• Beckman Institute

Latest Projects

Toward Simultaneous Recording of Electroencephalography and Functional Magnetic Resonance Imaging

Matthew Moore & Andrew Groll

As separate modalities, electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) provide critical means for insight into the human brain, but each method has a relative strength where the other has a relative weakness (i.e. temporal vs. spatial resolution). In this project, preliminary steps in successfully acquiring and processing simultaneous EEG-fMRI data were completed. A preselected EEG and fMRI compatible oddball task was implemented in each modality independently as well as simultaneously. Results indicate similar expected event-related potential (ERP) waveforms from the EEG recordings and expected blood oxygen level dependent (BOLD) signal changes from the fMRI recordings in the separate and simultaneous sessions. These results support the feasibility of implementing newly acquired multi-modal imaging equipment in future research and point to areas where data processing and artifact correction inherent to multi-modal imaging might be further adjusted and refined for consistent quality.

Validation of a Simulation Tool for Asynchronous Cortical Streams (STACS) using Cortical Cultures on Multi-Electrode Arrays

By Felix Wang & Amogh Belagodu

Learning is an inherently closed-loop process that involves the interaction between an intelligent agent with its environment. The underlying drive of this work is to provide a path toward a more complete understanding of the adaptive learning process that is observed in networks of spiking neurons. Multi-electrode arrays (MEAs) allow us to address this closed-loop system for learning. These devices allow for recording and stimulation over a network for an extended period of time, providing the ability to monitor the evolution of a network. Extending this to simulation allows for more in depth analysis than would be possible with the constraints of a live culture. However, as invaluable of a tool as simulators are in studying neural networks, they are predicated on their accuracy to the underlying physiology.

Thus, a crucial step in moving forward in this direction is the ability of a simulation tool to adequately capture the relevant functions of the biological system. We evaluate a Simulation Tool for Asynchronous Cortical Streams (STACS) by drawing comparison to the behavior of cortical cultures plated on multi-electrode arrays. In particular, we focus on network metrics that are invariant to the specific connections but rather based off of the general organizing principles that form the network topology. Most importantly, we provide a framework for embodiment and feedback through interfacing with the external environment similar to that provided through MEAs. By drawing comparison to in vitro cortical cultures, the aim is to validate the accuracy of the simulation tool in modeling evolving networks.

Because the focus of our study is learning, we chose the following metrics to examine between the cell cultures on the MEA and the generated network in STACS. Efficiency of spike propagation corresponds to the facilitation of communication between neurons that respond to a particular stimulus. That is, while there is an innate level of communication and chatter between neurons, this is increased when a stimulus is applied and the pathways corresponding to the stimulus is strengthened. Center of activity is used to indicate the localization of firing. For a homogeneous and randomly firing network, the center of activity will be roughly at the center of the array. However, as stimulus is applied to particular regions, the firing of the surrounding cells will occur at a greater and faster rate than the rest of the array, and so the center of activity will shift towards that region, indicative of learning by the network with respect to the stimulus. Polychronization in a network results from the interplay among inhibition, axonal delays, and synaptic plasticity. It exists in the form of stable polychronous neural groups that exhibit reproducible, time-locked patterns of spiking with millisecond precision. At a high level, polychronous firing shares similarity with the idea of a cell assembly, a network of neurons that tend toward spiking as a collective due to its strong synaptic connections. Similar to the firing response of a neuron to stimuli, polychronous firing can be thought of as the firing response of a neural network to stimuli.

For the validation process, we constructed an experimental pipeline to benchmark the agreement of neural simulation to the biological system using network level metrics above. During the MEA experimentation, certain parameters to the simulation may be adjusted, such as axonal delay or plasticity time constants. Others, such as the network composition of excitatory and inhibitory cells, must wait until the terminal immunohistochemistry of the cell culture. For our experiments, cultures were plated on the MEAs from dissociated cortical cells of embryonic rats, and stimulation and recording from the MEAs were performed through the NeuraLynx system. A similar approach to stimulation and recording was mirrored in the STACS simulation tool through a communication protocol, YARP. Preliminary experimental results for a simple application of stimuli and an initial set of parameters indicated that greater plasticity to stimuli in the simulated network was needed. In addition to decreasing spiking rates, this was noted through the contrast of the simulated network with the biological one with respect to the general direction of movement of its center of activity. Also brought into consideration through the experiments was an area of exploration to incorporate models to account for baseline activity in the network regardless of stimuli.