In this research we introduce a regularization based feature selection algorithm to benefit from sparsity and feature grouping properties and incorporate it into the medical image classification task. Using this group sparsity (GS) method, the whole group of features are either selected or removed. The basic idea in GS is to delete features that do not affect the retrieval process, instead of keeping them and giving these features small weights. Therefore, GS improves system by increasing accuracy of the results, plus reducing space and time requirements needed by the system.