, 7.4 g/kg), and high-dose (BXXXD-H, 14.8 g/kg) groups (n = 6 per group). During the subsequent intervention period, mice in control and model groups were gavaged with sterile water, while mice in BXXXD groups were gavaged once daily with the corresponding dose of BXXXD aqueous extract for 4 weeks. Histopathological changes in esophageal tissues were observed by hematoxylin and eosin (HE) staining. The fecal and esophageal microbiota were profiled via 16S rDNA high-throughput sequencing to evaluate bacterial diversity, community structure, and co-occurrence networks. BXXXD chemical fingerprints were analyzed using ultra-high-performance liquid chromatography coupled with quadrupole QExactive Orbitrap mass spectrometry (UHPLC-QE-MS). Serum short-chain fatty acids (SCFA) level was quantified by targeted metabolomics using gas chromatography-mass spectrometry (GC-MS). Transcriptomic analysis of esophageal tissues was performed to assess gene expression profiles. Results: Compared with model group, BXXXD-M group exhibited reduced mucosal hyperplasia and more orderly epithelial cell arrangement, with superior therapeutic effects in comparison with both BXXXD-L and BXXXD-H groups (P < 0.01). Microbiota analysis revealed that BXXXD increased the abundance of beneficial Enterococcus and reduced pathogenic Escherichia-Shigella in the esophagus. In the gut, BXXXD elevated the relative abundance of beneficial taxa, including Lactobacillus, Dubosiella, Bacteroides, and Faecalibacterium. Targeted metabolomics showed that BXXXD significantly reduced total serum SCFA level (P < 0.01). Transcriptomic analysis indicated that BXXXD downregulated the expression of genes associated with the progression, migration, and invasion of esophageal cancer, which were identified as kallikrein-related peptidase 6 (Klk6), defensin beta 4 (Defb4), family with sequence similarity 3 member B (Fam3b), carboxypeptidase A4 (Cpa4), serum amyloid A1 (Saa1), and chitinase-like 1 (Chil1) (P < 0.05). Conclusion BXXXD may reduce the expression levels of esophageal cancer-related genes and improve esophageal precancerous lesions through modulation of the gut microbiota and metabolites.