Training-Free Stimulus Encoding for Retinal Implants via Sparse Projected Gradient Descent

TL;DR

This paper introduces a training-free, sparse projected gradient descent method for stimulus encoding in retinal implants, improving image reconstruction fidelity without requiring neural network training.

Contribution

It formulates stimulus encoding as a sparse least-squares problem and develops an efficient solver exploiting perceptual sparsity, enhancing reconstruction quality in retinal implant simulations.

Findings

Achieved up to +0.265 SSIM increase in reconstructions.

Improved PSNR by +12.4 dB over baseline methods.

Reduced MAE by 81.4% on Fashion-MNIST.

Abstract

Retinal implants aim to restore functional vision despite photoreceptor degeneration, yet are fundamentally constrained by low resolution electrode arrays and patient-specific perceptual distortions. Most deployed encoders rely on task-agnostic downsampling and linear brightness-to-amplitude mappings, which are suboptimal under realistic perceptual models. While global inverse problems have been formulated as neural networks, such approaches can be fast at inference, and can achieve high reconstruction fidelity, but require training and have limited generalizability to arbitrary inputs. We cast stimulus encoding as a constrained sparse least-squares problem under a linearized perceptual forward model. Our key observation is that the resulting perception matrix can be highly sparse, depending on patient and implant configuration. Building on this, we apply an efficient projected residual norm steepest descent solver that exploits sparsity and supports stimulus bounds via projection. In silico experiments across four simulated patients and implant resolutions from $15\times15$ to $100\times100$ electrodes demonstrate improved reconstruction fidelity, with up to $+0.265$ SSIM increase, $+12.4\,\mathrm{dB}$ PSNR, and $81.4\%$ MAE reduction on Fashion-MNIST compared to Lanczos downsampling.