Original URL: http://www.theregister.co.uk/2013/08/20/mindreading_mri_spots_letters_in_the_brain/

Mind-reading MRI reads letters in the brain

Reading the reader

By Richard Chirgwin

Posted in Science, 20th August 2013 04:08 GMT

Researchers from Radboud University Nijmegen are claiming that with a sufficiently-sensitive MRI and decent mathematical modelling, they can reconstruct images of the brain recognising letters seen by the test subject.

Specifically, the researchers say they have “used data from the scanner to determine what a test subject is looking at”, according to this release, in a paper to be published in the journal Neuroimage.

“The researchers 'taught' a model how small volumes of 2x2x2 mm from the brain scans – known as voxels – respond to individual pixels. By combining all the information about the pixels from the voxels, it became possible to reconstruct the image viewed by the subject. The result was not a clear image, but a somewhat fuzzy speckle pattern. In this study, the researchers used hand-written letters,” the release states.

After their initial fuzzy pattern, the researchers added prior knowledge of what a letter should look like to the model. The Register can't help wondering if this doesn't sound like “looking up the answer in the back of the book”, so we'll quote the release directly and let you judge:

Scanned word

Seen in the brain: "thought" reconstruction from an MRI

“We gave the model prior knowledge: we taught it what letters look like. This improved the recognition of the letters enormously. The model compares the letters to determine which one corresponds most exactly with the speckle image, and then pushes the results of the image towards that letter. The result was the actual letter, a true reconstruction”, said researcher Marcel van Gerven.

The research aim, van Gerven says, is to find ways to model what the brain is experiencing, so that the models could be applied to “working memory or to subjective experience such as dreams or visualisations”.

The next step in the research is to increase the resolution from the current 1,200 voxels to 15,000 voxels. ®