Abstract:
Electrical impedance tomography (EIT) is an imaging modality which is sensitive to neural activity and can resolve multiple concurrent changes in impedance in a sample, two attributes which make it a promising candidate for investigation in peripheral nerve interface applications. In nerve cuffs for control of advanced robotic prostheses, replicating the complex dexterity of humans requires real time determination of multiple concurrent sites of neural activity in the peripheral nerves of the arm. The goals of this thesis address two areas associated with real time EIT in nerve cuffs: (i) improving the signal to noise ratio, and (ii) frequency multiplexing of EIT drive currents. In Chapters 1 and 2, an introduction to neural prosthetics and literature review of electrical impedance tomography of neural activity, respectively, are presented, followed by a short overview of the thesis in Chapter 3. In Chapters 4 – 6, a finite element (FE) model was used to generate data for the forward and inverse problems to evaluate feasibility of the application and determine experiment parameters. Chapters 7 – 9 present three in vitro studies on sciatic nerve of rat. In Chapter 7, the frequency roll off of the baseline impedance is measured and used to estimate the operating frequency range for frequency division multiplexing (FDM). In Chapter 8, the signal processing steps in FDM were analysed. In Chapter 9, a novel modification to the current source was investigated which improves the signal to noise ratio and potentially extends the operating frequency range. Modelling of nerve fibers under longitudinal and transverse currents indicated a highly anisotropic impedance that produced a tradeoff between the fraction change in impedance during neural activity and the start of the frequency roll-off from capacitive charge transfer. When this impedance anisotropy was input into a finite element model of EIT in a nerve cuff, longitudinal currents produced a larger signal to error ratio but poorer spatial resolution than transverse currents; although when the current-limits of neural tissue were accounted for (longitudinal: 30 μA; transverse: 150 μA), the signal to error ratio was broadly comparable for both currents. In vitro experiments on cadavers, with stimulation of the paw and EIT recording in the sciatic nerve, indicated the number of parallel drive currents in a frequency division multiplexed system is limited by a narrow operating frequency range and relatively broad frequency spectra of neural activity, which, in turn, limits the achievable spatial resolution. The results indicate further improvements in signal to noise ratio and additional methods of multiplexing drive currents are needed to realize EIT in a nerve cuff for neural prosthetics.