ICA-based sparse feature recovery from fMRI datasets

摘要

Spatial Independent Components Analysis (ICA) is increasingly used in thecontext of functional Magnetic Resonance Imaging (fMRI) to study cognition andbrain pathologies. Salient features present in some of the extractedIndependent Components (ICs) can be interpreted as brain networks, but thesegmentation of the corresponding regions from ICs is still ill-controlled.Here we propose a new ICA-based procedure for extraction of sparse featuresfrom fMRI datasets. Specifically, we introduce a new thresholding procedurethat controls the deviation from isotropy in the ICA mixing model. Unlikecurrent heuristics, our procedure guarantees an exact, possibly conservative,level of specificity in feature detection. We evaluate the sensitivity andspecificity of the method on synthetic and fMRI data and show that itoutperforms state-of-the-art approaches.