Society for Neuroscience Annual Meeting 2013, November 9-13, San Diego

A wireless ultra-low power high-density implantable ecog

One of 4 contributions at SFN2013 by the UC Berkeley group of Carmena, Maharbiz and Rabaey

P. LEDOCHOWITSCH, R. MULLER, W. LI, H.-P. LE, S. GAMBINI, A. KORALEK, J.M. CARMENA, M.M. MAHARBIZ, J. RABAEY

In this work, we present an implantable, fully wireless μECoG system for neural recordings and brain mapping in-vivo. The implant combines a low-power wireless IC [1] with a custom flexible antenna [2] and a microfabricated, high-density micro electrode array [3] into a tiny package that is mechanically flexible enough to conform to the curvature of the cortex. This device can be implanted through a small burr hole rather than a full craniotomy, and its wireless capability allows the surgical sight to be sealed completely, reducing the risk of infection and enabling experimentation in truly free, unencumbered and awake behaving subjects. We tightly integrate a high-density flexible MEMS electrode array with active circuits and a power-receiving antenna to realize a fully implantable system in an extremely small (mm-scale) footprint. The electrode array is based on a thin Parylene C substrate with 64 recording sites of sub-mm electrode pitch. The antenna is patterned in the same MEMS fabrication process as the flexible array, allowing greater antenna area, and therefore greater received power, without the expense of implanting a large rigid structure. The only rigid component is the integrated circuit measuring 2.4 mm x 2.4 mm. Careful attention is paid to circuit area reduction enabling a scalable platform for 64-channel recording and beyond. The circuit is fully integrated with 64 neural recording front-ends together with wireless power coupling and wireless data transmission circuitry for communication across the skull. We have assembled a prototype and present measured system performance characteristics as well as in-vivo measurements of surface cortical potentials.

[1] R. Muller, S. Gambini, and J. Rabaey, ISSCC, 2011.

[2] T. Bjorninen, R. Muller, P. Ledochowitsch, L. Sydanheimo, L. Ukkonen, M. M. Maharbiz, and J. M. Rabaey, IEEE Antennas and Wireless Propagation Letters, 2012.

[3] P. Ledochowitsch, R. J. Felus, R. R. Gibboni, A. Miyakawa, S. Bao, and M. M. Maharbiz, in IEEE MEMS, 2011.

Remind: Toward a remind prosthesis for the human hippocampus

One of 15 contributions at SFN2013 by the group of Dr. Theodore Berger

M.-C. HSIAO, P.-N. YU, D. SONG, C.Y. LIU, C.N. HECK, D. MILLETT, T.W. BERGER

The multi-input, multi-output (MIMO) model development, for both the stationary case and the non-stationary case, will constitute the core of a hippocampal prosthesis for patients with substantial amnesia due to epilepsy, stroke, dementia, or head trauma. In order to extend our experimental model beyond the non-human primate, we have engaged in a collaboration between the Departments of Biomedical Engineering, Neurosurgery, and Neurology. This collaborative effort provided a unique opportunity (i.e., mesial temporal lobe epileptic patients who went through temporal lobectomy) to study viable human hippocampal tissue in vitro (IRB approved #HS-10-00162). This allowed us to become more familiar with the spatial geometry and the electrophysiological properties of the human hippocampus. We have been able to record neuronal responses evoked by electrical stimulation in human hippocampal slices. Using multi-electrode array (MEA) technology, neural activity from different regions have been captured simultaneously. Detailed methodology and summary of the initial results are presented. This study will facilitate the future design of implantable electrodes and effective electrical stimulation patterns that will enable application of our MIMO modeling approach to human subjects.

REMIND: Volterra estimation of a spike-timing-dependent plasticity learning rule from spiking data

One of 15 contributions at SFN2013 by the group of Dr. Theodore Berger

B. ROBINSON, D. SONG, T.W. BERGER

The goal of a cognitive prosthesis is to re-instantiate functional input-output spatiotemporal spiking transformations between neural regions. In the hippocampus, this input-output relationship will likely change over time due to activity-dependent synaptic plasticities such as long-term potentiation/depression (LTP/LTD). In this study, we develop a mathematical framework to characterize an activity-dependent plasticity learning rule from spiking data for eventual application to the hippocampal cognitive prosthesis. We first simulate the spiking data from a single-input, single-output neuron whose synaptic strength changes over time following spike-timing-dependent plasticity (STDP). The model input is a 5Hz Poisson random spike train. Each input spike increases the output firing probability through a feedforward first-order Volterra kernel, which can be interpreted as the excitatory postsynaptic potential (EPSP). The amplitude of the feedforward Volterra kernel changes over time following an STDP rule that is expanded with a series of Volterra kernels. These plasticity Volterra kernels describe how the EPSPs change as a function of the input-output inter-spike intervals and the plasticity induction time course. Saturation of plasticity and its consequences on the model estimation are also considered. Simulation results show that, using maximum-likelihood estimation and Laguerre expansion, plasticity Volterra kernels can be estimated accurately from the input-output spiking data alone (i.e., measurement to the membrane potential is not required). In the future, this mathematical framework will be extended to multiple-input multiple-output systems for the application to spiking data recorded from behaving animals. Results of such studies could yield insights into the nature of long-term synaptic plasticity and would be essential for the incorporation of plasticity into the hippocampal prosthesis.

From connectome to model parameters: Combining measurements of morphology and brain activity

In Nanosymposium "410.Computation, Modeling, and Simulation V"

R.A. KOENE

  Following impressive proof-of-principle results in neural circuit analysis using connectome data (Briggman et al., 2011; Bock et al., 2011), the next drive in data acquisition aims at large-scale high-resolution brain activity mapping, as outlined in goals of the BRAIN Initiative. High-resolution dynamic reference data, combined with knowledge about the architecture of biological neural circuits should enable us to create reliable and realistic functional models of those neural circuits.

  Here we investigate a protocol for model approximation from morphological data, plus parameter tuning with reference points providing sample activity data.

  The resolution of morphological data and the spatial resolution at which reference activity is recorded are not identical. We consider the trade-off between increasing the measurement resolution in a brain activity map or increasing the difficulty of parameter tuning.

SFN-WBE 2013 Informal gathering of the Whole Brain Emulation community and carboncopies.org at SFN2013. Let's efficiently and enjoyably exchange our insights from this year's SFN meeting that are directly related to Whole Brain Emulation.