深度学习舌诊模型在消化道疾病中的应用研究进展

Research Progress on the Application of Deep Learning Tongue Diagnosis Models in Gastrointestinal Diseases

摘要: 舌诊作为中医四诊中望诊的代表手段之一，可令医者做到司外揣内，然而传统舌诊常受限于医生的经验和主观性而无法充分发挥其作用，基于深度学习构建的各类舌诊模型的出现为上消化道疾病的早期筛查和辨证施治提供了智能化支持。本文探究舌诊客观化发展历程，检索了近10年来发表的高质量研究，对舌象采集及处理、模型网络架构、模型优化等方面的进展及深度学习舌诊模型在消化道疾病中的应用研究进行了系统综述，发现现有研究多仍存在如样本量受限、多模态模型较少等不足，未来需朝着多中心、大样本、标准化采集迈进，可将舌诊模型与胃镜图像结合综合进行预测，提高智能舌诊模型的预测精度，还能进一步引入时序卷积网络(TCN)捕捉长期的舌象演变趋势，实现疗效与预后的评估，推动医学与人工智能学科交叉应用的发展。

Abstract: Tongue diagnosis, as one of the representative methods of inspection in the four diagnostic methods of traditional Chinese medicine, enables doctors to infer internal conditions from external observations. However, traditional tongue diagnosis is often limited by the doctor’s experience and subjectivity, thus failing to fully exert its potential. The emergence of various tongue diagnosis models based on deep learning has provided intelligent support for the early screening and syndrome differentiation treatment of upper gastrointestinal diseases. This paper explores the development process of objective tongue diagnosis, retrieves high-quality studies published in the past decade, and conducts a systematic review of the progress in tongue image acquisition and processing, model network architecture, model optimization, and the application research of deep learning tongue diagnosis models in digestive diseases. It is found that existing studies still have shortcomings such as limited sample size and few multimodal models. In the future, we should move towards a multi-center, large-sample, and standardized collection approach. We can combine the tongue diagnosis model with gastroscopy images for comprehensive prediction, thereby improving the prediction accuracy of the intelligent tongue diagnosis model. Additionally, we can further introduce the temporal convolutional network (TCN) to capture the long-term evolution trend of tongue conditions, enabling the assessment of therapeutic effects and prognosis, and promoting the development of interdisciplinary applications between medicine and artificial intelligence.