Abstract
Informational functional magnetic resonance imaging (fMRI) measures are widely used to infer the involvement of cortical areas in a particular task or phenomenon. One of the most popular of these methods, multi-voxel pattern analysis (MVPA) using linear support vector machine, quantifies the predictability of the presented stimulus based on the pattern of responses of a group of voxels. This method has been applied to the study of visual processing and perceptual learning in retinotopic visual areas. However, interpretation of MVPA results is limited by a lack of information about the bounds on decoding performance from feed-forward activity given the power of the analysis methods. To address this, we have developed a forward model of MVPA in primary visual cortex. This model begins with a representation of the V1 cortical surface and its mapping with the visual space (Rovamo and Virsu, 1984) and incorporates the Balloon Model (Buxton et al., 1998) to generate the BOLD response. Known components of BOLD signal are integrated, including spatial correlation in BOLD signal due to hemodynamic spread and the spatially and temporally correlated noise that is inherent in BOLD measurements. For each subject, signal-to-noise ratio and correlation structures were determined by tuning model parameters to achieve the best match between the BOLD-signal modulation during retinotopy and model output with retinotopic mapping stimuli as input. The model was then used to simulate BOLD fMRI data for letter stimuli presented at a given eccentricity, to be analyzed with MVPA. This model provides an upper bound for empirical MVPA performance if the observed BOLD signal is due to retinotopically organized feed-forward activity, and can thus aid in interpretation of MVPA results in retinotopic visual areas as due to predominantly feedforward versus post-stimulus activity such as feedback.
Meeting abstract presented at VSS 2013