Deep Learning Applications in Tongue Diagnosis

Deep learning techniques have revolutionised the automation and precision of tongue-based diagnostic systems, particularly in the context of Traditional Chinese Medicine (TCM) and broader medical screening. High-resolution digital imaging, combined with advanced neural architectures, enables objective assessment of tongue colouration, texture and geometry—attributes historically evaluated subjectively by practitioners. Deep convolutional neural networks (CNNs) extract hierarchical features from raw images, facilitating tasks such as segmentation of the tongue region, quality control of image acquisition, classification of pathological patterns and prediction of systemic conditions. Recent progress encompasses end-to-end frameworks that localise and delineate tongue boundaries, pipelines that assess and filter poor-quality images for database construction and multi-task models that simultaneously identify anatomical landmarks and diagnose disease indicators. These methods not only enhance consistency and reduce inter-observer variability but also pave the way for large-scale screening and remote assessment of health status across diverse populations.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Deep Learning Applications in Tongue Diagnosis publication trend

The graph below shows the total number of articles in deep learning applications in tongue diagnosis across all publications each year (not limited to Nature Index journals).

Technical terms

Convolutional neural network (CNN): A deep learning architecture that applies convolutional filters to image data to extract hierarchical features.

U-Net: A CNN architecture with symmetric encoder–decoder paths and skip connections, designed for precise image segmentation.

Segmentation: The process of partitioning an image into distinct regions, such as isolating the tongue from its background.

Region of interest (ROI): A specific subset of an image flagged for detailed analysis, often based on initial localisation steps.

Morphological processing layer: A computational layer that applies shape-based operations (e.g., dilation, erosion) to refine segmentation outputs.

References

1. Tongue image quality assessment based on a deep convolutional neural network. BMC Medical Informatics and Decision Making (2021).
2. TongueNet: A Precise and Fast Tongue Segmentation System Using U-Net with a Morphological Processing Layer. Applied Sciences (2019).
3. Tonguenet: Accurate Localization and Segmentation for Tongue Images Using Deep Neural Networks. IEEE Access (2019).