Maryam Shanechi, University of Southern California
With recent technological advances, we can now record neural activity from the brain, and manipulate this activity with electrical or optogenetic stimulation in real time. These capabilities have brought the concept of brain-machine interfaces (BMI) closer to clinical viability than ever before. BMIs are systems that monitor and interact with the brain to restore lost function, treat neurological disorders, or enhance human performance.
BMIs can be categorized into two broad classes. The first class of BMIs allow people to directly control external devices by modulating their neural activity. Examples of these are motor BMIs that restore movement to paralyzed patients or speech BMIs that enable locked-in patients to speak by controlling a voice synthesizer. In these BMIs, the brain acts as a controller; it uses neural activity to control an external device while receiving visual or auditory feedback. The second class of BMIs aim to control the brain state using stimulation input. Examples of these are deep brain stimulation systems for treatment of depression. In contrast to the first class, in these BMIs the brain is the system to be controlled based on feedback of neural activity.
While these are diverse applications, BMIs can all be viewed as feedback-control systems in which the brain is “in the control loop”. As such, BMIs can be designed using a common methodology at the interface of neuroscience, control theory, and statistical inference. Moreover, given the direct access BMIs provide to both monitor and manipulate neural activity, they can serve as a scientific tool to investigate neural mechanisms underlying brain function or dysfunction. This session brings together experts across neuroscience and engineering to discuss state-of-the-art approaches and future opportunities in three main BMI domains: motor BMIs, speech BMIs, and BMIs for treatment of depression.
Motor BMIs decode the intended movement of a subject from their neural activity. In recent years, the feedback-control view of BMIs combined with novel statistical models and adaptive algorithms have led to improved decoders that allow subjects to operate a computer interface or robotic arm using their thoughts alone. In addition to having the potential to provide a therapy for paralyzed people, BMIs also create a unique platform for basic neuroscience, which has been leveraged to discover fundamental neural mechanisms underlying neuroprosthetic control that may resemble those at play during natural motor control and skill learning. Speech BMIs aim to decode the verbal output from brain areas governing the uniquely human ability for language. Recent research has uncovered the cortical representations of the auditory and vocal tract movements that are responsible for fluent speech, with the promise to enable decoding in communication disorders.
In addition to BMIs for control of external devices, a new generation of BMIs are being developed with the goal of treating a dysfunctional brain state in neurological disorders such as depression. These BMIs would decode a patient’s neuropsychiatric state from neural activity as feedback to adjust the pattern of electrical stimulation applied to the brain for treatment. Recent computational methods enable the real-time decoding of mood from human intracranial activity and characterize the effect of stimulation on neural activity. These advances could enable new precisely-tailored therapies for neuropsychiatric disorders such as depression and anxiety.
Taken together, BMI systems can revolutionize treatment across a diverse set of neurological diseases and injuries and shed light on complex neural processes underlying behavior.