Traditional Chinese medicine (TCM) pulse diagnosis is a non-invasive approach used to infer cardiovascular status, but its interpretation is relatively subjective, limiting reproducibility and diagnostic precision. This review summarizes progress in digitized radial pulse assessment using modern sensors and artificial intelligence (AI), and evaluates reported applications in cardiovascular screening and decision support. We searched PubMed, IEEE Xplore, and Web of Science Core Collection from inception through November 30, 2025, for studies that acquired wrist/radial pulse signals with electronic devices and applied quantitative analysis or machine learning/deep learning to characterize pulse patterns or assess cardiovascular conditions. Across the literature, pressure-sensor arrays, wearable photoplethysmography (PPG) surrogates, and hybrid platforms enabled more standardized pulse acquisition, while AI models reported promising performance for tasks such as blood pressure estimation, hypertension screening, coronary artery disease identification, heart failure risk stratification, and arrhythmia detection. However, methodological heterogeneity, limited sample sizes, inconsistent labeling standards, and insufficient external validation remain key barriers to clinical translation. Overall, AI-enhanced digital pulse diagnosis may improve the objectivity of TCM pulse assessment and complement conventional cardiovascular diagnostics, provided that future studies adopt rigorous protocols, transparent reporting, and clinically meaningful prospective validation.

The philosophy of “treating disease before its onset” is a fundamental concept of traditional Chinese medicine (TCM), permeating its diagnostic and therapeutic framework, and is central to clinical practice. However, current TCM diagnostic and treatment models for the “pre-disease to disease” window period face several limitations, including the lack of comprehensive clinical parameters, difficulties in characterizing and integrating heterogeneous multimodal data, and insufficient dynamic precision in interventions and efficacy evaluations. To address these issues, guided by Professor Candong Li’s theory of TCM stateology, this study focuses on integrating objective multimodal data. It proposes a new model for personalized TCM diagnosis and treatment targeting the “pre-disease to disease” window period. This approach first proposes the idea of restructuring the conceptual framework of “symptom” and integrating multi-source heterogeneous data at macroscopic, mesoscopic, and microscopic levels to form a three-dimensional assessment indicator system. By integrating graph neural networks, convolutional neural networks, attention mechanisms, and knowledge graph-guided weight allocation, this approach enables collaborative representation, alignment, and fusion of multi-source data. Subsequently, it plans to construct a multimodal fusion model at both feature and decision levels, in order to establish mappings between indicators and TCM state elements, and to screen key indicators characterizing pathological evolution during the window period. Furthermore, it proposes a technical path for enhancing model interpretability using methods such as SHapley Additive exPlanations (SHAP) and Ablation-CAM++. Finally, with state assessment as the core, it proposes the concept of constructing a dynamic evaluation method for individualized diagnosis and treatment based on time-series data analysis using algorithms such as long short-term memory (LSTM) networks and gated recurrent units (GRUs). Moreover, a causal inference framework and semi-supervised learning strategies are introduced to enable quantitative evaluation of individual intervention effects and to provide interpretable therapeutic feedback, forming a complete technical path from data representation and fusion, weight adjustment, and interpretability analysis, to dynamic diagnosis feedback. This study aims to address deficiencies in the current TCM diagnosis and treatment model during the “pre-disease to disease” window period and to provide an operational framework for the clinical practice of TCM’s “treating disease before its onset”.

Objective: To explore whether digital facial and tongue diagnostic technologies can support the assessment of coronary heart disease (CHD) patients for coronary artery stenosis severity, and examine potential associations between digital tongue diagnosis features and myocardial biomarkers. Methods: The TFDA-1 face and tongue diagnosis instrument and the TDAS analysis system were used to perform intelligent visual examination and analysis of the facial and tongue in CHD patients who attended the Department of Cardiology at Shanghai Baoshan Hospital of Integrated Traditional Chinese and Western Medicine between October 2, 2023 and July 31, 2024. Variables were screened using principal component analysis (PCA) and multicollinearity analysis to construct four machine learning models, including random forest, LightGBM, decision tree, and naive Bayes, for the early prediction of coronary artery stenosis severity. Model performance metrics, including sensitivity, specificity, precision, F1 score, accuracy, and the area under the receiver operating characteristic (ROC) curve (AUC), were evaluated. Visual analyses were performed using the SHapley Additive exPlanations (SHAP) interpreter and decision curve analysis. For patients after percutaneous coronary intervention (PCI), a conceptual model linking cardiac biomarkers and tongue diagnosis was constructed using the partial least squares structural equation modeling (PLS-SEM), and its validity was assessed. Results: A total of 459 CHD patients were enrolled and assigned to a PCI group and a non-PCI group (which comprised two subgroups: mild stenosis or less group, moderate stenosis or greater group). For sublingual vein (SV) features, the PCI group had lower SV-a and SV-b than the other groups (P < 0.01 and P < 0.05, respectively). For tongue surface features, the PCI group had significantly higher tongue body (TB)-L, TB-a, and TB-b (P < 0.05, P < 0.01, and P < 0.001, respectively), as well as higher tongue coating (TC)-a and TC-b (P < 0.01 and P < 0.001, respectively). Age, SV-a, SV-b, creatine kinase-myocardial band (CK-MB), CK, TC-a, lip-L, and lip-b were incorporated in the machine learning models. The random forest model performed best, with an AUC of 0.924, an F1 score of 0.839, precision of 0.807, accuracy of 0.864, sensitivity of 0.873, and specificity of 0.839. Decision curve analysis indicated that both LightGBM and random forest had clinical utility. PLS-SEM confirmed the pathway relationships: myocardial biomarkers → TB and myocardial biomarkers → TC (coefficient = – 0.238, t = 2.239, P = 0.025, and coefficient = – 0.270, t = 2.522, P = 0.012, respectively). Conclusion This study developed a noninvasive early warning model for coronary artery stenosis in patients with CHD. It applied PLS-SEM to investigate the association between post-PCI cardiac biomarkers and tongue diagnosis, and validated the proposed association chain. These findings suggest that intelligent traditional Chinese medicine (TCM) visual diagnosis integrated with modern digital technology may support CHD risk assessment and comprehensive health management.

Objective: To address the lack of fine-grained clinical recognition for specific Yang deficiency syndrome subtypes and the limitations of conventional object detection models in extracting irregular, low-contrast tongue phenotypes. This study aims to develop an objective subtype recognition framework based on an improved You Only Look Once nano (YOLO11n) architecture, using a standardized visual phenotype matrix to translate macroscopic traditional Chinese medicine (TCM) descriptions into quantifiable clinical targets. Methods: This cross-sectional diagnostic study consecutively enrolled adult inpatients admitted to the Department of Thyroid and Breast Surgery, The First Affiliated Hospital of Wannan Medical University (Yijishan Hospital), between September 1, 2024 and June 1, 2025, who were suspected of having Yang deficiency constitution based on initial TCM consultation. Clinical tongue image data were collected for analysis. Based on an Expert Visual Phenotype Annotation Matrix, a five-category recognition system was established, including the following TCM syndrome subtypes: spleen-dampness exuberance syndrome, mild kidney Yang deficiency syndrome, upper heat and lower cold syndrome, simultaneous Yin-Yang deficiency syndrome, and Yin deficiency and fluid depletion syndrome (negative control). The proposed Yang deficiency YOLO (YD-YOLO) model, built upon the YOLO11n baseline, integrates the Cross Stage Partial with kernel size 2 (C3k2)-GhostBottleneck-Dynamic Convolution (GBDC) module into the backbone to adaptively extract low-contrast features, and embeds the multipath aggregation coordinate attention (MACA) mechanism into the neck to suppress background interference through multi-scale spatial coordination. Gradient-weighted class activation mapping (Grad-CAM) was used to visualize feature attribution and evaluate the biological plausibility of the model’s focus. Model performance was evaluated through ablation and comparative experiments using mean average precision (mAP), precision, recall, F1 score, inference speed (frames per second, FPS), overall accuracy, Cohen’s kappa, and the area under the receiver operating characteristic (ROC) curve (AUC). Results: Based on the final inclusion of 1 186 clinical cases, the YD-YOLO model had an overall accuracy of 91.5%, a Cohen’s kappa of 0.912, and an mAP@50 of 0.731 [higher than the YOLO11n baseline (0.681)], with AUC ranging from 0.91 to 0.97 across all TCM syndrome subtypes. Among the TCM syndrome subtypes, the mild kidney Yang deficiency syndrome had the highest mAP@50 (0.900), and the inference speed reached 89.00 FPS. Grad-CAM analysis showed that the model localized activation to key TCM pathological features, such as marginal tooth marks and focal root coatings, while suppressing non-diagnostic oral background noise. Conclusion The YD-YOLO model demonstrates the feasibility of deep learning for the fine-grained classification of TCM Yang deficiency subtypes. By integrating visual phenotype quantification with model interpretability, the proposed framework provides an objective basis for syndrome differentiation, supporting the development of standardized digital diagnostic systems and the provision of clinical decision support in TCM practice.

Yin-Yang theory is one of the core components of the foundational theory of traditional Chinese medicine (TCM), yet its classification as philosophical or scientific has long been controversial. To move beyond the longstanding either-philosophy-or-science debate surrounding Yin-Yang theory in TCM, this article, for the first time, proposes a three-level framework spanning macro, meso, and micro perspectives. At the macro level, Yin and Yang are a theoretical achievement of ancient Chinese natural philosophy, used to explain universal laws governing how the cosmos operates, which constitutes Yin-Yang theory in philosophy; at the meso level, Huangdi Neijing (《黄帝内经》, Inner Canon of Huangdi) translated Yin-Yang from a cosmological framework into medical discourse, making it a core methodological basis for clinical pattern differentiation and treatment, which constitutes Yin-Yang theory in medicine; at the micro level, Yin-Yang theory in TCM has increasingly converged with modern science, showing scientific features that are testable and reproducible, which constitutes Yin-Yang theory in science. The article further demonstrates that Yin-Yang theory in TCM differs markedly from Yin-Yang as a general philosophical doctrine in core application and practical orientation: in the course of its medical transformation, elements related to life phenomena were selectively incorporated, while, within a medical context, abstract propositions such as cosmological speculation were bracketed and rendered concrete, thereby achieving a practice-oriented transition from philosophy to medicine. Based on this, the present study conducts analyses from four aspects, namely philosophical roots, clinical application, modern scientific interpretation, and implications for life sciences; it substantiates the logical basis for Yin-Yang theory in TCM as the overarching framework for clinical pattern differentiation and treatment, reviews modern research progress from perspectives such as systems science and network regulation, and explores its potential value in advancing a new paradigm of state-based medicine. The study suggests that an accurate understanding of Yin-Yang theory in TCM requires moving beyond the either-philosophy-or-science binary and analyzing specific issues on a case-by-case basis. Across different perspectives, Yin-Yang theory in TCM has philosophical, medical, and scientific attributes. Only by grounding the theory in actual clinical efficacy and engaging with modern science can the innovative development of Yin-Yang theory in TCM be promoted, thereby providing valuable Eastern insights for building a life science system with original Chinese contributions.

Objective: To map the research landscape of artificial intelligence (AI)-assisted tongue diagnosis through bibliometric analysis and to quantify its diagnostic accuracy and clinical interpretability through a diagnostic test accuracy (DTA) meta-analysis. Methods: For the bibliometric analysis, the Web of Science Core Collection (WoSCC) was queried for English-language articles and reviews on AI-assisted tongue diagnosis published between January 1, 2014 and December 31, 2025, and analysed using Bibliometrix, VOSviewer, and CiteSpace, with major output dimensions including annual publication output and disciplinary distribution, journal and citation characteristics, country/region and institutional collaboration, author networks, keyword co-occurrence, and keyword burst detection. For the DTA meta-analysis, four databases [Scopus, PubMed, Web of Science, and China National Knowledge Infrastructure (CNKI)] were searched in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses of Diagnostic Test Accuracy (PRISMA-DTA) guidelines. A bivariate random-effects model hierarchical summary receiver operating characteristic (HSROC) was used to pool sensitivity and specificity, with subgroup analyses by disease category, AI model architecture, and sample-size strata. Methodological quality was assessed with the Quality Assessment of Diagnostic Accuracy Studies version 2 (QUADAS-2) tool, and publication bias was evaluated by Deeks’ funnel plot asymmetry test. Results: A total of 198 publications met the bibliometric eligibility criteria. Annual output increased 24.5-fold (from 2 in 2014 to 49 in 2025), with the period 2022 – 2025 alone accounting for 65.2% of all publications. China contributed approximately 83.5% of all institutional affiliations, with Shanghai University of Traditional Chinese Medicine and Jiatuo Xu being the most productive institution and author, respectively. Keyword analysis identified four thematic clusters (AI and deep-learning architectures, image processing and segmentation, traditional Chinese medicine (TCM)-specific applications, and disease-specific applications) and a temporal evolution from traditional machine learning to deep learning and transformer-based, explainable, and multimodal AI architectures. Sixteen DTA meta-analysis studies (14 755 participants) covering metabolic and hepatic disorders, oncological and oral lesions, cardiovascular risk, diabetes, and other clinical applications were included in the DTA meta-analysis. The pooled sensitivity was 90.3% [95% confidence interval (CI): 86.7% – 93.1%] and the pooled specificity was 93.0% (95% CI: 90.6% – 94.7%); the area under the summary receiver operating characteristic (SROC) curve (AUC) was 0.961. Heterogeneity was substantial (I2 = 95.8% for sensitivity; I2 = 92.1% for specificity). Subgroup performance was broadly consistent across disease categories, AI architectures, and sample-size strata, and Deeks’ test indicated no significant publication bias (P = 0.258). Conclusion AI-assisted tongue diagnosis has progressed rapidly and shows pooled diagnostic performance comparable to established screening modalities, supporting its potential as a complementary and easily accessible decision-support tool.

Objective: To elucidate the biological basis of traditional Chinese medicine (TCM) syndromes from the perspective of “same syndrome, different diseases” in patients with chronic hepatitis B (CHB), liver cirrhosis (LC), and hepatocellular carcinoma (HCC), thereby providing a complementary approach for the diagnosis and treatment of chronic liver diseases (CLD). Methods: To investigate the dynamic characteristics of TCM syndromes in CLD, transcriptomic profiling of peripheral blood mononuclear cells (PBMCs) was performed from patients with CHB, LC, or HCC presenting with three TCM syndromes: liver gallbladder dampness heat syndrome (LGDHS), liver depression spleen deficiency syndrome (LDSDS), and liver kidney Yin deficiency syndrome (LKYDS). These participants were recruited at Shuguang Hospital Affiliated to Shanghai University of Traditional Chinese Medicine between August 1, 2018 and December 31, 2021. Differentially expressed genes (DEGs) were identified using the random variance model (RVM) F test with false discovery rate (FDR) correction. Principal component analysis (PCA) and unsupervised hierarchical clustering were applied to visualize sample grouping. Dynamic network biomarkers (DNB) analysis was employed to detect critical transition stages during syndrome evolution, followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses to characterize the functional roles and pathway involvement of the DNB members. Random forest (RF) analysis and the area under the receiver operating characteristic (ROC) curves (AUC) were used to rank the importance of candidate genes. External validation was performed using microarray data from an independent CLD cohort (GSE89377) and RNA-seq data from The Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA-LIHC) dataset. Additionally, reverse transcription quantitative polymerase chain reaction (RT-qPCR) was performed on an independent cohort of LC patients to validate the expression levels of the candidate genes. Results: The study included a total of 132 participants. DNB analysis identified LDSDS stage as a critical tipping point during TCM syndrome evolution across CHB, LC, and HCC. The phosphoinositide 3-kinase/protein kinase B (PI3K-AKT) signaling pathway was consistently enriched in the DNB analysis across all three types of CLD, suggesting its potential involvement in the critical transition of TCM syndromes. Among the 24 core DNB members of the PI3K-AKT pathway, four genes—integrin subunit beta 1 (ITGB1), collagen type IV alpha 1 chain (COL4A1), collagen type IV alpha 2 chain (COL4A2), and DNA damage inducible transcript 3 (DDIT3)—were identified by RF analysis (Gini score > 1) and ROC analysis. ROC analysis demonstrated high discriminative ability for distinguishing LGDHS from LKYDS in CHB patients, with AUC of 0.7891 for ITGB1, 0.7070 for COL4A1, 0.7148 for COL4A2, and 0.8945 for DDIT3. In the independent CLD cohort (GSE89377), all four genes showed significant stepwise upregulation from normal to CHB, LC, and HCC (all P < 0.05). In the TCGA-LIHC dataset, their expression progressively increased with tumor stage. RT-qPCR validation in an independent LC cohort (30 LGDHS vs. 30 LKYDS) confirmed that ITGB1, COL4A2, and DDIT3 were significantly upregulated in LKYDS compared with LGDHS (P = 0.0152, 0.0186, and 0.0247, respectively), whereas COL4A1 showed a non-significant upward trend (P = 0.1201). Conclusion This study introduces a novel approach to understanding the molecular features underlying TCM syndrome evolution in CLD. The PI3K-AKT pathway and four identified genes (ITGB1, COL4A1, COL4A2, and DDIT3) play crucial roles in the transition from excess (LGDHS) to deficiency (LKYDS) via the critical LDSDS stage. These findings offer potential quantitative biomarkers and therapeutic targets for TCM syndrome differentiation and may help arrest syndrome progression in CLD.

Objective: o perform a comparative evaluation of Mamdani and Takagi-Sugeno (TS) fuzzy inference systems for predicting peripheral vascular resistance from photoplethysmogram (PPG) signals, incorporating diagnostic parameters from Persian medicine (PM) pulsology. Methods: Both fuzzy inference systems were implemented in MATLAB R2021b and validated using leave-one-out cross-validation (LOOCV) on clinical PM pulse diagnostic data collected from 35 healthy volunteers. The dataset included pulse frequency scale (1 – 6) and weakness scale (1 – 4) alongside corresponding PPG-derived peripheral vascular resistance indices (range: 0.019 – 0.983). The Mamdani system (with 35 rules) was designed using trapezoidal and triangular membership functions, a singleton fuzzifier, a product inference engine, and a centroid defuzzifier. The TS system (with 6 rules) was configured as a first-order model with Gaussian input membership functions. System performance was quantitatively evaluated using mean absolute error (MAE), root mean square error (RMSE), and the coefficient of determination (R2). Statistical significance of performance differences was assessed using a paired t test. Results: The Mamdani inference system showed much higher prediction accuracy than the TS system. Comparative analysis revealed a substantial advantage for the Mamdani system (MAE = 0.007 63 ± 0.001 20, RMSE = 0.008 58 ± 0.001 50, R2 = 0.998 40 ± 0.000 80) over the TS system (MAE = 0.015 40 ± 0.002 10, RMSE = 0.022 48 ± 0.002 80, R2 = 0.989 20 ± 0.001 50). A paired t test comparing the absolute errors of the 35 LOOCV folds confirmed statistical significance [t (34) = 5.07, P < 0.001]. Conclusion The findings suggest the potential suitability of the Mamdani system for certain precision-oriented analytical tasks. Both systems exhibit practical utility for integrative medicine diagnostics and wearable health-monitoring applications for healthy individuals, enabling a modern computational translation of PM pulsology.

Objective: To address the challenges of systematically identifying herb pairs in traditional Chinese medicine (TCM), we proposed HerbGL, a framework for predicting potential herb pairs that integrates network propagation and graph regularization. Methods: Based on the assumption that herbal actions induce subtle perturbations in biological systems, a framework named HerbGL was proposed. Random walk with restart (RWR) was first applied to the protein-protein interaction (PPI) network to reconstruct herb-specific perturbation effects and generate weighted subnetworks. Then, to quantify affinity between herb pairs, two network-proximity metrics, Closeness and PageRank, were computed from the weighted subnetworks to construct herb-pair affinity matrices. Finally, these matrices, together with known herb pairs (derived from co-occurrence analysis of TCM formulas with a threshold determined from the inflection point of the frequency distribution), were incorporated into a graph regularization model to predict potential herb pairs. Model performance was assessed through baseline comparison, ablation and robustness experiment under different ratios of positive and negative samples, using the area under the receiver operating characteristic curve (AUROC), the area under the precision-recall curve (AUPRC), accuracy, and precision as evaluation metrics. Furthermore, the predicted herb pairs were validated through both literature evidence and Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. Results: The weighted subnetworks constructed by RWR provided a refined simulation of herb-specific perturbation effects, which formed the basis for subsequent affinity modeling and prediction. Analysis of herb pair co-occurrence frequencies revealed a marked change around 150, which was selected as the threshold to distinguish herb pairs from non-herb pairs. HerbGL exhibited superior predictive performance compared with baseline models (AUROC = 0.970 5, AUPRC = 0.955 5, accuracy = 0.726 6, precision = 0.970 6). Ablation results showed that removing the Closeness and PageRank metrics substantially degraded performance (AUROC = 0.819 1, AUPRC = 0.876 8), confirming their necessity. Robustness evaluation under an imbalanced positive-to-negative sample ratio of 1 : 5 yielded AUROC = 0.969 6 and AUPRC = 0.840 4, indicating stable predictive ability. Moreover, multiple case studies further validated the rationality of the predicted herb pairs, such as Fangfeng (Saposhnikoviae Radix) and Qingpi (Citri Reticulatae Pericarpium Viride) which are recorded in Liangpeng Huiji (《良朋汇集》, Collection of Excellent Recipes) Vol. 3: Fangfeng Shengma Tang (防风升麻汤). Additionally, pathway enrichment analysis of the Renshen (Ginseng Radix et Rhizoma) and Lianqiao (Forsythiae Fructus) pair further supported the biological plausibility of their compatibility. Conclusion HerbGL offers an effective and biologically informed framework for identifying herb pairs in TCM. Beyond improving herb pair prediction, the framework also provides data support for research on herb compounds and mechanisms, thereby supporting data-driven exploration of TCM compatibility.

Objective: This study proposes a clustering framework for Chinese materia medica (CMM) based on a large language model (LLM), aiming to explore potential compatibility patterns among CMMs from the semantic perspective of CMM property theory. Methods: First, a CMM property knowledge base was constructed based on Chinese Materia Medica, including 567 commonly used CMMs characterized by four properties, five flavors, and meridian tropism. Then, 49 CMMs derived from 10 prescriptions for Zangdu (脏毒, pathogenic toxins) recorded in Waike Zhengzong (《外科正宗》, Orthodox Manual of External Medicine) and Yangke Xinde Ji (《疡科心得集》, Collected Insights on Ulcer Medicine) were selected as the experimental dataset. Five semantic representation methods—One-Hot, Word2Vec, Bidirectional Encoder Representations from Transformers (BERT), Beijing Academy of Artificial Intelligence General Embedding (BGE), and Qwen—were applied to encode CMM property information into vector representations. Subsequently, t-distributed Stochastic Neighbor Embedding (t-SNE) was used for nonlinear dimensionality reduction on high-dimensional semantic vectors, followed by k-means clustering (k = 7). Clustering performance was evaluated using the Silhouette Score (SS), Davies-Bouldin Index (DBI), and Calinski-Harabasz Index (CHI). Results: The Qwen-based clustering method, CMM-EmbedCluster, achieved the highest SS (0.607 4) and CHI (158.057 2), as well as the lowest DBI (0.499 5), indicating improved cluster separation and compactness compared with other methods. Visualization of CMM clustering results showed that the clusters were well separated in the low-dimensional space, with strong inter-cluster discrimination and high intra-cluster functional consistency. Further interpretability analysis of CMM clustering results revealed stable structural differences among clusters in terms of four properties, five flavors, and meridian tropism, forming functional partitions consistent with CMM property theory. Conclusion CMM-EmbedCluster utilizes an LLM to achieve semantic-level representation and clustering of CMMs within the framework of CMM property theory, providing support for exploring potential compatibility patterns among CMMs from the perspective of CMM property semantics.