Power-efficient hardware and brain-computer interfaces… Understanding the timescales of neural dynamics… Examining gut-to-brain progression of α-synuclein pathology in a rodent model of Parkinson’s disease… The neural origins of magneto- and electro-encephalography…
Improving intracortical brain–computer interfaces through new power designs
Recent technological advances have given intracortical brain-computer interfaces (iBCIs) enormous potential for widespread human clinical use. But current devices are limited by the number of recording channels, which largely depends on the power budget of the system. What if we could create new designs that save on power, but still allow for efficient brain-computer interface units? Researchers at Stanford University are working on exactly that. The lab of Dr. Krishna Shenoy
at Stanford University recently published a paper that examined which design specifications (specifically in regard to power) can be relaxed without sacrificing brain-interface device performance. Dr. Leigh Hochberg
from Brown University and Harvard Medical School is also working on this problem with Dr. Shenoy’s team. Here, researchers proposed a new power-efficient approach by analyzing signals collected from experimental iBCI measurements in rhesus macaques and a single BrainGate2 clinical trial participant with a spinal cord injury implanted with a 96-channel Utah multielectrode array. The paper focused on three questions: 1) what type of signal should be recorded from the brain; 2) how reliable this signal should be; and 3) how the resulting new specifications will affect the iBCI neural interface design, with particular emphasis on its power consumption, in order to understand the trade-offs between signal quality and decoder performance. Altogether, their analysis led them to propose less stringent, custom circuit design parameters related to signal collection, processing, and decoding that enable a more power-efficient iBCI hardware design. Overall, their minimalistic design may lead to improved iBCIs for clinical use and suggest that conventional iBCI parameters can be relaxed without loss of performance. Although this study primarily looked at decoding movement, these results have important implications for other types of clinically relevant neural interfaces, such as those used in retinal prostheses and the peripheral nervous system.
The brain is incredibly heterogeneous, motivating neuroscientists to seek out a better way to organize it into distinct areas. Perhaps unsurprisingly, each brain area performs its own unique neural computations. For example, we know that the timescale at which neural responses fluctuate differs across areas, and recent studies have proposed the progression of timescales in neural responses as an organizing principle. But up until now, the relationships between such timescales and their dependence on response properties of individual neurons have been largely unknown, making it impossible to link the neural basis of these principles to other aspects of neural processing and to behavior. The lab of Dr. Alireza Soltani at Dartmouth College has recently published a study that developed a novel and comprehensive method to simultaneously estimate multiple timescales in neuronal dynamics and integration of signals (along with selectivity to those signals) in a decision-making task. Essentially, the overarching goal of work such as this is to understand how activity at the level of individual neurons gives rise to more sophisticated computational processes. In this study, researchers analyzed the activity of 866 single neurons in four cortical areas (lateral intraparietal cortex; dorsomedial prefrontal cortex; dorsolateral prefrontal cortex; and anterior cingulate cortex, ACC) in rhesus macaques, which are implicated in decision making and reinforcement learning. During the decision-making task, they estimated the timescales associated with the intrinsic and seasonal autoregressive (AR) components of the neural responses. Other factors were captured by using two exponential memory-trace components (filters) to describe the fluctuations of each neuron’s response at specific timepoints. After the validation of this estimation method and fitting procedure, the researchers used many different models (32) to fit individual neurons’ response in order to identify the best model for each neuron based on multiple factors. Overall, their results suggest that there are separate mechanisms underlying the generation of multiple timescales. Interestingly, neurons with no choice selectivity tended to integrate reward feedback over time. This research is valuable because it demonstrates that the timescales of neural dynamics across cortical areas can in fact be used as an organizational principle to understand brain computational processes.
From the gut to the brain: the progression of protein aggregations associated with Parkinson’s disease
Neurodegenerative conditions such as Parkinson’s disease (PD) can be characterized in part by the aggregations of protein, especially α-synuclein (α-Syn) protein. In a world with a growing aged population, determining the pathogenesis of age-related brain disorders is particularly important. In regard to PD, there is mounting evidence to suggest that α-Syn protein accumulation associated with the motor symptoms of PD begins, not in the brain, but the gastrointestinal (GI) system. Dr. Viviana Gradinaru and her team at California Institute of Technology recently provide evidence for this relationship. Specifically, they asked whether this protein accumulation starts in the GI and ascends to the brain via enteric nervous system (ENS)-innervating vagal fibers. First, they injected α-Syn preformed fibrils in the wall of the duodenum of young adult mice. From there, they watched the progression of α-Syn proteins and their effects on the GI system. Next, they used an adeno-associated virus (AAV) to express the GBA1 gene (which encodes a lysosomal enzyme called GCase) to determine if it can ameliorate peripheral synucleinopathology. Inoculation of the GI tract led to α-Syn protein accumulation within the brainstem of aged mice, but not younger mice, demonstrating that GCase plays an important role in this process. This work is important because it shows that aging may put an individual at risk for gut to brain protein transmission processes. Overall, understanding how vulnerabilities within the peripheral systems lead to α-Syn accumulation show promise as an early-detection technique for neurodegenerative conditions like PD.
Human Neocortical Neurosolver software helps interpret the origin of human brain imaging techniques
Magneto- and electro-encephalography (MEG/EEG) technology provides useful information about the brain. This is exciting because without these forms of technology, it would be harder to understand the brain in living humans. However, it is not well understood how these larger, macroscopic signals relate to the underlying cellular- and circuit-level mechanisms. This limitation hugely constrains the ability of MEG/EEG techniques to reveal novel principles of information processing. Further, it limits our ability to translate MEG/EEG findings into new therapies for disorders of the brain, which are largely due to dysfunctional neural circuits. Dr. Stephanie Jones and her team of researchers at Brown University are working to bridge the gap between experimental and computational neuroscience in order to better study human brain dynamics. They are helping to achieve this goal with their recently published, novel software program called the Human Neocortical Neurosolver (HNN). This user-friendly program is helping researchers and clinicians interpret the neural circuit origins of MEG/EEG signals. The model is able to simulate electrical activity of cells in the neocortex; then, by feeding real human EEG/MEG data into the model, the researchers can predict patterns of circuit-level activity that may be the neural origins of the EEG/MEG signal data. Essentially, now these MEG/EEG signals are working harder than ever before to tell us more precise information about the brain at the level of the circuit. Moreover, we now have a refined and user-friendly software program that allows us to do this more effectively, and this tool is open source and available for installation today. Find out more by listening to Dr. Jones’ talk at the 2020 Allen Institute Brain Modeling Workshop here.