Optimization and Design Strategies for Wireless Battery Free Implantable Electronics Using Deep-Learning-Based Markerless Tracking

Abstract

Wireless, battery-free neural interfaces enable the removal of large, bulky batteries and tethers required for conventional neural interfaces, which cause movement artifacts and impact the behavior of animal models. Allowing fully implanted neural and organ interfaces enables advanced insight into social interactions and expands studies towards subjects that move in 3D, such as birds. The free motion of subjects places an emphasis on mechanical, electromagnetic, and optical design approaches that accommodate operation in highly mobile areas and reduction of device impact on the motion of the subject. To enable designs highly tailored to the animal model, this thesis explores the utilization of Deep Neural Network (DNN) based markerless tracking of animals to inform designs of electromagnetics and mechanics design of wireless, battery-free electronics.To enable a high level of miniaturization, wireless and battery-free device architectures design of communication and energy harvesting antennas is critical to realize miniaturization, operational range, and data rates to allow for multimodal operation in subjects that move in 3D. This thesis explores informing the primary antenna design by extracting the animal's behavior using DNN tracking to shape intensity profiles optimized towards maximum energy transfer to miniaturized implants. Conversely, the thesis will also explore DNN tracking for the assessment of neural interface’s behavioral impact on the animal models to inform mechanical designs of subcutaneous structures to enable seamless organ interfaces such as the musculoskeletal system.