Abstract-Objective: This study advances neuroprosthetic design by leveraging neural interface and artificial intelligence (AI) to enable seamless translation of motor intent into intuitive, lifelike control of prosthetic devices. Methods: Our approach is founded on three key innovations. First, we develop neuromodulation microchips and devices for acquiring high-fidelity peripheral nerve signals from microelectrode implants. Second, we design and train AI models based on recurrent neural networks (RNN) to decode neural patterns and accurately predict motor intent. Third, we create a prototype neuroprosthetic hand that integrates all components into a portable, self-contained system. Results: We demonstrate that acquired nerve signals contain distinct neural signatures for individual hand gestures, which can be decoded by deep learning AI with superior accuracy compared to traditional machine learning techniques. Clinical validation shows the decoder can predict 15 degrees of freedom (DOF) in regression tasks, with a mean squared error (MSE) ranging from 0.001 to 0.01. An optimized classification model achieves 95–96% accuracy in distinguishing movements of the five fingers. When deployed on the prototype neuroprosthetic hand, the AI model enables amputees to control individual prosthetic fingers precisely and intuitively in real time. Conclusion & Significance: This work bridges human peripheral nerves with advanced AI to enable precise, intuitive control of prosthetic devices. It represents a transformative step toward next-generation neuroprosthetics, with the potential to significantly improve the quality of life for individuals with motor impairments.

Index Terms—artificial intelligence, deep learning, peripheral nerve, neural decoder, neural interface, neuroprosthesis

Xem bài viết tại: https://ai.siu.edu.vn/wp-content/uploads/2024/12/Nguyen-Anh-Tuan.pdf

NEXT-GENERATION NEUROPROSTHETICS: BRIDGING THE HUMAN MIND AND ARTIFICIAL INTELLIGENCE

Bài viết khác

Bài viết khác