Objective The technique of image analysis was introduced to the study of tongue inspection of Chinese medicine to improve the level of quantification

objectification and repeatability of tongue inspection. Methods The representative regions of tongue body and coating were separated and collected automatically in the model of tongue manifestations. Eight characteristic quantity variables of tongue manifestations

including colour

shape

wetness-dryness

etc.

were computed and the tongue manifestations were analyzed in 49 patients with cerebrovascular diseases and 39 health people respectively. Results The correct rate reached 87.5% in the k-means and 92% in the method of artificial neural network respectively

which was obviously higher than that in the classification method according only to the colour characteristics of tongue body or coating. Conclusion Above eight characteristics reflect comprehensively the characteristics of tongue manifestations and are good for improving the application and research of the techniques of image analysis in tongue inspection of Chinese medicine.