Mind reading is more commonly associated with mentalists rather than scientists, although that may not be the case for very long. Functional magnetic resonance imaging (fMRI) scans allow scientists to see brain activity as it happens, essentially viewing thoughts on screen in real-time. However, deciphering those thoughts from the digital data obtained from fMRI scans is a challenge.
fMRI scans show changes in blood flow in the brain and blood flow changes are known to be associated with neuronal activity. When a person undergoes an fMRI scan and is asked to perform certain tasks, scientists can see which parts of the brain are active. fMRI scans produce an incredible amount of data, and analyzing those data is a mammoth task. Analyzing the data in real time is all but impossible without the use of complex algorithms and serious computing power.
fMRI scans were first performed about 20 years ago and considerable research has been invested in the technology. Numerous studies have been conducted using fMRI scans in an effort to decipher thoughts. Typically, those studies involve conducting fMRI scans on multiple individuals while they perform certain tasks and comparing the scans from those individuals. It is then possible to see which parts of the brain are being used for specific cognitive functions.
However, without the use of computers, analyzing the results of those studies can take days or weeks. By using computers, scientists can analyze chunks of the data and process them simultaneously, speeding up the process considerably. Advances in parallel processing and better algorithms have cut the time for analyses considerably.
Now a team of neuroscientists at Princeton University, with assistance from computer scientists at Intel, have developed the necessary algorithms and software to enable fMRI data to be analyzed in real-time.
To do that, the researchers broke down the fMRI scans into tiny 1 mm cubes, similar to the pixels in a digital image. Those cubes are called voxels and each contains a considerable amount of data. The voxels – of which there are around 100,000 in an fMRI image – are analyzed simultaneously.
Since each voxel can communicate with all of the others, there is a vast number of possible connections and interactions. Those interactions can change from second to second. Deciphering all of the data and decoding that information is a gargantuan task.
The Princeton University/Intel project started a couple of years ago. 10 computer scientists from Intel are now working on the project along with 20 scientists at Princeton University. The team is using more than $1.5 million of computer equipment that was provided by Intel to process the fMRI data.
The scientists at Princeton want to gain valuable insights into the functioning of the brain to improve understanding of how the brain thinks. Intel has a similar goal, although it is hoped that the information gathered from the study can be used to develop better artificial intelligence learning systems.
Numerous experiments have now been performed, including one in which a subject undergoes an fMRI scan while concentrating on a detail-rich image – a photograph of a crowded café. J. Benjamin Hutchinson, assistant professor at Northeastern University and former postdoctoral researcher at Princeton, designed the experiment. He was able to tell in real-time when the subject was concentrating on the picture and when the subject’s mind began to wander. The subjects would be rewarded for concentrating by increasing the colors of the image or making the focus sharper. It is hoped a similar technique could be used to aid concentration and weaken intrusive memories – essentially re-training the brain.
Another member of the research team, Kenneth Norman, professor of psychology at Princeton, said “As cognitive neuroscientists, we’re interested in learning how the brain gives rise to thinking,” Norman went on to say, “Being able to do this in real time vastly increases the range of science that we can do.”
The potential of the project and the technology is considerable, with many possible applications in science and medicine. Being able to analyze thoughts could help physicians make medical diagnoses faster and scientists could learn how memories are formed and accessed and how the brain thinks. Ultimately, scientists could use the technology to see what people are thinking in real time.
According to the researchers, the advanced computational analysis techniques such as parallel processing, algorithmic optimization and machine learning “is enabling a new generation of experiments and analyses that could transform our understanding of some of the most complex—and distinctly human—signals in the brain: acts of cognition such as thoughts, intentions and memories.”
The team hopes that it will soon be possible to generate an image on screen that reflects a special memory that is being retrieved by a patient and to do that in real-time. While being able to see what people are thinking about using an fMRI scan may be a long way off, the team is making significant progress toward that goal.
A paper on the project – Computational approaches to fMRI analysis – has recently been published in Nature Neuroscience.