Answers are grounded in this paper and cite their sources.
Systematic review assesses deep learning's effectiveness in TCM, revealing significant validation gaps.
Traditional Chinese Medicine (TCM) diagnosis is characterized by substantial subjectivity and inter-practitioner variability. Although deep learning (DL) offers potential for standardization and intelligent diagnostic support, its methodological rigor and clinical applicability remain unclear. This systematic review evaluated the applications, methodological quality, and translational readiness of DL in TCM disease diagnosis. Eight databases were searched from January 2010 to May 2026, and 21 studies were included. Quality assessment was conducted using TRIPOD-AI and QUADAS-2. The review identified a technological evolution from conventional classification models toward knowledge graph-enhanced frameworks, large language models, and multimodal integration. However, major methodological limitations were observed. The mean TRIPOD-AI reporting rate was only 50.5%, with deficiencies in transparency and validation reporting. QUADAS-2 assessment showed high risk of bias in the Index Test and Reference Standard domains in 47.6% of studies, and none performed independent external validation. In conclusion, despite rapid technological advances, DL-based TCM diagnostic research remains limited by insufficient validation, poor reporting transparency, and reference standard bias. Future research should prioritize multi-center external validation, prospective real-world testing, standardized reference standards, and theory-driven explainable AI approaches.
No takes yet. Share an insight, caveat, or question.
Lu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: