Christopher Tyler¹; ¹Smith-Kettlewell Eye Research Institute

The procedure for neural population receptive field (pRF) estimation from non-invasive functional Magnetic Resonance Imaging signals (Dumoulin & Wandell, 2008) is now well established in many fields of application. We proposed parallel procedure for estimating the local neural temporal population response (Tyler & Likova, 2009), which allows estimation local neural response to its native temporal resolution from fMRI signals. The procedure involves modeling the temporal structure of the neural population response (NPR). This procedure relies on the fact that neural responses are inherently rectifying, carrying no response for negative, or hyperpolarizing, intracellular neural signals. In the population, therefore, the negative lobes of the neural temporal responses are carried by a positive signal in off-cells. Since both on- and off-cells (or Gabor RFs having all phases of the carrier wave) have the same sign of energy requirements, the BOLD metabolic response to local neural activation is inherently rectifying, providing the nonlinearity required for estimation of the temporal NPR kernel from BOLD signals as a function of stimulus duration. We model the sensitivity of the procedure as a function of the NPR parametrization and fMRI signal-to-noise ratio. This shows that the larger the difference in time constant between the BOLD MRK and the NPR kernel, the better the NPR kernel estimation. The estimable parameters are the neural transient time constant and gain, the transient/sustained response ratio, and the off-response polarity and gain. The results show that the typical fMRI signal-to-noise ratio is sufficient to determine the time constant for a typical neural population transient of 0-100 ms to a resolution of ~25 ms. However, there is no simple relationship between the summation curve property and the underlying nonlinearities and time constants. To determine the specific parameter values, the parametrized model NPRK was optimized to the full dataset for V1 and hMT+.