Randomized Structural Sparse Based Feature Selection Algorithm On Brain MRI Data

Sparse models have been widely used in the analysis of medical imaging, as well as many other areas such as image processing, signal processing and computer vision. One of the challenges is the high-dimension but small-sample cases in human brain MRI data. LASSO model has been used in the variable selection problem of high dimensional data, but it can’t solve the problem of high correlations between variables, considering the special continuity of MRI data. And the problem of error control is still to be further considered.In this paper, we consider voxel selection for functional Magnetic Resonance Imaging(fMRI) brain data with the aim of finding a more complete set of probably correlated discriminative voxels, thus improving interpretation of the discovered potential biomarkers. In order to deal with the difficulty, stability selection has received a great deal of attention lately, especially due to its finite sample control of false discoveries and transparent principle for choosing a proper amount of regularization. However, it fails to make explicit use of the correlation property or structural information of these discriminative features and leads to large false negative rates. In other words, many relevant but probably correlated discriminative voxels are missed. Thus, we propose a new variant on stability selection “randomized structural sparsity”, which incorporates the idea of structural sparsity. Comparing with other state-of-the-art methods, MRI data experiments demonstrate that our method can be superior in controlling for false negatives while also keeping the control of false positives inherited from stability selection.