Palm medicine is an important part of Tradition Chinese Medicine (TCM), which has been widely practiced in China and southeast country of Asia. Palmprint is composed of many lines and some special structures which imply a number of diseases. In this paper a fuzzy approach is proposed to recognize one of special structures in palmprint which is a key process in automated palm diagnosis system. Firstly, a palm image is preprocessed and all palmprint lines are extracted. Secondly, the extracted palm-lines are transformed to an undirected graph according to the connection of the points on the palm-lines. Thirdly, three features are extracted from this graph and their membership functions are defined. Finally, these three features are utilized to recognize one special structure which is called mi structure. Applying our approach to 200 palmprint images, the experimental results are encouraging.