Motivation: Traditional univariate task-fMRI analysis relies on isotropic Gaussian smoothing and thus suffers from low spatial specificity. Local constraint multivariate methods could better model activation shapes but requires parameter selections that are challenging at group level. Goal(s): Improve sensitivity and specificity in task-fMRI group-level activation detection. Approach: A spatially-adaptive group-level detection method was developed to utilize oriented spatial filters for comprehensive subject-specific activation mapping and optimize filter contributions at the group level. Results: The method demonstrated superior sensitivity and spatial specificity compared to traditional isotropic Gaussian smoothing. Impact: Improve the sensitivity and specificity of task-fMRI group-level activation detection.
Zhuang et al. (Tue,) studied this question.