A Review of Research on Computer Vision-Based Analysis and Diagnosis of Tongue Patterns in Traditional Chinese Medicine

Abstract

As the key method of TCM syndrome differentiation and treatment, objectification and intelligence of tongue diagnosis are the core of TCM modernization. This paper systematically reviews the technical development in the field of tongue diagnosis analysis and diagnosis of traditional Chinese medicine based on computer vision, and constructs a three-dimensional framework to understand the development of this field from the perspective of paradigm shift. The research combs the research status at home and abroad, and analyzes the progress in tongue image acquisition and preprocessing, feature extraction and analysis, and diagnosis model construction. Secondly, according to the technological generation, the evolution of research methods is divided into three stages: traditional image processing stage, deep learning stage, multimodal fusion stage and cognitive computing stage, and the core methods, technical principles and evolution logic of each stage are deeply analyzed. Based on this, we find out several main problems that the research is facing now: data standardization, whether the model can be understood by people, and clinical verification. Finally, it describes the future direction from three aspects: technical trend, clinical application expansion and standardization development. This research shows that this field is changing from the traditional mode to the cognitive computing mode driven by data and knowledge. Building high-quality multi-modal data sets and understandable dual-drive models is a specific aspect that needs to be paid close attention to in the future.