Abstract: In view of brain tumor images, we proposed a retrieval method based on content. Firstly, we augmented the region of interest (ROI) of brain tumor images by using the image expansion algorithm. Secondly, based on the feature extraction method of pointbased spatial pixel intensity, we described the local characteristics of the augmented ROI. Finally, we introduced the aggregated feature algorithm to improve the discrimination and compression ability of image features. The results of the experimental comparison of three brain tumor image retrieval methods show that the algorithm can obtain high retrieval accuracy and prove that the algorithm can achieve efficient and accurate brain tumor image retrieval.

Key words: brain tumor image retrieval, aggregated feature, spatial pixel intensity, contentbased image retrieval, bagofwords model