Quantifying High Dimensional Recordings of Neural Surface Potentials

Thumbnail Image



Journal Title

Journal ISSN

Volume Title

Repository Usage Stats



Chronically reliable neural implants are needed to provide long lasting, high fidelity interfaces for clinical use and for basic science. Minimally invasive implants, such as micro-electrocorticographic arrays (µECoG), have been proposed in this field to minimize immune response and glial scarring reactions and promote long-term recording stability. µECoG electrodes record field potential from the surface of the pia or dura. Because of their low sensitivity to direct neuronal discharges, the neural "output" of µECoG electrodes is ambiguous. In addition, the correct sampling resolution for µECoG arrays is not precisely known, due to incomplete knowledge of the spatial correlation of surface potential. This dissertation proposes and validates µECoG characterization metrics to assist in comparing the quality of recording output between novel and reference devices, and to enable longitudinal quality tracking in chronic implants. Among this methodology is a model of spatial field variation that sheds new light on spatial bandwidth, and the prediction and denoising capabilities from correlated sampling.

In the first section of this dissertation, the rat auditory cortex model is introduced as the in vivo validation target for subsequent work. A comprehensive set of metrics, sensitive to spatial and temporal variation, are developed to summarize the background and evoked response µECoG signal. µECoG is validated as accurately reflecting auditory physiology in the acute setting, and is shown to record consistent signal in implants for 30-60 days. The second section applies these signal validation techniques to demonstrate the compatibility of neural recordings from two fundamentally different amplification technologies: a high impedance voltage amplifier and a transimpedance amplifier. The third section is a thorough investigation of recordings from µECoG arrays implanted chronically for over one year. The quantification methodologies are adapted to longitudinal quality tracking and are compared to the common status indicator of electrode impedance. The fourth section explores the problem of spatial sampling in µECoG by directly quantifying the efficiency of predicting unsampled field potential under various spatial bandwidth and noise conditions.

This set of results in acute and chronic settings provides a robust confirmation of some simple but reliable metrics of µECoG recordings. The use of these statistics in ongoing device validation work should provide nuanced and multivariate accounting of the strengths and weaknesses of neural interfaces. However, future work is required to improve sensitivity to finer scale spatio-temporal features of surface potential fields.





Trumpis, Michael (2018). Quantifying High Dimensional Recordings of Neural Surface Potentials. Dissertation, Duke University. Retrieved from https://hdl.handle.net/10161/17436.


Dukes student scholarship is made available to the public using a Creative Commons Attribution / Non-commercial / No derivative (CC-BY-NC-ND) license.