6 108 6. Gaze-contingent processing improves mobility performance environment. The use of finger-steered movement with a sensitive trackpad may have led to difficulties interpreting the virtual self-motion. This holds especially for phosphene vision that lacks natural optical flow cues. One limitation of the scene recognition task, was the limited generalizability of the 3D environments. Although there were large differences in the used scenes in terms of design and objects, the used environments reflect only a subset of the diversity (e.g. cultural diversity) of real-world environments. Due to the fundamental low-level visual differences between the experimental conditions, it is unlikely that the results are much influenced by the specific choice of environments. There are several (partly inherent) limitations of simulated prosthetic vision with sighted subjects. Our VR simulation study differs inherently from reality. Collision detection was based on the head position and no physical interactions with the environment were present. Although efforts were undertaken to maximize the realism of the phosphene simulation, it nevertheless misses some of the true complexities of artificial phosphene vision, such as temporal dynamics, irregular shapes, and interactions between electrodes. Note that the functional quality of the implemented phosphene simulation is optimistic compared to that of existing contemporary prostheses (e.g., seeFernández et al., 2021). In reality, the creation of many phosphenes with a wide field of view will be associated with both surgical and technical challenges that remain to be addressed. Furthermore, blind individuals are likely to use different strategies than sighted study subjects (especially for mobility). Although we aimed to measure visually-guided mobility, the prospective users of visual prostheses may rely on multi-modal sensory information (e.g. using a cane). Lastly, and related, although the current study implemented an immersive simulation with complex 3D tasks, it is important to acknowledge that nevertheless there is a gap with the actual clinical situation. the tasks that were tested are still relatively controlled compared to the real life situation, where orientation and mobility is usually a closed-loop interaction with the environment. With the progression of prosthetic technology it will become possible to perform clinical experiments with more complex visually-guided activities of daily living. In general, further clinical work can improve the realism and relevance of simulation studies. Vice versa, simulation studies can give further directions for progressing the technology towards the clinical setting. 6.5. Conclusion In this mixed reality simulated prosthetic vision study, we evaluated the benefits of eye tracking-based gaze compensation for head-steered prostheses in the context of mobility and orientation. We found that gaze-contingent processing improved the performance in all experimental tasks. Likely, this improvement is mostly owed to compensation for spatial updating conflicts and only to a lesser extent to extended freedom of visual scanning. We found that neglecting the effects of eye-movements in simulations of headsteered prostheses can yield overoptimistic results. This should be taken into account in future simulation work. Taken together, we conclude that for cases in which eye-tracking is feasible, gaze-contingent image processing is expected to improve the functional and subjective quality of head-steered prostheses.
RkJQdWJsaXNoZXIy MTk4NDMw