A content-based scheme to retrieve lung Computed Tomographic images (CT) is presented in this paper. The proposed system consists of 3 main modules. The Input Module, the Query Module and the Retrieval Module. In Input Module based on six selected regions of database images features are extracted and fed into k-means to initialize clusters. In Query Module, from the selected query image user is allowed to select six interested regions of square or rectangular or polygonal shape manually. The features of selected regions are extracted using Moment Invariants. Based on computed features of query image similar images are clustered. In the Retrieval Module, we merge clusters from the same relevance class using a minimum distance criterion. The process is repeated until the number of remaining clusters becomes relevant to the image categories in the retrieved window and the system will output a set of candidate images that are similar to the query image.