5.5. Conclusion 5 79 of the perceptual experience of an implant user, and in reality the distinction may be less strict. The conscious perception of phosphenes requires considerable training and the detection process is influenced by attention (Fernández et al., 2021). Although our implementation is effective for modeling the psychometric data, alternative implementations could also be considered. The perceptual effect of different simulated phosphene threshold implementations for sighted subjects remains to be evaluated in future SPV work. Secondly, the leaky integrator and the memory trace that are implemented in our simulator might be an oversimplified model of tissue activation in the visual cortex and some non-linear dynamics might be missed. For instance, all data used in this study to fit and validate the model used symmetric, biphasic pulse trains, while other pulse shapes might lead to different neural or behavioural responses (Merrill et al., 2005). Also, several studies reported that higher stimulation amplitudes may give rise to double phosphenes (Brindley & Lewin, 1968; Dobelle & Mladejovsky, 1974; Oswalt et al., 2021; Schmidt et al., 1996), or a reduction of phosphene brightness (Schmidt et al., 1996). Furthermore, in contrast to the assumptions of our model, interactions between simultaneous stimulation of multiple electrodes can have an effect on the phosphene size and sometimes lead to unexpected percepts (Bak et al., 1990; Dobelle & Mladejovsky, 1974; Fernández et al., 2021). Although our software supports basic exploratory experimentation of non-linear interactions (seeFigureS3), by default, our simulator assumes independence between electrodes. Multi-phosphene percepts are modeled using linear summation of the independent percepts. These assumptions seem to hold for intracortical electrodes separated by more than 1 mm (Ghose & Maunsell, 2012), but may underestimate the complexities observed when electrodes are nearer. Further clinical and theoretical modeling work could help to improve our understanding of these non-linear dynamics. A third limitation is that our simulator currently only models responses of V1 stimulation. Future studies could explore the possible extension of modeling micro-stimulation of the LGN (Pezaris & Reid, 2007) and higher visual areas, such as V2, V3, V4 or inferotemporal cortex (IT). In previous NHP research, reliable phosphene thresholds could be obtained with the stimulation of the LGN, V1, V2, V3A and middle temporal visual area (MT) (Murphey & Maunsell, 2007; Pezaris & Reid, 2007). Furthermore, IT stimulation has shown to bias face perception (Afraz et al., 2006). Similar effects have been confirmed in human subjects, and previous work has demonstrated that electrical stimulation of higher order visual areas can elicit a range of feature-specific percepts (Lee et al., 2000; Murphey et al., 2009; Schalk et al., 2017). Our simulator could be extended with maps of other visual areas with clear retinotopy, and an interesting direction for future research will be the implementation of feature-specific percepts, including texture, shape and colour. 5.5. Conclusion We present a framework for the biologically plausible simulation of phosphene vision. The simulator models psychophysical and neurophysiological findings in a wide array of experimental results. Its phenomenologically accurate simulations allow for the optimisation of visual cortical prostheses in a manner that drastically narrows the gap between simulation and reality, compared to previous studies of simulated phosphene vision. It can operate in real time, therefore being a viable option for behavioural experiments with sighted volunteers. Additionally, owing to the PyTorch implementation and the
RkJQdWJsaXNoZXIy MTk4NDMw