Research Seminar - April 3, 2001

Abstract

This discussion will consider how electrical nonsynaptic communication can influence aspects of neural processing. The research presented proposes that the extracellular space (ECS) surrounding a collection of neurons maintains an electrical potential resulting from currents produced by the axons of these neurons. For the exploration of electrical nonsynaptic communication and its functionality, an artificial neural network (ANN), which is inspired by the fly's early visual processing system, was developed.

As with the fly's lamina, the ANN is a hexagonal array of artificial synaptic cartridges (ALCs). Each ALC maintains an artificial extracellular space (AECS) potential produced by currents from the stimulus-receiving artificial neurons, called artificial photoreceptors (ARs). The ANN applies diffusive mechanisms to model electrical nonsynaptic communication. Aspects of electrical nonsynaptic communication included diffusive interactions between the ARs and their respective AECS as well as between adjacent ALCs. It will be shown that this arrangement of nonsynaptic channels produces an intercellular route linking ALCs. Consequently, the AR potential in one ALC can affect the AECS potential in adjacent ALCs.

As each synaptic cartridge in the fly's lamina contains two large, second-in-order processing neurons, called large monopolar cells (LMCs), the ANN also includes two artificial large monopolar cells (ALMCs). The ALMC receive inputs from six ARs, each with a distinct visual field. The consideration of how the AECS potential may affect the input into each ALMC, led to potential differencing input. Thus, ALMC inputs rely on AR-AECS potential differences. Results presented will demonstrate that potential differencing reduces the background potential present in the input potentials. It will also be demonstrated that potential input differencing results in inputs with on-center off-surround characteristics.

Although the ANN is a simplification of the fly's lamina, this research will also demonstrate that biological mechanisms can be investigated using simpler computational models. To investigate nonsynaptic mechanisms possibly occurring within the fly's early visual processing, a computer model for the ANN was developed, tested, and analyzed. The ANN is computationally expensive. To improve the performance of the computer program, parallel programming methods were applied to the computer model. Parallel programming methods and performance measurements will be discussed.