Scientists in Kyoto have reproduced an image based on how the brain responds – effectively reading the brain.
Yoichi Miyawaki and colleagues at the ATR Computational Neuroscience Laboratories have published a report in the journal Neuron, where they used functional Magnetic Resonance Imaging (fMRI) to observe the changes in brain activity when a subject is looking at an image. Using complicated computer modelling, they were able to train the system to recognise what happens in the brain in response to certain images, and the system was then able to accurately reproduce images that the subject had not seen before.
The images were simple, high contrast pictures, a 10 by 10 grid with each ‘pixel’ containing a shade from white to black, on which they could produce random patterns called a contrast map, geometric shapes or letters. In the training stage, they showed the subject 440 random images for 6 seconds each while observing the fMRI, which recorded brain activity by looking at blood flow in the visual cortex, which is located at the back of your brain.
Their computer model went to work by analysing small, overlapping portions of the fMRI in 3 dimensions, and then putting together the data from each portion to reconstruct how the brain responded to each image. Using multiple portions allowed them to build up an idea of the brain’s response, with the overlaps reducing the error, as each section was sampled more than once. Once it was trained on the random contrast maps, they moved on to showing the subject more defined images, such as letters and geometric shapes.
From the fMRI data, the computer model was able to reconstruct the images as seen by the subject, such as spelling out the word Neuron. It’s only able to work in black and white at them moment, and relies on the images having quite high contrast, but the team think that improving measurement accuracy should enable it to work in colour, and one day may be able to produce images from dreams, or even a direct readout of someone’s feelings.
This is a good start, but they have a long way to go before you can upload your dreams at night to watch later on!