Imaging datasets typically include a large number of features (e.g. voxels or Regions of Interest, ROIs), and have relatively small sample sizes.
Including all possible features in a regression model can easily lead to overfitting and a consequent lack of generalizability. A large number of
features also make it difficult to perform feature selection, and thereby identify the relevant voxels or ROIs. We report a novel Machine
Learning method for automated data-driven feature selection with large neuroimaging dataset