USC Viterbi’s Maryam Shanechi aims to develop AI models to improve neuromodulation and enable adaptive treatment for cerebellar disorders
Illustration of the brain (credit: Gemini)
Imagine not being able to properly say a word or take a step. For patients with cerebellar disorders and other motor-related neurological diseases, this is their reality, as loss of basic motor control can significantly impair everyday function and gradually erode independence.
Every step a person takes, word they say and hand movement used to pick up an object depends on the cerebellum, the brain region responsible for muscle control, coordination and balance. Damage to this small structure at the back of the brain can impair nearly every coordinated movement a person makes.
Yet current treatments still fail to effectively address these disorders, a key gap that requires alternative therapy options for patients.
A promising alternative therapy option is neuromodulation, a technology that focuses on altering neural activity. By delivering targeted stimulation such as electrical impulses directly to specific areas of the brain, it can restore function or manage neurological conditions such as movement disorders.
However, existing neuromodulation therapies are “open-loop,” meaning they lack real-time brain monitoring and the precision necessary to provide adaptive care. Historically, identifying exactly where and when to apply stimulation has required extensive trial and error.
USC Researcher Maryam Shanechi
USC researcher Maryam Shanechi is leading the AI Core, the central AI effort to address the critical lack of therapeutic precision and the trial-and-error nature of neuromodulation therapies by building artificial intelligence (AI) models for cerebellar disorders in an upcoming research project. The AI team also includes Yisong Yue from California Institute of Technology (Caltech), who serves as co-investigator.
This study is funded by the Raynor Cerebellum Project (RCP), a program of the Once Upon a Time foundation with $50 million pledged toward the mission to improve the lives of people with cerebellar disorders. The Shanechi lab is part of the RCP Institute, which specifically brings together multidisciplinary teams of researchers to develop an AI-directed neuromodulation approach using both invasive techniques, such as deep brain stimulation, and noninvasive methods.
Following a highly competitive national request for applications, Shanechi was selected to lead the AI Core for the entire institute and was awarded $2.4 million for this research.
The five-year project officially began in late 2025.
Shanechi is the Founding Director of the USC Center for Neurotechnology. As Alexander A. Sawchuk Endowed Chair in Electrical and Computer Engineering and professor at USC, Shanechi holds joint appointments across the USC Viterbi School of Engineering’s Ming Hsieh Department of Electrical and Computer Engineering, Thomas Lord Department of Computer Science and Alfred E. Mann Department of Biomedical Engineering, as well as the USC Mark and Mary Stevens School of Computing and Artificial Intelligence. She brings extensive experience developing AI solutions for clinical neuroscience problems.
By creating an AI system that decodes abnormal brain activity and adjusts therapy as symptoms change, just like a “brain thermostat,” her research moves the field away from decades of trial and error and toward an intelligent, data-driven approach to neurotechnology for cerebellar disorders. The solutions her work will produce represent a paradigm shift in the treatment of cerebellar disorders.
Current treatments for cerebellar disorders, such as cerebral palsy and spinocerebellar ataxia, primarily consist of pharmaceutical drugs and behavioral therapies, including speech therapy for patients with difficulty articulating speech.
However, for many patients, these treatments remain insufficient with low efficacy. Especially because these approaches are not designed to target the specific neural sources of a patient’s symptoms and cannot adapt as a patient’s condition fluctuates throughout the day, which overall makes them less effective therapy options.
As a result, many patients continue to experience significant functional impairment despite available treatments.
One of the new and emerging treatment approaches for cerebellar disorders is neuromodulation, a technology that alters neural activity by delivering electrical, magnetic or focused ultrasound stimulation directly to specific areas of the brain.
In the field, neuromodulation is often seen as a highly promising treatment because it has the potential to directly target the source of the problem by regulating the specific brain activity linked to abnormal symptoms.
Despite its promise, neuromodulation as currently practiced is “open-loop,” meaning it is delivered without real-time monitoring of how a patient’s brain is responding to therapy. “Think of it like a thermostat,” explained Shanechi. “There is no thermostat to sense the brain’s state.” Without that feedback, physicians cannot adjust the dosage of stimulation based on a patient’s moment-to-moment brain activity, so therapy remains static even as symptoms fluctuate. This makes neuromodulation one-size-fits-all rather than adaptive, and it also makes it harder to tell whether a given treatment is working, contributing to low efficacy for many patients.
Going beyond open-loop approaches requires solving a deeper scientific and computational problem. Researchers have not yet fully identified the abnormal neural activity patterns that give rise to symptoms such as balance, gait and reaching difficulties. Without understanding what these disorder “signatures” look like, scientists cannot decode or target them precisely, creating a major bottleneck for treatment development. Even once signatures are known, neuromodulation still requires identifying the exact anatomical locations in the brain to stimulate, as well as determining the precise timing and dosage for intervention, what researchers describe as a “where, when, and how” challenge. Historically, identifying these parameters has required extensive trial and error.
Underlying all of these barriers to treatment is the complexity of neural data itself. Making sense of the high-dimensional, multimodal data required to guide neuromodulation, including intracranial electroencephalogram (iEEG), functional magnetic resonance imaging (fMRI) and functional ultrasound data, remains a massive computational challenge. Researchers still lack AI models capable of connecting these different forms of complex brain data to a patient’s actual behavioral and symptom states in an intelligent system.
Shanechi’s team aims to build an AI system that addresses the barriers facing neuromodulation for treating cerebellar disorders.
This system will be trained on multimodal data — including iEEG, fMRI and functional ultrasound recordings — collected over multiple days in the context of cerebellar disorders.
The system will be designed to identify the neural signatures of cerebellar disorders, relating complex neural activity patterns to specific behavioral symptoms, such as gait, balance and speech impairments. Decoding these symptom states directly from brain activity is the prerequisite for any adaptive therapy for cerebellar disorders and a longstanding bottleneck in the field, allowing for a level of precision that traditional drug-based treatments cannot achieve.
The system will also be designed to address the “where, when, and how” of intervention, helping identify the brain regions to sense and stimulate, the optimal timing, and the appropriate dosage. To do so, the system will predict the effect of stimulation on a patient’s neural and behavioral response, replacing trial-and-error with data-driven targeting.
Together, the system could enable adaptive neuromodulation that responds to a patient’s symptoms as they fluctuate in cerebellar disorders. This is what researchers describe as the “brain thermostat” that has been missing.
By combining artificial intelligence and neuroscience, Shanechi’s work aims to move the field away from static, trial-and-error treatment approaches and toward intelligent, adaptive neurotechnologies that physicians can test for immediate clinical improvement in cerebellar disorders.
For patients with cerebellar disorders, this could mean therapies that respond dynamically to their changing needs throughout the day, rather than treatments that remain fixed while symptoms fluctuate.
More broadly, the AI systems Shanechi’s team is building could accelerate progress on other neurological conditions where similar scientific and computational barriers have slowed treatment development for decades.
Published on June 11th, 2026
Last updated on June 11th, 2026
USC researchers discover neurons have memory and use a single neuron’s activity to map entire brain networks, first to achieve…
Algorithm can enhance decoding accuracy of brain signals in real-time brain-computer interfaces.
How advanced technology offers new hope in the fight against a devastating disease.
A breakthrough for paralyzed patients? New algorithm from Maryam Shanechi’s lab can improve brain-computer interfaces and discover new brain patterns.
AI is running on mobile phones, sensors, and home appliances thanks to the engineering innovations that make the most of…
Findings Published In Scientific Reports Employs New Methods And Tools For Decoding How Cognition, Creativity And Even Disease Evolve In…
USC Researcher to Lead $2.4M AI Core for Raynor Cerebellum Project Treating Cerebellar Disorders
Unlocking the Inner Ear: Transforming How Scientists Study Hearing and Balance
From 5G to ‘seeing around the corner’: USC Engineer Shapes the Present and Future of Wireless Communication
