From convolutional neural networks to large foundation models: A systematic review of deep learning for intelligent tongue diagnosis.

Abstract

As a core component of inspection in Traditional Chinese Medicine (TCM), the tongue contains rich physiological and pathological information. Advances in deep learning technology have significantly enhanced the accuracy and interpretability of intelligent tongue diagnosis. This paper presents a systematic review of intelligent tongue diagnosis based on deep learning. First, in terms of methodology, we elaborate on four mainstream paradigms of intelligent tongue diagnosis, examining the algorithmic models, technical principles, and applicable scenarios within each paradigm. Second, regarding task types, we categorize the research into three mainstream diagnostic tasks, clarifying the technical characteristics and typical algorithmic choices for each. Third, we investigate two major application domains of intelligent tongue diagnosis and detail the specific applications within these sub-fields. Finally, by synthesizing existing research with recent trends, we outline future research directions for tongue diagnosis. This review establishes a systematic foundation for building clinically trustworthy and deployable intelligent tongue diagnosis frameworks.