by Daniel Ben Levi, AP Lang student and Features Contributor
photo courtesy of Carnegie Mellon Engineering
Questions surrounding the brain functions behind language learning, speech comprehension, and cognitive disorders have puzzled researchers for decades, but in the last few years, advancements in brain imaging technologies and artificial intelligence have started to produce answers. Now, a new technology out of Carnegie Mellon’s Biomedical Functional Imaging and Neuroengineering Laboratory may push us closer to unlocking the secrets of the brain.
Created by Dr. Bin He, a professor of Biomedical Engineering, and a dedicated team of researchers, the new brain imaging technology is a novel form of functional neuroimaging, combining electromagnetic encephalography (EEG) with deep learning to produce a highly accurate 3D map of brain activity. Like previous EEG technologies, the novel tech employs electromagnetic waves to track brain activity over time, but unlike older models, Dr. He’s new form of imaging also allows for extremely accurate spatial reasoning, allowing researchers to not only see changes in the brain’s activity but also identify where those changes are occurring.
The technology employs a cap that is placed on the subject’s head in order to track electronic and magnetic signals in different areas of the brain. The cap contains a number of contact points, filled with gel to improve conductivity, that pick up brain activity at high speeds. When a signal is received, a novel artificial intelligence algorithm extrapolates the location of the signal on the scalp to pinpoint the location of activity inside the brain. Dr. Bin He is excited to see the applications of the new brain imaging methodology, explaining that researchers will be able to use it to better study “cognition, perception, and various motor functions, along with epilepsy, pain, or mental diseases in clinical applications.”
Magnetoencephalography (MEG) is a close counterpart of the EEG technology that Dr. He is working to improve; MEG also has a high sampling rate but relatively poor spatial reasoning capabilities. “With an extremely fast sampling rate of 1000 scans per second, MEG technology can provide fantastic data for tracking brain activity over time,” notes Professor Laura Gwilliams a Psychologist at NYU. Professor Gwilliams researches the brain’s ability to process speech, hoping to one day recreate the computations that the brain performs in order to convert sound to meaning. “We know that this is possible, we just don’t know what the computations are yet,” she says.
If successful, Gwilliams’s research would allow for both extremely advanced AI assistants and better diagnoses of brain injury or disorder. “If you understand the computations of the healthy brain,” she explains, “you can isolate problems in the unhealthy brain, which allows for more specific treatment.”
In a recent study, Gwilliams and her team utilized MEG technology to discover how the brain processes speech sounds over time. Training a machine learning algorithm on data from 21 participants, the researchers were able to link patterns of neural activity to certain sounds. Interestingly, they found that there wasn’t one single pattern that was tied to a sound – the number of possible patterns was tied to the number of time samples around the sound. Furthermore, based on whether or not the participant could predict the following utterance, stored sounds were either discarded quickly or kept for longer periods of time.
“This is very exciting,” Gwilliams remarks, “because this gives us insight into how the processing of speech sounds connects to the processing of words.” The study represents an important step forward in understanding the relationship between brain function and speech.
During our interview, Professor Gwilliams noted that one major drawback of using MEG technology was its inability to pinpoint the exact areas of the brain in which the activity was occurring – a ballpark estimate compared to an exact answer. For her most recent study, location wasn’t a focus, so this didn’t pose any problems, but it’s clear that if the spatial reasoning of an fMRI scan could be combined with the sampling rate of EEG or MEG, collecting precise data would be significantly easier, allowing for further discovery.
This is where technologies like Dr. He’s come into play, giving researchers the best of both worlds. Dr. Gwilliams notes her excitement for such advancements, exclaiming that AI “is definitely the future! Neuroscience is in such an exciting position because of all the modern advances in AI that could push forward data analysis and the construction of cognitive models.”
Halie Olson, a graduate student in MIT’s Brain and Cognitive Sciences Program who researches child brain development, supports Gwilliams’ enthusiasm. Olson explains that in her work, precise spatial reasoning is extremely important, making technologies like Dr. He’s that provide more accurate imaging immensely valuable.
“The differences that we can see in the brain right now are dependent on the resolution of our tools, ” Olson explains. “The more precise our tools get, the more we’re going to be able to differentiate between specific neural networks.” Interestingly, Olson notes that advancements in imaging tech have allowed AI researchers to rethink their design processes, leading to more advanced imaging technologies and a constant cycle of complementary refinement.
Brain imaging technologies have also played a revolutionary role in how we approach language instruction research in the past few years. A recent study conducted by Dr. Rachel Romeo and her team at MIT’s McGovern Institute for Brain Research found that socioeconomic background may influence the types of struggles students face in learning to read.
Using fMRI technology, the researchers mapped the brain functions of 155 children, some with and some without reading disorders, while they read, allowing them to pinpoint the location of brain activity during the task. They found that while children from a higher socioeconomic background struggled most with phonological processing, the piecing together of sounds into words, children from lower socioeconomic backgrounds had more difficulty with orthographic processing, rapidly naming visually presented letters or words.
The team offered a few possible reasons for the contrast, one being that children from lower socioeconomic backgrounds have less accessibility to books, but regardless of the reason, the findings call into question modern practices in aiding those with reading disorders; it may be time to personalize learning even more, using these novel findings as a baseline.
“Language acquisition curricula have been revolutionized in the past few years,” Dr. Deborah Hahn, a French language instructor at Newton South High School, explains. She understands the importance of small group, proficiency-oriented learning, telling me that “people used to learn languages out of books, memorizing all of their nuanced patterns but lacking the oral practice needed to achieve fluency. Now, teachers focus more on getting students to use language to communicate meaning instead of solely mastering grammar.”
When asked what she’d like to see improved in the modern curriculum, Dr. Hahn replied that “small programs utilizing proficiency testing and authentic resources will probably have the most success, but a shift towards that would require much funding and training.”
This is why studies such as Dr. Romeo’s are so important. Scientifically confirmed evidence exemplifies the need for change, and that evidence can only be conclusively retrieved with the brain imaging technologies of the 21st century. If Dr. He’s novel techniques were applied to Romeo’s experiment, the more precise temporal imaging would likely allow for an understanding not only of the differences in brain region activation, but also in actual function, shedding further insight into how reading disorders could better be overcome.
To learn more about how advancements in brain imaging tech could aid clinical psychologists in providing better treatments, I spoke with Dr. Mark Kline, a licensed psychologist with over 30 years of experience.
Kline is a strong advocate of the cognitive revolution, arguing for a stronger connection between psychologists and neuroscientists in order to increase the rate at which novel cognitive research is employed in clinical treatment. “You can’t really practice in this field unless you’re willing to rule out known biological factors,” Kline explains. “Brain imaging technologies change my diagnostic evaluations; if a patient has a neurobiological disorder that we can identify, we can employ biological treatments in addition to psychological ones.”
He goes on to describe the benefits of modern advancements in brain imaging technologies, stating that “The higher the quality of your brain scan, the more accurate your diagnosis will be. If you had a screening EEG that could pick up areas of lesser activation that could be associated with depression or schizophrenia, that could be really helpful in guiding psychopharmacological treatment.”
In addition to the incorporation of biological factors that Kline describes, some psychologists have begun using novel, evidence-based treatments such as cognitive behavioral therapy and dialectical behavioral therapy, using our expanded knowledge of the brain to fine-tune their practice. By harnessing our understanding of the brain activity behind cognitive disorders and conducting studies on the effectiveness of their treatments with brain imaging tech, these psychologists help their patients deal with mood disorders and suicidal ideation more effectively.
Dr. Jessica Engel, a psychology instructor at Newton South High School, shares Kline’s enthusiasm for modern advancements in brain imaging technology, remarking that “the advent of fMRI allowed us to see what areas of the brain activate when subjects engage with different stimuli, allowing us to map blood flow activity to draw correlations.” Those correlations have driven many fascinating discoveries, explains Engel, putting emphasis on the finding that the brain doesn’t actually shut down during sleep, which would have been entirely impossible without the advancements of the cognitive revolution. She is confident that continued efforts to improve imaging will be fruitful, stating that “their effects will probably be more significant than we could ever imagine.”
One thing is certain: no matter which one of these researchers, practitioners, or teachers you ask, it is clear that the fields of psychology, neuroscience, and linguistics are undergoing a major transformation, one that will certainly change the way we understand our world in the coming decades.
- Huge thank you to Jessica Engel, Laura Gwilliams, Deborah Hahn, Mark Kline, and Halie Olson for the insightful interviews!
- Works cited:
- Sun, Rui, et al. “Deep neural networks constrained by neural mass models improve electrophysiological source imaging of spatiotemporal brain dynamics.” Proceedings of the National Academy of Sciences 119.31 (2022): e2201128119.
- Gwilliams, L., King, JR., Marantz, A. et al. Neural dynamics of phoneme sequences reveal position-invariant code for content and order. Nat Commun 13, 6606 (2022).
- Rachel R. Romeo, Tyler K. Perrachione, Halie A. Olson, Kelly K. Halverson, John D.E. Gabrieli, Joanna A. Christodoulou. Socioeconomic dissociations in the neural and cognitive bases of reading disorders. Developmental Cognitive Neuroscience, 2022; 58: 101175