Cortical surface maps show the distributions of weight magnitudes for a foveal and a peripheral patch ( Figure 4A). We first examined the distributions of voxel weights of local decoders in comparison with the conventional retinotopy. see Figure S1 for comparison with conventional algorithms without sparse voxel selection). Thus, the decoder is not explicitly informed about the retinotopy mapping. The decoder uses all the voxels from the early visual areas as the input, while automatically pruning irrelevant voxels. Hence, each local decoder serves as a “module” for a simple image component, and the combination of the modular decoders allows us to represent numerous variations of complex images. As each local element has fewer possible states than the entire image, the training of local decoders requires only a small number of training samples. The stimulus state at each local element (C i, C j, …) is predicted by a decoder using multivoxel patterns (weight set for each decoder, w i, w j, …), and then the outputs of all the local decoders are combined in a statistically optimal way (combination coefficient, λ i, λ j, …) to reconstruct the presented image. We assume that an image is represented by a linear combination of local image elements of multiple scales (colored rectangles). Here, we present an approach to visual image reconstruction using multivoxel patterns of fMRI signals and multiscale visual representation ( Figure 1A). The results suggest that our approach provides an effective means to read out complex perceptual states from brain activity while discovering information representation in multivoxel patterns. Reconstruction was also used to identify the presented image among millions of candidates. Binary-contrast, 10 × 10-patch images (2 100 possible states) were accurately reconstructed without any image prior on a single trial or volume basis by measuring brain activity only for several hundred random images. In this study, we reconstructed visual images by combining local image bases of multiple scales, whose contrasts were independently decoded from fMRI activity by automatically selecting relevant voxels and exploiting their correlated patterns. Constraint-free visual image reconstruction is more challenging, as it is impractical to specify brain activity for all possible images. ![]() Previous fMRI studies have predicted a perceptual state by classifying brain activity into prespecified categories. Perceptual experience consists of an enormous number of possible states.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |