KOCH VISION Journal Club
<p>Jan 20, 2010 12:06 PM
 
Visual Image Reconstruction from Human
Brain Activity using a Combination of
Multiscale Local Image Decoders
 
Miyawaki, Uchida ...   Kamitani   all from Kyoto  Natl INstit of INfo and Communc Tech.
 
NEURON  60  915-929  Dec 11 2008.
 
 
Perceptual experience consists of an enormous num-
ber of possible states. <b>Previous fMRI studies have</b>
<b>predicted a perceptual state by classifying brain</b>
<b>activity into prespecified categories</b>. Constraint-free
visual image reconstruction is more challenging, as
it is impractical to specify brain activity for all possible
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 vox-
els and exploiting their correlated patterns. Binary-
contrast, 103 10-patch images (2100
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. Reconstruction was also used to identify
the presented image among millions of candidates.
<b>The results suggest that our approach provides an ef-</b>
<b>fective means to read out complex perceptual states</b>
<b>from brain activity while discovering information</b>
<b>representation in multivoxel patterns</b>
 
 </p>