Ding Yidan, Udompanyawit Chalisa, Zhang Yisha, He Bin
Department of Biomedical Engineering, Carnegie Mellon University, Pittsburgh, PA, USA.
Department of Electrical and Computer Engineering, Carnegie Mellon University, Pittsburgh, PA, USA.
Nat Commun. 2025 Jun 30;16(1):5401. doi: 10.1038/s41467-025-61064-x.
Brain-computer interfaces (BCIs) connect human thoughts to external devices, offering the potential to enhance life quality for individuals with motor impairments and general population. Noninvasive BCIs are accessible to a wide audience but currently face challenges, including unintuitive mappings and imprecise control. In this study, we present a real-time noninvasive robotic control system using movement execution (ME) and motor imagery (MI) of individual finger movements to drive robotic finger motions. The proposed system advances state-of-the-art electroencephalography (EEG)-BCI technology by decoding brain signals for intended finger movements into corresponding robotic motions. In a study involving 21 able-bodied experienced BCI users, we achieved real-time decoding accuracies of 80.56% for two-finger MI tasks and 60.61% for three-finger tasks. Brain signal decoding was facilitated using a deep neural network, with fine-tuning enhancing BCI performance. Our findings demonstrate the feasibility of naturalistic noninvasive robotic hand control at the individuated finger level.
脑机接口（BCIs）将人类思想与外部设备相连，为运动功能受损的个体以及普通人群提高生活质量带来了潜力。非侵入式脑机接口广泛适用于广大用户，但目前面临一些挑战，包括映射不直观和控制不精确等问题。在本研究中，我们展示了一种实时非侵入式机器人控制系统，该系统利用单个手指运动的动作执行（ME）和运动想象（MI）来驱动机器人手指运动。所提出的系统通过将预期手指运动的脑信号解码为相应的机器人动作，推动了先进的脑电图（EEG）-脑机接口技术的发展。在一项涉及21名身体健康且有脑机接口使用经验的用户的研究中，我们在双指MI任务中实现了80.56%的实时解码准确率，在三指任务中实现了60.61%的实时解码准确率。使用深度神经网络促进了脑信号解码，微调提高了脑机接口的性能。我们的研究结果证明了在个体手指水平上实现自然主义非侵入式机器人手部控制的可行性。
