Shift-invariant rank-(L, L, 1, 1) BTD with 3D spatial pooling and orthonormalization: Application to multi-subject fMRI data

Li-Dan Kuang, Hao-Peng Zhang, Hao Zhu, Shiming He, Wenjun Li, Yan Gui, Jin Zhang, Jianming Zhang

Abstract: Highlights:•A novel shift-invariant rank-(L, L, 1, 1) BTD with 3D spatial pooling and orthonormalization for multi-subject fMRI data separation.•Two loading matrices of shared SMs, shared TCs, subject-specific time delays and strengths can be decomposed by the proposed method.•The proposed 3D weighted spatial pooling preprocessing compresses and smooths multi-subject fMRI data, and assigns a higher weight to in-brain voxels but a lower weight to out-brain voxels, which reduces the size and improves robustness to noise.•To deal with the high spatiotemporal variability, the proposed methods relaxes the rank-(L, L, 1, 1) BTD model of the reduced fMRI data by incorporating temporal shift-invariance and spatial orthonormality constraints.