A corner-based intention identification strategy for medical image retrieval is proposed. Since corner feature is easy to be understood by users, suitable to be stored and processed by computers, and corner feature has invariance of rotation and translation. The paper uses corners feature for medical image retrieval. By means of corners, extra pre-processing operations of human-computer Interactions can be available, which make doctors customize query conditions in a flexible manner. Doctor's intention could be understood more clearly, which could improve the accuracy and efficiency of retrieval dramatically. Experiments show that the corner-based intention identification strategy is successful.