Caleb Kemere

Awake Hippocampal Sharp-Wave Ripples Support Spatial Memory

Spatial Memory Perturbation The hippocampus is important for learning and memory. However, it is not clear which patterns of neural activity in the hippocampus support specific mnemonic functions. Jadhav et al. (p. 1454, published online 3 May) developed a real-time analysis system to detect and selectively interrupt a certain type of hippocampal neuronal network event—sharp-wave ripples—during learning. In awake animals, loss of sharp-wave ripples and associated memory replay activity caused a learning deficit specific to spatial working memory but had no effect on reference memory. This learning deficit was present despite the preservation of place-field representations and replay activity during rest. The neuronal “replay” of past experience may allow animals to retrieve specific memories and use them to guide behavior. The hippocampus is critical for spatial learning and memory. Hippocampal neurons in awake animals exhibit place field activity that encodes current location, as well as sharp-wave ripple (SWR) activity during which representations based on past experiences are often replayed. The relationship between these patterns of activity and the memory functions of the hippocampus is poorly understood. We interrupted awake SWRs in animals learning a spatial alternation task. We observed a specific learning and performance deficit that persisted throughout training. This deficit was associated with awake SWR activity, as SWR interruption left place field activity and post-experience SWR reactivation intact. These results provide a link between awake SWRs and hippocampal memory processes, which suggests that awake replay of memory-related information during SWRs supports learning and memory-guided decision-making.

Power feasibility of implantable digital spike-sorting circuits for neural prosthetic systems

A new class of neural prosthetic systems aims to assist disabled patients by translating cortical neural activity into control signals for prosthetic devices. Based on the success of proof-of-concept systems in the laboratory, there is now considerable interest in increasing system performance and creating implantable electronics for use in clinical systems. A critical question that impacts system performance and the overall architecture of these systems is whether it is possible to identify the neural source of each action potential (spike sorting) in real-time and with low power. Low power is essential both for power supply considerations and heat dissipation in the brain. In this paper we report that state-of-the-art spike sorting algorithms are not only feasible using modern complementary metal oxide semiconductor very large scale integration processes, but may represent the best option for extracting large amounts of data in implantable neural prosthetic interfaces.

Detecting Neural-State Transitions Using Hidden Markov Models for Motor Cortical Prostheses

Neural prosthetic interfaces use neural activity related to the planning and perimovement epochs of arm reaching to afford brain-directed control of external devices. Previous research has primarily centered on accurately decoding movement intention from either plan or perimovement activity, but has assumed that temporal boundaries between these epochs are known to the decoding system. In this work, we develop a technique to automatically differentiate between baseline, plan, and perimovement epochs of neural activity. Specifically, we use a generative model of neural activity to capture how neural activity varies between these three epochs. Our approach is based on a hidden Markov model (HMM), in which the latent variable (state) corresponds to the epoch of neural activity, coupled with a state-dependent Poisson firing model. Using an HMM, we demonstrate that the time of transition from baseline to plan epochs, a transition in neural activity that is not accompanied by any external behavior changes, can be detected using a threshold on the a posteriori HMM state probabilities. Following detection of the plan epoch, we show that the intended target of a center-out movement can be detected about as accurately as that by a maximum-likelihood estimator using a window of known plan activity. In addition, we demonstrate that our HMM can detect transitions in neural activity corresponding to targets not found in training data. Thus the HMM technique for automatically detecting transitions between epochs of neural activity enables prosthetic interfaces that can operate autonomously.

Signal Processing Challenges for Neural Prostheses

Cortically controlled prostheses are able to translate neural activity from the cerebral cortex into control signals for guiding computer cursors or prosthetic limbs. While both noninvasive and invasive electrode techniques can be used to measure neural activity, the latter promises considerably higher levels of performance and therefore functionality to patients. The process of translating analog voltages recorded at the electrode tip into control signals for the prosthesis requires sophisticated signal acquisition and processing techniques. In this article we briefly review the current state-of-the-art in invasive, electrode-based neural prosthetic systems, with particular attention to the advanced signal processing algorithms that enable that performance. Improving prosthetic performance is only part of the challenge, however. A clinically viable prosthetic system will need to be more robust and autonomous and, unlike existing approaches that depend on multiple computers and specialized recording units, must be implemented in a compact, implantable prosthetic processor (IPP). In this article we summarize recent results which indicate that state-of-the-art prosthetic systems can be implemented in an IPP using current semiconductor technology, and the challenges that face signal processing engineers in improving prosthetic performance, autonomy and robustness within the restrictive constraints of the IPP.

Neural stimulation and recording with bidirectional, soft carbon nanotube fiber microelectrodes.

The development of microelectrodes capable of safely stimulating and recording neural activity is a critical step in the design of many prosthetic devices, brain-machine interfaces, and therapies for neurologic or nervous-system-mediated disorders. Metal electrodes are inadequate prospects for the miniaturization needed to attain neuronal-scale stimulation and recording because of their poor electrochemical properties, high stiffness, and propensity to fail due to bending fatigue. Here we demonstrate neural recording and stimulation using carbon nanotube (CNT) fiber electrodes. In vitro characterization shows that the tissue contact impedance of CNT fibers is remarkably lower than that of state-of-the-art metal electrodes, making them suitable for recording single-neuron activity without additional surface treatments. In vivo chronic studies in parkinsonian rodents show that CNT fiber microelectrodes stimulate neurons as effectively as metal electrodes with 10 times larger surface area, while eliciting a significantly reduced inflammatory response. The same CNT fiber microelectrodes can record neural activity for weeks, paving the way for the development of novel multifunctional and dynamic neural interfaces with long-term stability.

Rapid and Continuous Modulation of Hippocampal Network State during Exploration of New Places

Hippocampal information processing is often described as two-state, with a place cell state during movement and a reactivation state during stillness. Relatively little is known about how the network transitions between these different patterns of activity during exploration. Here we show that hippocampal network changes quickly and continuously as animals explore and become familiar with initially novel places. We measured the relationship between moment-by-moment changes in behavior and information flow through hippocampal output area CA1 in rats. We examined local field potential (LFP) patterns, evoked potentials and ensemble spiking and found evidence suggestive of a smooth transition from strong CA3 drive of CA1 activity at low speeds to entorhinal cortical drive of CA1 activity at higher speeds. These changes occurred with changes in behavior on a timescale of less than a second, suggesting a continuous modulation of information processing in the hippocampal circuit as a function of behavioral state.

Neurons to Silicon: Implantable Prosthesis Processor

A processor architecture for neural prosthesis control is described. It implements real-time neural decoding from a permanently implanted electrode array to reduce the data rate from 80Mb/s to 20b/s, minimizing the wireless communication requirements. The neural signals are digitized by a 100-channel 100kS/s adaptive-resolution ADC array consuming 1muW per channel

Hippocampal awake replay in fear memory retrieval

Hippocampal place cells are key to episodic memories. How these cells participate in memory retrieval remains unclear. After rats acquired a fear memory by receiving mild footshocks in a shock zone on a track, we analyzed place cells when the animals were placed on the track again and displayed an apparent memory retrieval behavior: avoidance of the shock zone. We found that place cells representing the shock zone were reactivated, despite the fact that the animals did not enter the shock zone. This reactivation occurred in ripple-associated awake replay of place cell sequences encoding the paths from the animals current positions to the shock zone but not in place cell sequences within individual cycles of theta oscillation. The result reveals a specific place-cell pattern underlying inhibitory avoidance behavior and provides strong evidence for the involvement of awake replay in fear memory retrieval.

Optimal estimation of feed-forward-controlled linear systems

The neuroprosthetic interface must infer an intended movement from the neural activity that would accompany it in healthy individuals. We show that an optimal estimator for a controlled system such as that responsible for human movements jointly estimates the goal and the trajectory of point-to-point movements. We demonstrate that this paradigm can achieve orders of magnitude of increased accuracy in regimes in which the interface has low SNR. With high SNR, our technique proves reliably more accurate than a typical approach which ignores the controlled nature of the system under observation. Furthermore, we show that even when the system violates the model assumptions of feedforward linear control with additive noise, system performance remains appreciably better than the alternative.

Model-based decoding of reaching movements for prosthetic systems

Model-based decoding of neural activity for neuroprosthetic systems has been shown, in simulation, to provide significant gain over traditional linear filter approaches. We tested the model-based decoding approach with real neural and behavioral data and found a 18% reduction in trajectory reconstruction error compared with a linear filter. This corresponds to a 40% reduction in the number of neurons required for equivalent performance. The model-based approach further permits the combination of target-tuned plan activity with movement activity. The addition of plan activity reduced reconstruction error by 23% relative to the linear filter, corresponding to 55% reduction in the number of neurons required. Taken together, these results indicate that a decoding algorithm employing a prior model of reaching kinematics can substantially improve trajectory estimates, thereby improving prosthetic system performance.