One of the long-term goals of the field of neuro-imaging is to understand what a person is thinking just by looking at the pattern of his or her brain activity–in essence, reading the mind. While that feat is still a long way off, scientists at the University of New Mexico have taken an important step by refining neuro-imaging techniques to the point where they can reliably detect a single thought forming in an individual’s brain.
The technique could be used to improve clinical applications of neuro-imaging, such as patient diagnosis, or to study cognitive processes that are fleeting or irreproducible, such as learning a new skill. “This could open up a whole new dimension of how fMRI could be used,” says Peter Bandettini, director of the fMRI core facility at the National Institutes of Health in Bethesda, MD.
Functional magnetic resonance imaging (fMRI) measures the amount of blood flow to different parts of the brain, thereby indicating which brain areas are most active. Imaging the brain while someone performs mental tasks, such as remembering words or doing math, gives insight into the parts of the brain crucial for these cognitive processes.
But brain activity is very “noisy,” meaning scientists must distinguish the relevant neural signals from background activity, which might come from a subject’s breathing, moving, or even daydreaming. To detect brain activity associated with a specific task, then, most fMRI studies average brain scans from repeated tests in dozens of people.
Stefan Posse and colleagues at the University of New Mexico are developing new ways to collect and analyze fMRI data that allow them to detect brain activity from a single thought. They’ve created their highly sensitive imaging methods by taking more pictures in a shorter amount of time and by developing new algorithms to integrate those images and to reduce background noise.
As described in a paper last month in the journal Neuroimage, Posse’s team asked eight volunteers lying in a scanner to think of a word beginning with a letter flashed on a screen above their faces. They then recorded the activity in Broca’s area, a part of the brain involved in the generation of language.
The researchers found they could detect activity in this region after a single trial about as reliably as in previous studies that averaged results over multiple trials. “We can actually see it on the scanner in real time,” says Posse.
Experts say the findings are promising, although they still need to be confirmed. “It’s important to try this in other brain regions and with other tasks to see where it works,” says Christopher deCharms, founder of Omneuron, a brain imaging startup in Menlo Park, CA.
The researchers are now trying to collect additional information from these brief brain activity patterns. “Sometimes, we would see a second bump in Broca’s area,” says Posse. “Subjects then told us they had thought of a second word.”
Eventually, the researchers hope to be able deduce even more complicated characteristics–such as the type of word the person generated, whether the word made them feel happy or angry, and, ultimately, more complex thoughts. “If you can see activity generated by a single word, maybe you can also see activity from a longer sequence of thoughts, then complex brain processes,” says Posse. “The idea is to be able to decompose the stream of thought processes into individual thoughts.”
Brain-imaging experts say a technique that reliably measured single thoughts could open up a new world of experiments. “If we can succeed in measuring data from a single trial, it gives us a more powerful method than what’s been available,” says deCharms. “You could monitor performance in a task, like surgery or flying a plane, if you wanted to understand how performance changes second by second.”
And this single-thought method could be used to study the learning process itself, which “happens very quickly,” says deCharms. “You can only have a first impression once.” To study a specific learning process, scientists would need to measure the difference in brain activity between the first time someone performs a new skill and the second or third. “If you have to average over 20 trials, you lose a lot of insight,” says deCharms.
Single-trial brain-imaging techniques could also be useful in the clinic. Currently, fMRI is rarely used for diagnostic or therapeutic purposes because it is difficult to gather reliable data from an individual brain image. But as fMRI techniques such as Posse’s allow more-sensitive imaging, doctors will be better able to make medical decisions from individual brain scans. In addition, says Bandettini, the ability to generate reliable images in shorter amounts of time will make the process easier for both doctors and patients.
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
The Biggest Questions: What is death?
New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
How to fix the internet
If we want online discourse to improve, we need to move beyond the big platforms.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.