7 113 Chapter 6 further underscores the relevance of accounting for human biology in the optimization of visual prostheses, focusing on the effects of eye movements. In contrast to natural vision, which is steered with gaze, most prosthetic designs use a head mounted camera. Phosphenes - similar to ‘eye floaters’ - are locked to eye-movements, causing unstable and disorienting vision and prior research has suggested a possible role of eye-tracking for compensation. In two separate VR experiments with a total of 42 sighted participants the effects of gaze locked vision are compared to a gaze-contingent compensation condition with eye-tracking, and a control condition with head-steered vision where gaze effects are ignored. Simulated gaze-contingent phosphene vision was found to improve mobility and orientation performance compared to gaze-locked vision. Furthermore, both the gaze-locked and gaze-contingent study conditions were outperformed by the control condition with ignored gaze. The results of this study indicate that gaze-locked vision in head-steered prostheses can be debilitating for complex visuallyguided activities of daily living such as mobility and orientation. Ignoring these gaze effects in the phosphene simulation can produce overoptimistic results in SPV studies. The inclusion of a compensatory eye-tracking system is strongly endorsed for prospective users of head-steered prostheses with an unimpaired oculomotor system. Conclusion The studies in this dissertation demonstrate that digital simulations can be used for the accelerated evaluation and optimization of visual prosthetics. The presented empirical results provide a nuanced view on several hardware and software requirements, such as the number of implanted electrodes, the inclusion of an eye-tracking system and the choice of scene simplification. In addition, several methodological directions are provided to evaluate the multitude of (interacting) design factors. Functional prototyping can help to evaluate specific designs in naturalistic tasks such as mobility. Through evaluation with virtual patients, the quality of prosthetic vision can be more automatically optimized, making use of deep-learning-based parameter optimization. Biologically plausible simulations can help to close the gap between abstract simulations and clinical reality. The results in this dissertation contribute to the interdisciplinary research field of neuroprosthetics, augmenting the existing clinical experimentation with a more computational focus. The results in this dissertation underscore a significant role of digital simulations for the non-invasive early-stage prototyping and optimization of visual neuroprosthesis.
RkJQdWJsaXNoZXIy MTk4NDMw