SOURCE RECORD / S0197
采用深度学习解码的非侵入式脑机接口连续追踪
Continuous tracking using deep learning-based decoding for noninvasive brain-computer interface
- 作者或机构
- Forenzo D, Zhu H, Shanahan J, Lim J, He B
- 发布日期
- 2024-04-30
- 来源类型
- peer_reviewed_human_study
- 技术路线
- non-invasive EEG motor imagery with deep-learning decoding
- 功能场景
- continuous 2D cursor control
- 权利状态
- mirror_allowed / CC BY 4.0
证据摘要
证据支持：深度学习解码用于在线连续二维追踪。关键数字：最后一次会话中，EEGNet 和 PointNet 相对传统解码器均为 p<0.0001，两种深度模型之间为 p=0.2768；性能随个人训练数据增加而提高。局限：健康且有经验的使用者、特定 2D 虚拟光标任务和逐会话累积数据；跨人预训练未显著改善早期表现，不能外推患者或居家使用。
关键数字
last-session NMSE: EEGNet and PointNet each versus traditional decoder p<0.0001; EEGNet versus PointNet p=0.2768; performance improved as subject-specific data accumulated
局限
Healthy experienced users; task-specific 2D virtual cursor; models accumulated session data; pretraining on other subjects did not significantly improve early performance; not patient or home evidence