Neural spike classification acceleration with RRAM technologies

Abstract

Spike classification is an area of critical importance in neurological medicine. The behaviour of neurons is critical in diagnosis of disease, understanding neural structures, and operating prostheses. The probe technology used to gather in situ data from neurons has seen significant advances in the past decade, but the technology required to process this vast amount of data lags behind and this thesis aims to address the data processing aspect of this issue. A novel analogue circuit that conducts most of the sorting in the pre-processing stage is presented, demonstrating the feasibility of such a system using memristive devices. This thesis covers the progress made during this research project; the development of an new instrument for the testing of memristors and memristor related circuits, the design of a new analogue cell for use in template matching systems, and the testing and simulation of this cell both in isolation and in a simple template matching system. The circuit developed demonstrated comparable energy dissipation to current state of the art spike sorting systems, without the need to digitise the signals being processed. This development opens the path to the fabrication of an integrated memristor based spike sorting system suitable for neural signal processing.

More information

Identifiers

ORCID for Themis Prodromakis: orcid.org/0000-0002-6267-6909

Catalogue record

Export record