Menu
Neurostimulation and BCI has become increasingly relevant as a accepted means to treat a wide variety of different neurological disorders. Drug resistant psychiatric disorders, such as depression and OCD, and movement disorders, including Parkinson’s disease, are some of the notable afflictions with government approved therapies available. As such, the associated hardware devices and surgical procedures have advanced considerably.
Despite these improvements, the algorithms that govern how stimulation is applied haven’t changed in well over half a century. And BCI technology is subject to error and drift. Current neurostimulation methods are simplistic, not patient-specific, and manually modulated by a clinician. As such, patient outcomes are highly inconsistent, with around 30% or more of individuals experiencing no response to the therapy for most targeted conditions.
At RyvivyR, we design neural foundation models and machine learning methods specialized for analyzing and interpreting neural data, as well as controlling neural responses in real-time. Our approaches are robust, ensuring specific patient-to-patient nuances are accounted for, gearing up for the next generation of individualized neurostimulation treatment.
Our proprietary software is installed on existing invasive and non-invasive stimulation devices, set up in closed-loop with a neural recording device. These algorithms modulate the applied stimulus administered by the hardware in real-time based on the everchanging pattern of neural activity recorded from the brain. Stimulation can be applied to escape a pathological pattern of neural activity, while simultaneously building a map of the electrical activity pattern that can be readily saved for evaluation of prognosis.
Neural data can be recorded across many different modalities, including invasive electrophysiology, non-invasive EEG, imaging, stimulation-response data, and emerging BCI systems. Despite differences in how these signals are collected, they often reflect a shared underlying structure -- the evolving dynamics of the nervous system. Our methods for creating modality-agnostic neural foundation models are designed to learn this shared structure across diverse datasets, devices, and clinical contexts. By training models that generalize across recording modalities, patient populations, and disease states, we aim to create flexible AI systems that can support discovery, personalization, and translation across neuroscience, neurotechnology, and clinical care.
Disorders of consciousness (DOC) are a class of disorders that inhibit a victim's wakefulness and or self-awareness. They often arise from a traumatic brain injury (TBI), resulting in coma, vegetative states, and minimally conscious states, as well as from complications due to general anesthesia. The latter preventing a victim to properly wake up or to suffer from post operative delirium. Traditional treatments are often ineffective, with a 5% success rate at best.
It is well-accepted in neuroscience that the brain exists in a variety of states, represented by the patterns of electrical activity that are expressed by populations of neurons. Furthermore, brain function (cognition, perception, movement, sleep, etc.) is often the result of switching between these different states in time -- a phenomenon called metastability. Understanding these neural dynamics is the key to developing many future, personalized treatment options However, learning such dynamics is notoriously difficult, requiring more sophisticated tools to help researchers and clinicians.
It is our goal that with our combined efforts, yours and ours, we can further rid the world of disability, while continually learning about the neural processes that make us who we are.
