By Divvij Chandna
In the last few years, there have been various technologies used to recreate images as they are perceived by the human brain. Most of these techniques are quite inefficient and expensive, therefore, it is not practical to use them for any purpose. But now, Changde Du’s team at Research Center for Brain-Inspired Intelligence in Beijing, China and Kai Miller at Stanford University have come up with different methods of their own which make this process much easier than it was before.
The common aspect of each of their methods is that they process data by using deep-learning techniques, thus “training” the algorithms to read images with ease. The algorithms aren’t required to be told that the subjects are looking at something and are capable of detecting that on their own.
The software developed at Stanford measures two types of signals from the brain: event-related broadband and event-related potentials. The former is a result of neutrons acting out of sync where as the latter occurs when they are working together. After this, the algorithm is trained by showing certain pictures multiple times and then it shows the images seen by the brain with only a half a second delay.
The software created in China uses frequency magnetic resonance imaging (fMRI) scans. fMRI is a procedure which uses MRI technology that measures brain activity by detecting changes associated with blood flow. These are used to transfer the activity from voxels (three dimensional pixels) inside the brain to pixels in an image. The program finds out the correlations between data stored in the voxels which results in clearer images and also requires much less processing.
The applications of this technology in the future are huge, this is just a small step away from being able to detect what people are thinking or dreaming and reading their minds. By using this technology the dynamics of many areas of the brain can be explored and cures to many brain diseases could be found.