Abstract
Inferring neuronal activity from BOLD response is hard. Their relationship is masked by the nonlinearity between stimulus and neuronal activity. Furthermore, neuronal activity is multi-faceted and difficult to quantify even with direct, invasive simultaneous recording methods. Now consider a hypothetical system with two identical groups of non-interacting neurons that are co-localized in the same capillary bed and control the same local vascular network. Further suppose that we can manipulate each group independently with stimuli. This system would be ideal for mapping the relationship between neuronal activity and BOLD response, circumventing the unknown composition of neuronal activity and the unknown relationship between stimulus and neuronal activity – presenting identical stimuli to both groups of neurons will, by definition, double neuronal activity in the local region.
We found such a system in low-level visual areas of humans born without optic chiasm. For achiasmatic subjects, we and others have found that each fMRI voxel in V1-V3 has two population receptive fields (pRFs) symmetrically located across the vertical meridian (Hoffmann et al., 2012 Neuron, Purington et al., 2012 VSS). We have further demonstrated that the co-localized neurons of these pRFs do not interact (Bao et al., 2012 VSS). By presenting stimuli to one or both of the pRFs in these pairs, we found that resulting BOLD responses to cross-pRF spatial and temporal summations manifest the same dynamic nonlinearity inherent in biophysical models of hemodynamic response, such as the Balloon model (Buxton et al., 1998 MRM), devoid of any neuronal nonlinearity. We have thus empirically demonstrated that the Balloon model and models with similar nonlinearity accurately describe the relationship between BOLD response and neuronal activity. The visual cortex of an achiasmatic subject is a powerful tool for untangling hemodynamic and neuronal nonlinearity in phenomena such as fMRI adaptation, surround suppression, and spatial summation.
Meeting abstract presented at VSS 2013