Item

Abstract

Functional resonance imaging (fMRI) has been used to measure the retinotopic structures of the human visual cortex in vivo. Recently, this stream of research has been advanced by introduction of a computational model to fit a predefined population receptive field (pRF) model to fMRI signals observed. This method has advantages over the previous methods by providing receptive field (RF) size as well as more accurate retinopic maps. However, this model is limited because this method need assume the pRF as a certain model (e.g., circular Gaussian). To overcome this limitation, in the present study, we introduce a new method to visualize the pRF structure prior to modeling. This method estimates the pRF structure by fitting a set of weights representing the pRF topography in space to observed fMRI signals. For that, let vector p and s represent a pRF topography and a stimulus aperture. When visual stimuli present through the aperture, the pRF response is given as r = ps. As the pRF response is observed in the form of fMRI signal, it is required to convolve it with a canonical hemodynamic response function h. Therefore, the final pRF prediction x is given: x = h*r = h*(ps) Here, * denotes convolution. From this model, vector p is estimated by using the ridge regression. Application of our method yielded clear pRF structures which include the pRF center and surround regions. In addition, some pRF centers looked elliptic while the previous method assumed the pRF is isotropic.Therefore, this approach allows scientists to select a more appropriate pRF model based on the pRF topography observed. This application resulted in more accurate eccentricity map than the one by the previous method (directly-fitting circular Gaussian model). Furthermore, we could observe pRF properties such as elongation and orientation of the pRF center, and the surround suppression.