Linear constraint minimum variance beamformer functional magnetic resonance inverse imaging

Abstract

Accurate estimation of the timing of neural activity is required to fully model the information flow among functionally specialized regions whose joint activity underlies perception, cognition and action. Attempts to detect the fine temporal structure of task-related activity would benefit from functional imaging methods allowing higher sampling rates. Spatial filtering techniques have been used in magnetoencephalography source imaging applications. In this work, we use the linear constraint minimal variance (LCMV) beamformer localization method to reconstruct single-shot volumetric functional magnetic resonance imaging (fMRI) data using signals acquired simultaneously from all channels of a high density radio-frequency (RF) coil array. The LCMV beamformer method generalizes the existing volumetric magnetic resonance inverse imaging (InI) technique, achieving higher detection sensitivity while maintaining whole-brain spatial coverage and 100 ms temporal resolution. In this paper, we begin by introducing the LCMV reconstruction formulation and then quantitatively assess its performance using both simulated and empirical data. To demonstrate the sensitivity and inter-subject reliability of volumetric LCMV InI, we employ an event-related design to probe the spatial and temporal properties of task-related hemodynamic signal modulations in primary visual cortex. Compared to minimum-norm estimate (MNE) reconstructions, LCMV offers better localization accuracy and superior detection sensitivity. Robust results from both single subject and group analyses demonstrate the excellent sensitivity and specificity of volumetric InI in detecting the spatial and temporal structure of task-related brain activity.