Abstract
Purpose. Prior functional brain imaging studies have measured the location and amplitude of cortical responses to chromatic contrast. We aim to develop a model of cortical response that captures the effects of color direction and contrast. We build on the observation that psychophysical chromatic sensitivity is well-described by ellipsoidal isodetection contours. Specifically, we ask whether a model employing ellipsoidal isoresponse contours describes the BOLD response in human V1. Methods. We scanned 3 subjects while they viewed 12 Hz flickering full field symmetric chromatic modulations, restricted to the LM cone-contrast plane around a common background. We studied 8 color directions at 5 contrast levels, with responses for each direction/contrast pair measured in ten separate acquisitions. An anatomical template was used to restrict analysis to voxels in V1 corresponding to 0 to 20 degrees eccentricity. We fit the data with two models: i) a general linear model (GLM) in which each direction/contrast was modeled by its own regressor ii) a quadratic color model (QCM) in which the isoresponse contours were elliptical and the effect of overall contrast was modeled by a single sigmoid. Quality of fit was evaluated using cross-validation, with one acquisition left out in each cross-validation iteration. Results. Cross-validated RMSE, evaluated with respect to the time course of the BOLD response, was essentially the same for the full GLM and more restricted QCM. The isoresponse contours obtained from the QCM showed that more chromatic contrast is required for a modulation in the L+M contrast direction to produce the same response as a modulation in the L−M contrast direction, matching prior work. Conclusions. The QCM provides a good account of the BOLD fMRI response across LM color direction and contrast. This account in turn provides a quantitative description of the BOLD response to modulations of any direction/contrast within the LM plane.
Acknowledgement: NIH RO1 EY10016