Abstract
There is great interest in developing methods for identifying the VEPs of separate visual areas. This goal has been difficult to achieve because early visual areas are so closely spaced.
Our approach to this problem uses MRI and fMRI to constrain the solution. Previously, we recorded VEP's to each of 192 stimulus patches in a dartboard display.
The VEP dipole sources were initially constrained to the cortical locations specified by the fMRI. Then a nonlinear search allowed the dipoles to move by up to 3mm to improve the fit of the forward model FMeeg) to the VEP data, which achieved excellent agreement between the time functions for the right and left hemispheres for V1 and moderately good agreement for V2. This step was needed because of inaccuracies in both the fMRI and in the dipole forward model. As a result we have two types of forward models predictions: FMfmri based purely on MRI/fMRI information and FMeeg based on the best fit to a high resolution (192 patches) dataset.
To improve signal to noise ratios we replaced the 192 patch stimuli with large designer patches chosen to lie on relatively flat regions of V1 and V2 that individually activate close to 10 times the cortical area. We carried out large patch simulations using the data from our 192 patch experiments. As expected, the FMeeg improved V1/V2 time function isolation. In addition, the designer patch results were compared with those obtained using comparably large uniformly distributed patches (non fMRI constrained) that avoided the horizontal and vertical meridians. The intelligently designed stimulus patches performed substantially better than arbitrary placed patches at separating V1 and V2 responses.
The use of designer stimuli increases SNR, thereby reducing recording time, yet promises to fulfill the goal of separating the responses of multiple visual areas.