Optimal placement of high-channel visual prostheses in human retinotopic visual cortex

Recent strides in neurotechnology show potential to restore vision in individuals afflicted with blindness due to early visual pathway damage. As neuroprostheses mature and become available to a larger population, manual placement and evaluation of electrode designs becomes costly and impractical. An automatic method to optimize the implantation process of electrode arrays at large-scale is currently lacking.Approach:Here, we present a comprehensive method to automatically optimize electrode placement for visual prostheses, with the objective of matching pre-defined phosphene distributions. Our approach makes use of retinotopic predictions combined with individual anatomy data to minimize discrepancies between simulated and target phosphene patterns. While demonstrated with a 1000-channel 3D electrode array in V1, our pipeline is versatile, potentially accommodating any electrode design and allowing for design evaluation.Main results:Notably, our results show that individually optimized placements in 362 brain hemispheres outperform average brain solutions, underscoring the significance of anatomical specificity. We further show how virtual implantation of multiple individual brains highlights the challenges of achieving full visual field coverage owing to single electrode constraints, which may be overcome by introducing multiple arrays of electrodes. Including additional surgical considerations, such as intracranial vasculature, in future iterations could refine the optimization process.Significance:Our open-source software streamlines the refinement of surgical procedures and facilitates simulation studies, offering a realistic exploration of electrode configuration possibilities.