TY  - JOUR
AU  - Gao, Longfei
AU  - Jin, Xuebo
PY  - 2026
DA  - 2026/03/10
TI  - A Tongue Image Dataset with Pathological Annotations for AI-assisted Diagnosis in Traditional Chinese Medicine
JO  - Biomedical Informatics and Smart Healthcare
T2  - Biomedical Informatics and Smart Healthcare
JF  - Biomedical Informatics and Smart Healthcare
VL  - 2
IS  - 1
SP  - 5
EP  - 19
DO  - 10.62762/BISH.2026.303296
UR  - https://www.icck.org/article/abs/BISH.2026.303296
KW  - tongue diagnosis
KW  - traditional Chinese medicine
KW  - medical image dataset
KW  - deep learning
KW  - object detection
AB  - Tongue diagnosis is a core component of Traditional Chinese Medicine (TCM) with important clinical application value, yet its standardization is severely hampered by the subjectivity of manual interpretation and the lack of unified imaging acquisition protocols. Worse still, the scarcity of large-scale annotated datasets has become a key bottleneck restricting the development of artificial intelligence (AI)-assisted TCM tongue diagnosis technology. To address these critical issues, this study constructs a high-quality standardized dataset dedicated to AI-driven TCM tongue diagnosis research. The dataset contains 6,719 high-resolution tongue images collected under strictly standardized conditions, and all images are annotated with 20 pathological symptom categories in line with TCM theoretical systems. Each image is attached with an average of 2.54 clinical labels, all of which have been double-verified and confirmed by licensed TCM practitioners to ensure clinical authenticity and annotation accuracy. In order to facilitate academic research and industrial applications, we have used three mainstream annotation formats (COCO, TXT, XML) to annotate the data, making the dataset compatible and universal. To verify the practical value and effectiveness of the dataset for AI model training, we conducted a comprehensive benchmark test on it using twelve classic deep learning detection models, including multiple variants of YOLOv5/v7/v8 as well as SSD and MobileNetV2. The experimental results fully demonstrate that the dataset can effectively support the training and performance evaluation of AI models for tongue diagnosis. As a high-quality public data resource, this dataset lays a solid and critical foundation for developing reliable computational analysis tools in the field of TCM, alleviating the long-term data shortage problem that hinders the digital development of TCM tongue diagnosis, and promoting the deep integration of AI technology with TCM research and clinical practice through standardized and high-quality diagnostic image data.
SN  - 3068-5524
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Gao2026A,
  author = {Longfei Gao and Xuebo Jin},
  title = {A Tongue Image Dataset with Pathological Annotations for AI-assisted Diagnosis in Traditional Chinese Medicine},
  journal = {Biomedical Informatics and Smart Healthcare},
  year = {2026},
  volume = {2},
  number = {1},
  pages = {5-19},
  doi = {10.62762/BISH.2026.303296},
  url = {https://www.icck.org/article/abs/BISH.2026.303296},
  abstract = {Tongue diagnosis is a core component of Traditional Chinese Medicine (TCM) with important clinical application value, yet its standardization is severely hampered by the subjectivity of manual interpretation and the lack of unified imaging acquisition protocols. Worse still, the scarcity of large-scale annotated datasets has become a key bottleneck restricting the development of artificial intelligence (AI)-assisted TCM tongue diagnosis technology. To address these critical issues, this study constructs a high-quality standardized dataset dedicated to AI-driven TCM tongue diagnosis research. The dataset contains 6,719 high-resolution tongue images collected under strictly standardized conditions, and all images are annotated with 20 pathological symptom categories in line with TCM theoretical systems. Each image is attached with an average of 2.54 clinical labels, all of which have been double-verified and confirmed by licensed TCM practitioners to ensure clinical authenticity and annotation accuracy. In order to facilitate academic research and industrial applications, we have used three mainstream annotation formats (COCO, TXT, XML) to annotate the data, making the dataset compatible and universal. To verify the practical value and effectiveness of the dataset for AI model training, we conducted a comprehensive benchmark test on it using twelve classic deep learning detection models, including multiple variants of YOLOv5/v7/v8 as well as SSD and MobileNetV2. The experimental results fully demonstrate that the dataset can effectively support the training and performance evaluation of AI models for tongue diagnosis. As a high-quality public data resource, this dataset lays a solid and critical foundation for developing reliable computational analysis tools in the field of TCM, alleviating the long-term data shortage problem that hinders the digital development of TCM tongue diagnosis, and promoting the deep integration of AI technology with TCM research and clinical practice through standardized and high-quality diagnostic image data.},
  keywords = {tongue diagnosis, traditional Chinese medicine, medical image dataset, deep learning, object detection},
  issn = {3068-5524},
  publisher = {Institute of Central Computation and Knowledge}
}