Nuclear fusion is only 20 years away — and it has been for the last 50 years. So goes the joke in the physics community, reflecting the ever-elusive nature of this scientific breakthrough. In the realm of neurology, we face a similar paradox.
For years, neurologists have anticipated a revolution in the understanding and treatment of neurological disorders, but progress has felt incremental. Per a recent study in The Lancet, the need for such progress is more urgent than ever: Neurological conditions are now the leading cause of ill-health worldwide. It is imperative, then, that we utilize and discover new tools to derail this trend.
Modern artificial intelligence (AI) may provide the innovation we need. AI offers smarter diagnostics, personalized treatments, and new technology that can transform clinical neurology and neurosurgery. But what makes its advent a revolution is that AI presents new ways to harness centuries of neurological study and data, allowing us to understand the brain, intelligence, and humanity like never before.
A Glimpse of the Future
Enter Noland Arbaugh.
In 2016, Arbaugh was on his summer break from college, working as a camp counselor, when he suffered a freak diving accident which severed his spinal cord. He was rendered quadriplegic. Now 30 years old, Arbaugh has become the first human patient to receive a brain-computer interface (BCI) developed by neurotech company Neuralink. The device, named “the Link,” aims to create a direct communication link between the brain’s neuronal activity and external devices. AI serves as the translator, decoding complex electrical data.
In a groundbreaking procedure, Neuralink’s team implanted the Link onto Arbaugh’s motor cortex. This device, capable of processing vast amounts of data, enables him to control a computer by merely thinking about the actions he wants to perform. Despite facing some complications such as loss of many of the contacts between the BCI and cortex, Arbaugh has been able to regain a level of autonomy he had not experienced in years — from browsing the web to messaging family and even playing video games. BCIs like the Link demonstrate how AI can be used to understand a patient’s specific, personalized brain activity and the transformative potential this holds for neurological injury. But BCIs are just the start.
One of AI’s greatest contributions to neurology will be personalized treatment plans tailored to each individual patient’s needs and specific condition. For instance, AI algorithms can analyze a Parkinson’s patient’s gait, tremors, and response to medication to tailor deep brain stimulation (DBS) parameters and electrode placement, improving motor function and reducing medication side effects. Similarly, advancements in multimodal AI techniques show how an AI system could integrate imaging, biomarkers, and clinical data for the care of neurodegenerative disease, like Alzheimer’s, creating customized treatment plans that improve outcomes compared to traditional approaches. Indeed, these new systems not only enable personalized treatment, but improve our understanding of disease.
AI will also enable the neurology field to become far more proactive, via predictive analytics. By analyzing massive datasets, AI can detect early signs of neurological disorders like epilepsy and recommend preventive interventions to stop disease progression. This predictive capability gets health care ahead of the disease curve. Startups like Neuralight, Vytal, and ViewMind are developing AI-powered platforms to detect early cognitive decline from smartphone-based eye tracking tests and virtual reality assessments. Other startups are pioneering AI systems on the diagnostics front. One example is Enlitic, which uses AI to rapidly analyze medical imaging to accurately detect strokes, tumors, and neurodegenerative diseases. Improving both precision and timeframes for diagnoses is crucial for effective treatment. For example, AI-enabled diagnostic technology can allow EMTs and other paramedics to make critical decisions about whether a stroke is hemorrhagic or not before the patient even reaches the hospital. With the brain every minute counts.
Through the Looking Glass (of AI)
Physics and Neurology share other commonalities beyond the joke mentioned up top. In his 1977 book, “The Biological Origin of Human Values,” physicist George Edgin Pugh wrote, “If the human brain were so simple that we could understand it, we would be so simple that we couldn’t.” Through the looking glass of AI, perhaps we can bypass the ‘we.’
As a Harvard Medical School student interested in neurology, I am especially excited about new research conducted at home. One place that’s at the forefront of AI innovation is the Kempner Institute. Established in 2021 by a $500 million grant from the Chan Zuckerberg Initiative, the Kempner’s mission is to understand the basis of intelligence in natural and artificial systems. This dual focus enables new approaches toward answering the big questions of neuroscience.
Central to the Kempner’s approach is understanding brain mechanisms by using AI systems to crunch biological data and neural processes. Kanaka Rajan, a founding faculty member at the Kempner, explains, “One of the major changes … is that we have grown to work with much larger-scale data. We first started looking at hundreds of neurons in one brain area, and now we’re looking at thousands of neurons across multiple brain areas in many different species.” This increase in scale, made possible by AI research, is crucial for identifying the mechanisms underlying various neurological disorders, paving the way for more effective treatments. Such insights could revolutionize our approach to diseases like Alzheimer’s and Parkinson’s, offering hope for interventions that target these conditions more precisely.
The Kempner Institute is also uncovering mechanisms of brain plasticity and circuit functionality. They are drawing parallels to artificial intelligence, using one system to improve and understand the other. In an exciting collaboration with Google DeepMind, Harvard researchers developed an AI “virtual rodent” that accurately simulates the behavior of a real rodent — and whose artificial neural network activity can be used to predict real neural activity. Utilizing AI to model brain activity and identify circuits can allow us to simulate complex brain activity even in humans without invasive procedures. Bernardo Sabatini, co-director of the Kempner, has been observing perturbations in such circuits and identifying how they contribute to neuropsychiatric disorders. This capability is a boon for developing new therapeutic strategies, allowing clinicians to predict responses to treatments under various conditions and tailor interventions accordingly.
Reflections (of Humanity)
While the Kempner has focused on software, another recent Harvard lab came into the spotlight for their work with human “hardware”: The Lichtman Lab teamed up with Google to release 1.4 petabytes (a million gigabytes) of neuronal connection data from a real human brain. That corresponds to 16,087 neurons — one cubic millimeter of brain. In total, the average human brain is considered to be nearly 100 trillion neurons in size. This is in addition to the genetic information, neural imaging, and comprehensive health records that contribute to the 50 petabytes of data generated at our hospitals every year.
Such complexity is, as Dr. Pugh quipped above, unfathomable to humans. And such massive datasets show how the integration of AI in neurology is not just useful or a passing fad, but necessary to understand the brain at the fundamental level. Leveraging AI to detect patterns allows us to make sense of such data in ways that escape conventional analysis.
In a field defined by complex interplays between genetic and environmental factors over time, AI offers a new approach. In the next five years, expect to see new automated diagnostic tools, life-changing machines, and personalized treatments as we bring existing AI capabilities to our neurology patients. But in the next 20, expect to see natural intelligence’s greatest creation — its artificial counterpart — unlock the secrets of the human mind.
And hey, we might achieve nuclear fusion along the way!
The author would like to thank Rishab Jain for his insights which contributed to the piece. Rishab is a researcher in the Department of Neurosurgery at Massachusetts General Hospital working on understanding speech through single-neuron recordings in humans. He studies computational neuroscience at Harvard University.
The author does not have any disclosures to report or any relationships with the companies or specific research labs referenced in the article.
What advances in AI for neurology most excite you? Shout them out in the comments!
Aditya Jain is a third-year medical student at Harvard Medical School. His previous works include "The Future is STEM" and medical fiction shorts for In Vivo Magazine. He is a published researcher on the applications of artificial intelligence in medicine. When he's not busy with rotations, he enjoys playing guitar, reading sci-fi, and nature hiking. He tweets @adityajain_42. Aditya is a 2023–2024 Doximity Op-Med Fellow.
Illustration by Diana Connolly
