Visual prostheses aim to approximate biomimetic visual function by electrically simulating surviving retinal neurons. Improving the spatial resolution of electrically elicited artificial vision remains a critical challenge for retinal prostheses. We investigate how dynamic virtual channel (DVC) parameters shape retinal ganglion cell (RGC) population responses to improve spatial precision and activation efficiency in epiretinal stimulation. We developed a computational modeling framework to quantify DVC performance using a hierarchical optimization strategy. First, static virtual channels (SVCs) were used to map how current ratio (α) and stimulus intensity govern RGC activation, defining an optimal SVC parameter space. Building on this baseline, DVC protocols were refined by evaluating the combined effects of inter-virtual–channel interval (ΔT), α, and intensity. This strategy significantly reduces the complexity of DVC parameter optimization. Under SVC stimulation, increasing intensity improved the linearity of receptive field (RF) centroid displacement with α, while α and intensity jointly set RF centroid location and activated area. Under DVC stimulation, ΔT strongly modulated RGC activation, especially at short intervals. Initializing from SVC-optimized parameters, tuning ΔT and intensity produced more confined activation at lower stimulus intensities than SVC, indicating that DVC can serve as a novel stimulation strategy to enhance spatial precision and activation efficiency in retinal stimulation. This study provides the first systematic analysis of retinal DVC stimulation and a practical optimization framework for next-generation prostheses.

Temporal Interference Stimulation (TIS) is a low invasive technique for retinal modulation that utilizes the interference of high-frequency electric fields. Compared to conventional transcorneal electrical stimulation, TIS, by virtue of its enhanced depth penetration capabilities, enables selective modulation of deep-seated central retina, thereby establishing itself as a promising precise-therapeutic modality for diverse retinal pathologies. Clinically, retinal pathologies frequently coexist with concurrent lens pathologies. However, the functional efficacy of TIS in ocular systems with lens pathologies remains undocumented in current scientific literature. To explore intraocular electrical convergence under TIS, this study constructed multi-layered finite element models of normal aged and lens diseased human eyeball (NAE and LDE) with detailed anterior segment structures. The results firstly demonstrated a stable ground setting with compromising computational efficiency in NAE model. Subsequently, we found that lens lesions elevate the Lens Potential Convergence (LPC) up to 36.05% ± 0.06% (SD) of retinal potential levels in LDE model (compared with 28.10% ± 0.09% (SD) in NAE model as a reference). Using Axial Current Steering (ACS) by adjusting the current ratio between electrodes within paired channels, LPC was successfully reduced to 3.62% ± 0.01% (SD), accompanied by a 28% reduction in 90% retinal potential peak width surprisingly. This work elucidates the interference mechanisms of TIS targeting in enriched eye model and validates the effectiveness of the proposed ACS strategy, providing theoretical support for the future clinical application of TIS.