Self-optimizing framework for natural sensory feedback through transcutaneous electrical nerve stimulation

The absence of sensory input following limb amputation significantly impairs prosthesis embodiment and natural interaction with the environment. Bidirectional prostheses integrating sensory feedback with electrical stimulation have shown promise in enhancing user control, task precision, and overall satisfaction. However, optimizing stimulation parameters to elicit naturalistic sensations remains a major challenge, particularly for transcutaneous electrical nerve stimulation (TENS), which often produces tingling rather than pressure-like sensations. In this study, we used the SELFOPT framework, a subject-centric approach that empowers users to iteratively refine stimulation parameters in real time, bypassing experimenter intervention. We compare SELFOPT to a traditional ramp-up approach (RAMP) in a study involving 25 healthy participants, optimizing TENS parameters for soft and heavy tap sensations. Our results demonstrate that SELFOPT-derived parameters elicit significantly more natural and pressure-like sensations compared to RAMP. Analysis of parameter selection revealed a complex relationship between pulse amplitude, pulse width, and pulse frequency for eliciting a natural sensation. Furthermore, SELFOPT provides retrospective insights into optimal parameter regions, contributing to a deeper understanding of sensory encoding. Beyond enhancing naturalness, SELFOPT presents a versatile framework for individualized parameter tuning in neuroprosthetic applications. Its interactive nature eliminates experimenter-subject bottlenecks, supports multi-parameter optimization, and enables real-time adaptation. These findings pave the way for more intuitive and embodied prosthetic technologies, with broader implications for neurostimulation-based sensory feedback systems.