February 2022 issue
February 28, 2022 | Neuroscientists from across the globe participated in the annual meeting of the American Association for the Advancement of Science earlier this month. The meeting was conducted entirely online because of the coronavirus pandemic.
A session devoted to decoding brain signals was moderated by Kareem Zaghloul from the National Institute for Neurological Disorders and Stroke. Helen Bronte-Stewart from Stanford University gave a presentation on the subject of neural controllers for closed-loop neurostimulation in Parkinson’s disease. “One of the wonderful things about implanting deep brain stimulation leads is that we have access to the neural activity in deep brain structures in human subjects,” she said. “If we’re going to develop neural controllers for closed loop DBS we need to record these signals and decode which neural signals are relevant to pathological motor behaviors and which are not.”
Bronte-Stewart stressed the relevance of beta-band oscillopathy—a disorder of neuronal oscillations—in PD. It’s evident in the bilateral sensorimotor network and it increases with disease progression, more so in the affected subthalamic nucleus, she said. The beta abnormality is indirectly related to the hypokinetic aspects of the disease, such as bradykinesia—slowness of movement—and rigidity. It’s attenuated during resting tremor and importantly, also attenuated in a dose-dependent manner by both DBS and dopaminergic medication. Interestingly, the degree of attenuation is related to the degree of improvement in both bradykinesia and rigidity, she said. She believes that beta band power is a useful neural controller for closed-loop DBS. She also believes beta band burst durations are a useful signal.
Helen Mayberg from Mt. Sinai gave a presentation on tracking brain dynamics to optimize DBS for depression. “Treatment resistant depression is more than just a regular depression,” she said. “It’s really agony that robs attention and energy.”
Mayberg pointed out that severe depression can lead to motor impairment. “The idea that you can’t move away from mental pain is, in fact, the disorder,” she said. “As we think about the application of invasive neuromodulation to this kind of nebulous or ethereal problem—the inability to put thoughts into action because of pain—we need to reduce it to elements that might be tractable and measurable.
Mayberg stressed that we need to have an understanding of what “well” looks like. Is it just the absence of pain or is wellness not relapsing? What brain readouts do we want?
Mayberg described some of her early work with DBS of the sub colossal cingulate—near area 25—with 130 Hz stimulation delivered 24/7 in an effort to decrease activity there, which had an anti-depressant effect. She related that over time, patients that get well stay well. This was true in patients that stopped responding to any available treatments, including electroconvulsive therapy. She noted that the SCC and adjacent white matter is the locus of convergence of four white matter bundles that acts as a depression switch. “When you stimulate there patients’ depressive state actually turns off within a minute,” she said. She reported that recovery isn’t linear—different patients have a variable recovery curve that extends over weeks to months.
In order to devise a closed-loop neuromodulation therapy for treatment-resistant depression, Mayberg said that we need to find useful biomarkers. “How do we distinguish sick vs. well? How can we discriminate at the brain level distress vs. impending relapse?” she asked.
With two electrodes implanted, her team looked at LFP readouts before and after stimulation in the SCC white matter. Using machine learning, they were able to build a classifier that could distinguish between the naive depressed state and the changed reset state. That signal had features that were dominant in left beta and right alpha frequency ranges. The magnitude of left beta decrease correlated with patients’ improvement weeks later, though not after six months. They also built a different classifier that could track early and late responders. Using video interviews of patients, her team identified specific facial movements that were indicative of sick vs. well. Both the LFP signals and the facial classifier are good candidates for closed loop DBS inputs, she said.
Leigh Hochberg from Mass General and Brown University gave a presentation on the subject of intracortical brain-computer interfaces for restoration of communication and mobility. He began with a statement of goals. “For somebody who is progressively losing the ability to move or speak—somebody with ALS for example, I’d like to tell that person that they’ll never lose the ability to communicate,” he said. “For somebody who has just had a brainstem stroke and has become locked in, I’d like to be able to tell that person they will be able to communicate again easily tomorrow. And for somebody who may have had a spinal cord injury or a stroke, I’d like to be able to assure them that they’re going to be able to move again easily, intuitively, rapidly, tomorrow.”
Hochberg said that BCIs have three major components: a neural sensor, a decoder of some sort, and an effector. Key considerations are which signals from the brain should be recorded and which brain areas to place electrodes. He noted that there are a dozen or more related to the intention to move one’s hand.
Describing the “black box” of computational neuroscience as one of the most fun jobs in the world, Hochberg explained the process of understanding the language of the nervous system, including studying the firing rate of individual neurons.
He said there have been 14 people in the BrainGate trial so far plus a dozen or so others implanted with similar or systems. He showed a video of one user with ALS who was able to type on a tablet computer just by thinking. Another user was able to type 94 characters per minute with greater than 90 percent accuracy.
Hochberg ended his talk with a video of the first-in-human broadband intracortical wireless recording.
