8 118 8. General discussion it could be advantageous to make use of readily available information and informed models to confine the search space. For instance, it could be relevant to combine features of available image processing models or to optimize only a subset of image processing parameters (seeChapter3). Choosing the right level of complexity in the optimization framework requires an informed balance between complexity and control. Optimizing specific hardware parameters In addition to evaluating a confined set of image processing parameters, in silico frameworks can be useful to optimize specific hardware parameters. Besides the number of electrodes, which were computationally evaluated inChapters3and4, related research has studied the optimization of, for instance, electrode locations (Bruce & Beyeler, 2022). Recent work demonstrated that even the phosphene simulation itself can be computationally optimized using human feedback, to better resemble patient-specific characteristics (Granley et al., 2023). 8.3. The biology of visual neurostimulation Biological models of visual neurostimulation As the development of visual prosthetics matures towards a clinical testing phase, the incorporation of realistic models of experimental data can become more vital (seeChapters5and6). Many patient-specific design questions can only be answered with realistic phosphene simulations (Beyeler et al., 2019; Caspi et al., 2018; Fine & Boynton, 2023; Granley et al., 2022b; Paraskevoudi & Pezaris, 2021) that model findings from clinical experiments (Armenta Salas et al., 2022; Beyeler et al., 2019; Bosking et al., 2017b; Winawer & Parvizi, 2016). The work in this dissertation contributes to closing the gap between abstract simulations and the clinical reality, exploring the advantages of simulating at different levels of biological detail. Detailed models can address case-specific scenarios To understand the benefits of detailed simulations, it is important to recognize that the factors software design, hardware design, patient characteristics, and functional context are highly interrelated (see Chapter 2andChapter 4). The functional outcome is the result of a multitude of parameters with potential interactions. A more detailed simulation can help to capture the complete functional result. Where more general questions can be addressed with abstract or qualitative simulations (e.g., Sanchez-Garcia et al., 2020; Vergnieux et al., 2017), more detailed simulations can evaluate specific stimulation requirements and even patient-specific characteristics (seeChapter6), Granley et al., 2023). Including software parameters, hardware parameters and patient-aware contextual parameters in the optimization can help to bring the value of simulation research closer to individual patient requirements. Biological plausibility versus phenomenological accuracy Unlike retinal stimulation, which is understood in more detail (Beyeler et al., 2019), cortical prosthetic vision is still characterized by many unknowns and irregular findings (Bosking et al., 2018; Fernández et al., 2021). On the one hand it is of vital importance to improve our biological understanding of the observed irregularities in cortical phosphene
RkJQdWJsaXNoZXIy MTk4NDMw