Fig. 5: Using computational models to identify features of the stimulus that drive neural responses in each fROI.
From: Computational models of category-selective brain regions enable high-throughput tests of selectivity
a Schematic of the algorithm used to generate importance maps. We convolve each image with 2000 masks (left) and run the masked images through our brain model for each fROI to generate the predicted response for each image. The importance map (right) is the generated by taking a weighted sum of the masks with these predicted responses. Permission was obtained from the person in the photograph to publish the image used in this figure. b Importance maps for the left and right FFA for the images predicted to generate high responses. The warm colors on the overlaid colormaps indicate regions that most strongly activate the FFA and cool colors indicate regions that do not activate the FFA. c Same as b but for left and right EBA. d Same as b but for left and right PPA. e Importance maps for three representative photographs with faces, bodies and background scenes and importance maps for each fROI. Permission was obtained from the persons in the photographs to publish the image used in this figure.
