Abstract
Purpose: :
Subretinal microphotodiode-arrays (MPDAs) are designed to stimulate the retinal network in patients suffering from Retinitis Pigmentosa (RP) by converting visual information into topically ordered patterns of electrical stimulation. In order to estimate and visualize the most probable perceptions of patients carrying subretinal implants, a modeling approach has been chosen.
Methods: :
A previous model (Stett, Mathmood 03), describing light dependent activation of stimulation electrodes of the MPDA and the spatial activity distribution (Gaussian function, σ) in the retinal network elicited by single electrodes, was extended to incorporate temporal features of RGC activity following subretinal stimulation. Frequency dependency of RGC accommodation upon repetitive stimulation (Jensen, JNE 07) was accounted for. Random changes in charge injection (variable effectiveness of electrodes) and in retinal current spread (inhomogeneous tissue conductibility) were integrated into the model. For this purpose model parameters were multiplied with random numbers derived from a normal distribution specified prior to each simulation. The model was used to simulate processing of moving scenes by the MPDA and delivered respective RGC output in dependence of the stimulation frequency in real-time.
Results: :
Visual perceptions could be modeled based on experimental findings and patient reports. Reconstruction of a patient report on a phosphene extend yielded σ = 35% of electrode size. Identification of velocity of moving objects could be perceived if electrodes were activated with frequencies around 10Hz, assuming an input frame rate of 10 to 20 fps. Despite random fluctuations up to 30% (for charge injection and retinal activity spreading) identification of simple moving objects remained accurate. However, due to retinal network adaptation, high-frequency stimulation resulted in quickly fading visual perceptions. At 10Hz-stimulation frequency no modeled visual perception remained after 4s, which is in principal agreement with findings from the clinical pilot study.
Conclusions: :
Moving scenes displayed by the software facilitated formulation of expectations towards actual stimulation of the retinal network by subretinally implanted MPDAs. The model was able to produce valuable and quantifiable representations of temporal features of visual perceptions evoked in blind patients carrying subretinal MPDAs. New stimulation modes allowing visual percepts to become temporally more stable are required however.
Keywords: retinal degenerations: hereditary • computational modeling • perception