Deep learning in traditional Chinese medicine
Clinical Snapshot
PICO Framework
| P — Population | Not formally defined — the review encompasses TCM research broadly, including clinical practice data, medical imaging datasets, medicinal substance investigations, and TCM literature corpora |
| I — Intervention | Deep learning (DL) methods applied to TCM data domains: medical image processing, medicinal substances investigation, data fusion, and natural language processing (NLP) |
| C — Comparator | No formal comparator specified; implicitly compared to conventional computational or manual TCM research methods where discussed |
| O — Outcomes | Not formally pre-specified; narrative outcomes include DL model performance in TCM data mining tasks, identification of active compounds, image-based diagnosis support, and NLP-based literature extraction |
Bottom Line
This paper is a narrative review — not a systematic review or meta-analysis — examining how deep learning methods are being applied across four domains of traditional Chinese medicine research: medical image processing, medicinal substance investigation, data fusion, and natural language processing. While the topic is timely and the breadth of coverage is useful for orientation, the review carries significant methodological limitations. There is no structured search strategy, no pre-specified PICO question, no formal quality appraisal of included studies, and no quantitative synthesis. The risk of selection and publication bias is high. The authors themselves acknowledge that high-quality TCM data remain a critical bottleneck for meaningful DL development. For Australian clinicians, this review offers no actionable clinical guidance. DL-assisted TCM tools are not TGA-approved, not PBS-listed, and not endorsed by RACGP or CMBA clinical guidelines. The paper is best regarded as a scoping narrative for researchers at the intersection of computational methods and integrative medicine, not as evidence to inform clinical decision-making. Senior clinicians should treat its conclusions as hypothesis-generating at best.
Key Findings
P Value: Not reported
Effect Size: Not reported — no pooled effect sizes calculated; individual study metrics not systematically extracted
Primary Outcome: Narrative summary of DL applications across four TCM domains: (1) medical image processing, (2) medicinal substances investigation, (3) data fusion, and (4) natural language processing
Nnt Or Sensitivity: Not applicable — no diagnostic accuracy, therapeutic efficacy, or prognostic metrics are synthesised. Individual DL model performance figures (e.g., accuracy, AUC) from cited studies are referenced narratively but not tabulated or pooled
Confidence Interval: Not reported
Clinical Application
DL tools described (image recognition for tongue/pulse diagnosis, compound-target prediction, NLP for classical text mining) are largely at proof-of-concept or early research stage. Clinical deployment requires validated datasets, regulatory approval, and integration into existing clinical workflows — none of which are addressed in this review TCM is practised in Australia under the Chinese Medicine Board of Australia (CMBA) and regulated by AHPRA since 2012. Approximately 4,500 registered Chinese medicine practitioners operate nationally. However, this review has no direct relevance to TGA-listed TCM products, PBS-subsidised medicines, or RACGP clinical guidelines. DL-assisted TCM tools described are not TGA-approved or clinically validated for Australian practice. RACGP guidelines do not currently endorse DL-assisted TCM diagnostics. The review may be of academic interest to integrative medicine researchers at Australian universities but should not influence clinical practice without substantially higher-quality evidence. Primarily relevant to TCM researchers, computational pharmacologists, and health informaticians working within TCM frameworks. Not directly applicable to clinical patient care without substantial further validation in prospective clinical studies
Abstract
As an alternative medicine approach, traditional Chinese medicine (TCM) is mainly built on an ancient understanding of the world. In order to translate the promises of TCM into modern therapies, rigorous scientific techniques are applied, which generates big data from experiments, clinical practices and medical literature. Recently, deep learning (DL) has attracted the attention of TCM researchers due to its application in data mining. In this review, we summarized recent examples of DL application in the study of TCM, including medical image processing, medicinal substances investigation, data fusion and natural language processing. We find that DL methods have considerable potential to extract valuable information from TCM data, but high-quality TCM data are still needed for further development of DL techniques. Please cite this article as: Liu LP, Yang C, Cen SX, Suo TC. Deep learning in traditional Chinese medicine. J Integr Med. 2026; 24(4):471-480.
References
- 1.Liu, L.-P., Yang, C., Cen, S.-X., & Suo, T.-C. (2026). Deep learning in traditional Chinese medicine. Journal of Integrative Medicine, 24(4), 471–480. https://doi.org/10.1016/j.joim.2026.03.001
Related Research
Journal of molecular neuroscience : MN
Scalp Acupuncture Combined with Mouse Nerve Growth Factor Acupoint Injection for Post-stroke Cognitive Impairment
13 July 2026
BMJ supportive & palliative care
Efficacy and safety of acupuncture for breast cancer-related insomnia: a systematic review and meta-analysis
1 July 2026
Medicine
Efficacy and safety of traditional Chinese medicine decoctions as an adjuvant treatment for rheumatoid arthritis: A systematic review and meta-analysis
28 June 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service