This thesis investigates the design challenges and underlying mechanisms of spatially selective vagus nerve stimulation using temporal interference stimulation (TIS). While TIS shows promise for non-invasive and spatially selective neuromodulation, a limited understanding of the mechanistic underpinnings of neural stimulation enabled by high-frequency interfering electric fields, and fundamentally larger power expenditure related to the generation of low frequency intermodulation products, limits its applicability to specific cases where the featured spatial selectivity is an absolute necessity. 

The objective of this research is to develop an electrophysiological model to accurately represent experimental measurements of nonlinear neuron responses under influence of TIS, and to develop a system architecture and output stage design with power supply regulation for a power-efficient and spatially selective temporal interference-based vagus nerve stimulator. 

The proposed model generates an output waveform by means of nonlinear approximation of electric field interference, up to a specific order of a Taylor expansion. The model is shown to accurately predict intermodulation products present in experimental data of TIS for various input frequencies. A way to define the relevance of waveform shape on the stimulation efficiency of temporal interference stimulation is also hypothesized, which shows that the efficiency of both sinusoidal (at even orders of the Taylor expansion) and square waveforms is dependent on input Taylor expansion order, approaching 10% efficiency when the order of the expansion is high. Limitations of the model include the lack of experimental data to validate the model outputs for other waveforms besides sinusoidal ones, and simplified assumptions made in the determination of Taylor expansion coefficients, which may limit the applicability of a generalized version of the model.

The principles behind temporal interference-based stimulators and the effect of power supply regulation on power-efficiency are discussed, leading to a full system architecture for the power-efficient implementation of TIS-based stimulators. The presented indirect feedback output stage design, with power supply regulation, features over 3σ-accuracy to the required load current, and consistently outperforms conventional cascode current mirror approaches in simulations and extensive testing of its performance as a TIS output stage. The discussion on the presented results feature some important points on the compliance of the posited output stage, most prominent of which is its reliance on large source impedances, which a current-mode stimulator should rather do without.

Future research is encouraged to measure more varied input waveform shapes, and determine proper values for the Taylor expansion coefficients of the posited electrophysiological model. The model should also be further expanded into a spatially distributed version, which takes into account the spatial component of electric field propagation, and the consequential variations in modulation depth of the stimulation waveform. Additionally, the presented output stage requires further validation for different process corners, at higher operational temperatures, and some of the early design assumptions may need to be re-evaluated, particularly the choice of using indirect feedback. These steps will contribute to a viable, spatially selective and power-efficient output stage for temporal interference-based stimulators, to provide patients with better quality of life in the future.