Purchase this article with an account.
Cheryl Olman; Failures of inference: challenges for interpreting localized fMRI measurements of visual features. Journal of Vision 2011;11(11):1199. doi: 10.1167/11.11.1199.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Functional MRI studies in recent years have shown that even low-resolution data reflect information represented in the neural code on a scale much smaller than the size of the voxel (decoding). However, our recent localized measurements of the fMRI response to individual Gabor elements, either alone or flanked by other Gabor patches, demonstrate unique challenges for for interpretation of responses to spatial patterns. In one case the fMRI response to low-contrast Gabor patches was not reliably suppressed by high-contrast parallel flankers, even though other studies using extended sine-wave gratings have shown that the BOLD response reflects iso-orientation surround suppression. A computational model incorporating both local/untuned and long-range/orientation-tuned inhibition resolves this discrepancy by demonstrating that the net balance of activity between neurons representing the stimulus orientation and neurons in surrounding orientation columns depends crucially on the size of the stimuli. Since fMRI reflects the responses in the entire neural population, the population-average response can obscure responses in the neural population encoding the stimulus and result in a lack of correlation between fMRI data and behavioral measurements of the responses in neurons encoding the stimulus. Contributions of local inhibitory neural activity to the BOLD response can further complicate interpretation of localized measurements. In a second experiment, we found that the relative amplitudes of the estimated fMRI response to 5% and 10% contrast Gabor patches depend on the distance of the voxel from the center of the neural representation of the stimuli. Therefore, the size of the region of interest for analysis has a significant impact on whether the fMRI response is linearly related to underlying neural activity. Taken together, these studies suggest that accurate inference of localized neural activity from fMRI data requires a detailed understanding of local neural population responses across the cortical region of interest.
This PDF is available to Subscribers Only