Finer discrimination of brain activation with local multivariate distance

The organization of human brain function is diverse on different spatial scales. Various cognitive states are always represented as distinct activity patterns across the specific brain region on fine scales. Conventional univariate analysis of functional MRI data seeks to determine how a particular cognitive state is encoded in brain activity by analyzing each voxel separately without considering the fine-scale patterns information contained in the local brain regions. In this paper, a local multivariate distance mapping (LMDM) technique is proposed to detect the brain activation and to map the fine-scale brain activity patterns. LMDM directly represents the local brain activity with the patterns across multiple voxels rather than individual voxels, and it employs the multivariate distance between different patterns to discriminate the brain state on fine scales. Experiments with simulated and real fMRI data demonstrate that LMDM technique can dramatically increase the sensitivity of the detection for the fine-scale brain activity patterns which contain the subtle information of the experimental conditions.