Stochastic Brain Function Revealed by Individual Trial Analysis of Functional Magnetic Resonance Imaging

Conventional functional magnetic resonance imaging (fMRI) using blood oxygen level dependent contrast relies on spatiotemporal signal averaging to detect activation. The use of averaging requires that the signal of interest consists of a fixed, reproducible hemodynamic response when presented with identical stimuli. This assumption is examined with a novel, nonparametric, wavelet-based approach, termed specified resolution wavelet analysis, which precludes the need for averaging under specialized conditions. High resolution voxels (approaching the size of cortical columns) centered in the primary visual cortex were examined with a simple, event-related visual checkerboard paradigm. In the absence of averaging, even highly active voxels show inconsistent activation to identical stimuli. This nondeterministic pattern of activation could not be explained by fluctuations in behavior, physiology, movement, test insensitivity, or any other of a number of common confounds in conventional fMRI experiments. An alternative model of brain activation based on stochastic recruitment from a pool of redundant functional units is advanced to explain this inconsistency. This model is consistent with conventional approaches to fMRI interpretation in the limit of spatiotemporal averaging. The implications of this model for neuroplasticity and behavioral consistency are discussed.