Distinct Lateral Prefrontal Regions Are Organized in an Anterior-Posterior Functional Gradient.

The dorsolateral prefrontal cortex (dlPFC) is composed of multiple anatomically defined regions involved in higher-order cognitive processes, including working memory and selective attention. It is organized in an anterior-posterior global gradient where posterior regions track changes in the environment, whereas anterior regions support abstract neural representations. However, it remains unknown if such a global gradient results from a smooth gradient that spans regions or an emergent property arising from functionally distinct regions, that is, an areal gradient. Here, we recorded single neurons in the dlPFC of nonhuman primates trained to perform a memory-guided saccade task with an interfering distractor and analyzed their physiological properties along the anterior-posterior axis. We found that these physiological properties were best described by an areal gradient. Further, population analyses revealed that there is a distributed representation of spatial information across the dlPFC. Our results validate the functional boundaries between anatomically defined dlPFC regions and highlight the distributed nature of computations underlying working memory across the dlPFC.SIGNIFICANCE STATEMENT Activity of frontal lobe regions is known to possess an anterior-posterior functional gradient. However, it is not known whether this gradient is the result of individual brain regions organized in a gradient (like a staircase), or a smooth gradient that spans regions (like a slide). Analysis of physiological properties of individual neurons in the primate frontal regions suggest that individual regions are organized as a gradient, rather than a smooth gradient. At the population level, working memory was more prominent in posterior regions, although it was also present in anterior regions. This is consistent with the functional segregation of brain regions that is also observed in other systems (i.e., the visual system).

Comparing representations and computations in single neurons versus neural networks.

Single-neuron-level explanations have been the gold standard in neuroscience for decades. Recently, however, neural-network-level explanations have become increasingly popular. This increase in popularity is driven by the fact that the analysis of neural networks can solve problems that cannot be addressed by analyzing neurons independently. In this opinion article, I argue that while both frameworks employ the same general logic to link physical and mental phenomena, in many cases the neural network framework provides better explanatory objects to understand representations and computations related to mental phenomena. I discuss what constitutes a mechanistic explanation in neural systems, provide examples, and conclude by highlighting a number of the challenges and considerations associated with the use of analyses of neural networks to study brain function.

Neural correlates of learning in a linear discriminant analysis brain-computer interface paradigm.

Objective.With practice, the control of brain-computer interfaces (BCI) would improve over time; the neural correlate for such learning had not been well studied. We demonstrated here that monkeys controlling a motor BCI using a linear discriminant analysis (LDA) decoder could learn to make the firing patterns of the recorded neurons more distinct over a short period of time for different output classes to improve task performance.Approach.Using an LDA decoder, we studied two Macaque monkeys implanted with microelectrode arrays as they controlled the movement of a mobile robotic platform. The LDA decoder mapped high-dimensional neuronal firing patterns linearly onto a lower-dimensional linear discriminant (LD) space, and we studied the changes in the spatial coordinates of these neural signals in the LD space over time, and their correspondence to trial performance. Direction selectivity was quantified with permutation feature importance (FI).Main results.We observed that, within individual sessions, there was a tendency for the points in the LD space encoding different directions to diverge, leading to fewer misclassification errors, and, hence, improvement in task accuracy. Accuracy was correlated with the presence of channels with strong directional preference (i.e. high FI), as well as a varied population code (i.e. high variance in FI distribution).Significance.We emphasized the importance of studying the short-term/intra-sessional variations in neural representations during the use of BCI. Over the course of individual sessions, both monkeys could modulate their neural activities to create increasingly distinct neural representations.

Information sparseness in cortical microelectrode channels while decoding movement direction using an artificial neural network.

Implanted microelectrode arrays can directly pick up electrode signals from the primary motor cortex (M1) during movement, and brain-machine interfaces (BMIs) can decode these signals to predict the directions of contemporaneous movements. However, it is not well known how much each individual input is responsible for the overall performance of a BMI decoder. In this paper, we seek to quantify how much each channel contributes to an artificial neural network (ANN)-based decoder, by measuring how much the removal of each individual channel degrades the accuracy of the output. If information on movement direction was equally distributed among channels, then the removal of one would have a minimal effect on decoder accuracy. On the other hand, if that information was distributed sparsely, then the removal of specific information-rich channels would significantly lower decoder accuracy. We found that for most channels, their removal did not significantly affect decoder performance. However, for a subset of channels (16 out of 61), removing them significantly reduced the decoder accuracy. This suggests that information is not uniformly distributed among the recording channels. We propose examining these channels further to optimize BMIs more effectively, as well as understand how M1 functions at the neuronal level.

A nonlinear hidden layer enables actor-critic agents to learn multiple paired association navigation.

Navigation to multiple cued reward locations has been increasingly used to study rodent learning. Though deep reinforcement learning agents have been shown to be able to learn the task, they are not biologically plausible. Biologically plausible classic actor-critic agents have been shown to learn to navigate to single reward locations, but which biologically plausible agents are able to learn multiple cue-reward location tasks has remained unclear. In this computational study, we show versions of classic agents that learn to navigate to a single reward location, and adapt to reward location displacement, but are not able to learn multiple paired association navigation. The limitation is overcome by an agent in which place cell and cue information are first processed by a feedforward nonlinear hidden layer with synapses to the actor and critic subject to temporal difference error-modulated plasticity. Faster learning is obtained when the feedforward layer is replaced by a recurrent reservoir network.

STEER: 3D Printed Guide for Nerve Regrowth Control and Neural Interface in Non-Human Primate Model.

OBJECTIVE: Peripheral neural interface (PNI) with a stable integration of synthetic elements with neural tissue is key for successfulneuro-prosthetic applications. An inevitable phenomenon of reactive fibrosis is a primary hurdle for long term functionality of PNIs. This proof-of-concept study aimed to fabricate and test a novel, stable PNI that harnesses fibro-axonal outgrowth at the nerve end and includes fibrosis in the design. METHODS: Two non-human primates were implanted with Substrate-guided, Tissue-Electrode Encapsulation and Integration (STEER) PNIs. The implant included a 3D printed guide that strove to steer the regrowing nerve towards encapsulation of the electrodes into a fibro-axonal tissue. After four months from implantation, we performed electrophysiological measurements to test STEER's functionality and examined the macro and micro- morphology of the outgrowth tissue. RESULTS: We observed a highly structured fibro-axonal composite within the STEER PNI. A conduction of intracranially generated action potentials was successfully recorded across the neural interface. Immunohistology demonstrated uniquely configured laminae of myelinated axons encasing the implant. CONCLUSION: STEER PNI reconfigured the structure of the fibro-axonal tissue and facilitated long-term functionality and stability of the neural interface. SIGNIFICANCE: The results point to the feasibility of our concept for creating a stable PNI with long-term electrophysiologic functionality by using simple design and materials.

Intention Estimation Based Adaptive Unscented Kalman Filter for Online Neural Decoding.

The commonly used fixed discrete Kalman filters (DKF) in neural decoders do not generalize well to the actual relationship between neuronal firing rates and movement intention. This is due to the underlying assumption that the neural activity is linearly related to the output state. They also face the issues of requiring large amount of training datasets to achieve a robust model and a degradation of decoding performance over time. In this paper, an adaptive adjustment is made to the conventional unscented Kalman filter (UKF) via intention estimation. This is done by incorporating a history of newly collected state parameters to develop a new set of model parameters. At each time point, a comparative weighted sum of old and new model parameters using matrix squared sums is used to update the neural decoding model parameters. The effectiveness of the resulting adaptive unscented Kalman filter (AUKF) is compared against the discrete Kalman filter and unscented Kalman filter-based algorithms. The results show that the proposed new algorithm provides higher decoding accuracy and stability while requiring less training data.

Decoding movement direction from cortical microelectrode recordings using an LSTM-based neural network.

Brain-machine interfaces (BMIs) allow individuals to communicate with computers using neural signals, and Kalman Filter (KF) are prevailingly used to decode movement directions from these neural signals. In this paper, we implemented a multi-layer long short-term memory (LSTM)based artificial neural network (ANN) for decoding BMI neural signals. We collected motor cortical neural signals from a nonhuman primate (NHP), implanted with microelectrode array (MEA) while performing a directional joystick task. Next, we compared the LSTM model in decoding the joystick trajectories from the neural signals against the prevailing KF model. The results showed that the LSTM model yielded significantly improved decoding accuracy measured by mean correlation coefficient (0.84, p < 10-7) than the KF model (0.72). In addition, using a principal component analysis (PCA)-based dimensionality reduction technique yielded slightly deteriorated accuracies for both the LSTM (0.80) and KF (0.70) models, but greatly reduced the computational complexity. The results showed that the LSTM decoding model holds promise to improve decoding in BMIs for paralyzed humans.

Minimally dependent activity subspaces for working memory and motor preparation in the lateral prefrontal cortex.

The lateral prefrontal cortex is involved in the integration of multiple types of information, including working memory and motor preparation. However, it is not known how downstream regions can extract one type of information without interference from the others present in the network. Here, we show that the lateral prefrontal cortex of non-human primates contains two minimally dependent low-dimensional subspaces: one that encodes working memory information, and another that encodes motor preparation information. These subspaces capture all the information about the target in the delay periods, and the information in both subspaces is reduced in error trials. A single population of neurons with mixed selectivity forms both subspaces, but the information is kept largely independent from each other. A bump attractor model with divisive normalization replicates the properties of the neural data. These results provide new insights into neural processing in prefrontal regions.

A Flexi-PEGDA Upconversion Implant for Wireless Brain Photodynamic Therapy.

Near-infrared (NIR) activatable upconversion nanoparticles (UCNPs) enable wireless-based phototherapies by converting deep-tissue-penetrating NIR to visible light. UCNPs are therefore ideal as wireless transducers for photodynamic therapy (PDT) of deep-sited tumors. However, the retention of unsequestered UCNPs in tissue with minimal options for removal limits their clinical translation. To address this shortcoming, biocompatible UCNPs implants are developed to deliver upconversion photonic properties in a flexible, optical guide design. To enhance its translatability, the UCNPs implant is constructed with an FDA-approved poly(ethylene glycol) diacrylate (PEGDA) core clad with fluorinated ethylene propylene (FEP). The emission spectrum of the UCNPs implant can be tuned to overlap with the absorption spectra of the clinically relevant photosensitizer, 5-aminolevulinic acid (5-ALA). The UCNPs implant can wirelessly transmit upconverted visible light till 8 cm in length and in a bendable manner even when implanted underneath the skin or scalp. With this system, it is demonstrated that NIR-based chronic PDT is achievable in an untethered and noninvasive manner in a mouse xenograft glioblastoma multiforme (GBM) model. It is postulated that such encapsulated UCNPs implants represent a translational shift for wireless deep-tissue phototherapy by enabling sequestration of UCNPs without compromising wireless deep-tissue light delivery.

Sparse Ensemble Machine Learning to Improve Robustness of Long-Term Decoding in iBMIs.

This paper presents a novel sparse ensemble based machine learning approach to enhance robustness of intracortical Brain Machine Interfaces (iBMIs) in the face of non-stationary distribution of input neural data across time. Each classifier in the ensemble is trained on a randomly sampled (with replacement) set of input channels. These sparse connections ensure that with a high chance, few of the base classifiers should be less affected by the variations in some of the recording channels. We have tested the generality of this technique on different base classifiers - linear discriminant analysis (LDA), support vector machine (SVM), extreme learning machine (ELM) and multilayer perceptron (MLP). Results show decoding accuracy improvements of up to ≈21 %, 13%, 19%, 10% in non-human primate (NHP) A and 7%, 9%, 7%, 9% in NHP B across test days while using the sparse ensemble approach over a single classifier model for LDA, SVM, ELM and MLP algorithms respectively. Furthermore, improvements of up to ≈7(14)%, 8(15)%, 9(19)%, 7(15)% in NHP A and 8(15)%, 12(20)%, 15(23)%, 12(19)% in NHP B over Random Forest (Long-short Term Memory) have been obtained by sparse ensemble LDA, SVM, ELM, MLP respectively.

Time-invariant working memory representations in the presence of code-morphing in the lateral prefrontal cortex.

Maintenance of working memory is thought to involve the activity of prefrontal neuronal populations with strong recurrent connections. However, it was recently shown that distractors evoke a morphing of the prefrontal population code, even when memories are maintained throughout the delay. How can a morphing code maintain time-invariant memory information? We hypothesized that dynamic prefrontal activity contains time-invariant memory information within a subspace of neural activity. Using an optimization algorithm, we found a low-dimensional subspace that contains time-invariant memory information. This information was reduced in trials where the animals made errors in the task, and was also found in periods of the trial not used to find the subspace. A bump attractor model replicated these properties, and provided predictions that were confirmed in the neural data. Our results suggest that the high-dimensional responses of prefrontal cortex contain subspaces where different types of information can be simultaneously encoded with minimal interference.

Towards Intelligent Intracortical BMI (i 2BMI): Low-Power Neuromorphic Decoders That Outperform Kalman Filters.

Fully-implantable wireless intracortical Brain Machine Interfaces (iBMI) is one of the most promising next frontiers in the nascent field of neurotechnology. However, scaling the number of channels in such systems by another 10× is difficult due to power and bandwidth requirements of the wireless transmitter. One promising solution for that is to include more processing, up to the decoder, in the implant so that transmission data-rate is reduced drastically. Earlier work on neuromorphic decoder chips only showed classification of discrete states. We present results for continuous state decoding using a low-power neuromorphic decoder chip termed Spike-input Extreme Learning Machine (SELMA) that implements a nonlinear decoder without memory and its memory-based version with time-delayed bins, SELMA-bins. We have compared SELMA, SELMA-bins against state-of-the-art Steady-State Kalman Filter (SSKF), a linear decoder with memory, across two different datasets involving a total of 4 non-human primates (NHPs). Results show at least a 10% (20%) or more increase in the fraction of variance accounted for (FVAF) by SELMA (SELMA-bins) over SSKF across the datasets. Estimated energy consumption comparison shows SELMA (SELMA-bins) consuming ≈ 9 nJ/update (23 nJ/update) against SSKF's ≈ 7.4 nJ/update for an iBMI with a 10 degree of freedom control. Thus, SELMA yields better performance against SSKF while consuming energy in the same range as SSKF whereas SELMA-bins performs the best with moderately increased energy consumption, albeit far less than energy required for raw data transmission. This paves the way for reducing transmission data rates in future scaled iBMI systems.

Clustering Neural Patterns in Kernel Reinforcement Learning Assists Fast Brain Control in Brain-Machine Interfaces.

Neuroprosthesis enables the brain control on the external devices purely using neural activity for paralyzed people. Supervised learning decoders recalibrate or re-fit the discrepancy between the desired target and decoder's output, where the correction may over-dominate the user's intention. Reinforcement learning decoder allows users to actively adjust their brain patterns through trial and error, which better represents the subject's motive. The computational challenge is to quickly establish new state-action mapping before the subject becomes frustrated. Recently proposed quantized attention-gated kernel reinforcement learning (QAGKRL) explores the optimal nonlinear neural-action mapping in the Reproducing Kernel Hilbert Space (RKHS). However, considering all past data in RKHS is less efficient and sensitive to detect the new neural patterns emerging in brain control. In this paper, we propose a clustering-based kernel RL algorithm. New neural patterns emerge and are clustered to represent the novel knowledge in brain control. The current neural data only activate the nearest subspace in RKHS for more efficient decoding. The dynamic clustering makes our algorithm more sensitive to new brain patterns. We test our algorithm on both the synthetic and real-world spike data. Compared with QAGKRL, our algorithm can achieve a quicker knowledge adaptation in brain control with less computational complexity.

Fully automated leg tracking of Drosophila neurodegeneration models reveals distinct conserved movement signatures.

Some neurodegenerative diseases, like Parkinsons Disease (PD) and Spinocerebellar ataxia 3 (SCA3), are associated with distinct, altered gait and tremor movements that are reflective of the underlying disease etiology. Drosophila melanogaster models of neurodegeneration have illuminated our understanding of the molecular mechanisms of disease. However, it is unknown whether specific gait and tremor dysfunctions also occur in fly disease mutants. To answer this question, we developed a machine-learning image-analysis program, Feature Learning-based LImb segmentation and Tracking (FLLIT), that automatically tracks leg claw positions of freely moving flies recorded on high-speed video, producing a series of gait measurements. Notably, unlike other machine-learning methods, FLLIT generates its own training sets and does not require user-annotated images for learning. Using FLLIT, we carried out high-throughput and high-resolution analysis of gait and tremor features in Drosophila neurodegeneration mutants for the first time. We found that fly models of PD and SCA3 exhibited markedly different walking gait and tremor signatures, which recapitulated characteristics of the respective human diseases. Selective expression of mutant SCA3 in dopaminergic neurons led to a gait signature that more closely resembled those of PD flies. This suggests that the behavioral phenotype depends on the neurons affected rather than the specific nature of the mutation. Different mutations produced tremors in distinct leg pairs, indicating that different motor circuits were affected. Using this approach, fly models can be used to dissect the neurogenetic mechanisms that underlie movement disorders.

Graded Memory: A Cognitive Category to Replace Spatial Sustained Attention and Working Memory
.

In this opinion article we challenge the commonly-held notion that visuospatial working memory and visuospatial sustained selective attention are two ontologically different cognitive categories. We start by discussing the general idea of cognitive categories, and then review some of the key behavioral and neural evidence both in favor of and against the separability of these processes. We then discuss a theoretical framework that could be useful for understanding the neural implementations of cognitive categories. We conclude that the evidence is insufficient to support the assumption that spatial working memory and spatial sustained attention are independent categories, and that further experimentation is necessary to determine the ontological independence of the two processes.

Robust Local Field Potential-based Neural Decoding by Actively Selecting Discriminative Channels.

Local field potentials (LFPs) have been proposed as a neural decoding signal to compensate for spike signal deterioration in invasive brain-machine interface applications. However, the presence of redundancy among LFP signals at different frequency bands across multiple channels may affect the decoding performance. In order to remove redundant LFP channels, we proposed a novel Fisher-distance ratio-based method to actively batch select discriminative channels to maximize the separation between classes. Experimental evaluation was conducted on 5 non-consecutive days of data from a non-human primate. For data from each day, the first experimental session was used to generate the training model, which was then used to perform 4-class decoding of signals from other sessions. Decoding achieved an average accuracy of 79.55%, 79.02% and 79.40% using selected LFP channels for beta, low gamma and high gamma frequency bands, respectively. Compared with decoding using full LFP channels, decoding using selected LFP channels in high gamma band resulted in an increase of 8.67% in accuracy, even if this accuracy was still 7.26% lower than that of spike-based decoding. These results demonstrate the effectiveness of the proposed method in selecting discriminative LFP channels for neural decoding.

Mixed selectivity morphs population codes in prefrontal cortex.

The prefrontal cortex maintains working memory information in the presence of distracting stimuli. It has long been thought that sustained activity in individual neurons or groups of neurons was responsible for maintaining information in the form of a persistent, stable code. Here we show that, upon the presentation of a distractor, information in the lateral prefrontal cortex was reorganized into a different pattern of activity to create a morphed stable code without losing information. In contrast, the code in the frontal eye fields persisted across different delay periods but exhibited substantial instability and information loss after the presentation of a distractor. We found that neurons with mixed-selective responses were necessary and sufficient for the morphing of code and that these neurons were more abundant in the lateral prefrontal cortex than the frontal eye fields. This suggests that mixed selectivity provides populations with code-morphing capability, a property that may underlie cognitive flexibility.

Boosting performance in brain-machine interface by classifier-level fusion based on accumulative training models from multi-day data.

The nonstationarity of neural signal is still an unresolved issue despite the rapid progress made in brain-machine interface (BMI). This paper investigates how to utilize the rich information and dynamics in multi-day data to address the variability in day-to-day signal quality and neural tuning properties. For this purpose, we propose a classifier-level fusion technique to build a robust decoding model by jointly considering the classifier outputs from multiple base-training models using multi-day data collected prior to test day. The data set used in this study consisted of recordings of 8 days from a non-human primate (NHP) during control of a mobile robot using a joystick. Offline analysis demonstrates the superior performance of the proposed method which results in 4.4% and 13.10% improvements in decoding (significant by one-way ANOVA and post hoc t-test) compared with the two baseline methods: 1) concatenating data from multiple days based on common effective channels, and 2) averaging accuracies across all base-training models. These results further validate the effectiveness of proposed method without recalibration of the model.

Stop state classification in intracortical brain-machine-interface.

Invasive brain-machine-interface (BMI) has the prospect to empower tetraplegic patients with independent mobility through the use of brain-controlled wheelchairs. For the practical and long-term use of such control systems, the system has to distinguish between stop and movement states and has to be robust to overcome non-stationarity in the brain signals. In this work, we investigates the non-stationarity of the stop state on neural data collected from a macaque trained to control a robotic platform to stop and move in left, right, forward directions We then propose a hybrid approach that employs both random forest and linear discriminant analysis (LDA). Using this approach, we performed offline decoding on 8 days of data collected over the course of three months during joystick control of the robotic platform. We compared the results of using the proposed approach with the use of LDA alone to perform direct classifications of stop, left, right and forward. The results showed an average performance increment of 22.7% using the proposed hybrid approach. The results yielded significant improvements during sessions where LDA showed a heavy bias towards the stop state. This suggests that the proposed hybrid approach addresses the non-stationarity in the stop state and subsequently facilitates a more accurate decoding of the movement states.