基于脑电图的步行与步态相位分类:范围综述
仅1项在机器人辅助步态训练中分类摆动与站立相,最高准确率83.06%;无研究在神经疾病人群中分类步态相位,领域缺乏方法学共识。
要点
- 步态相位分类研究少,多数仅分两相位且以跑台步行为主。
- 仅1项在机器人辅助步态训练中分类摆动与站立相,准确率83.06%。
- 无研究在神经疾病人群中分类步态相位,方法学缺乏共识。
核心结论
- 脑电图步态分类领域缺乏方法学共识,分类准确率变异大,神经疾病人群步态相位分类研究完全缺失。
研究设计与样本
- 预注册范围综述,检索SCOPUS、EMBASE、PubMed和Web of Science,从15,915项唯一研究中纳入62项。
主要发现
- 仅1项研究在机器人辅助步态训练中分类摆动与站立相,最高准确率83.06%。
- 步态相位研究中9项采用离线分析,1项在线分析对摆动与站立相分类准确率达82.3%。
- 感觉运动区电极最常用于分类,感觉运动区 consistently 表现步态相关判别性脑电特征。
关键数值与推荐
- 机器人辅助步态训练中摆动与站立相分类最高准确率83.06%。
- 在线分析摆动与站立相分类准确率82.3%。
临床怎么用
- 当前证据尚不足以支持脑电图步态分类在康复临床的常规应用,需方法学共识后再推进。
局限性
- 领域缺乏方法学共识,分类准确率变异大,且无神经疾病人群步态相位分类研究。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
Brain-computer interfaces (BCIs) have the potential to optimise robotic-assisted gait rehabilitation by closing the loop between user and device. Developing such systems requires reliable methods for classifying walking and its component phases to enable real-time feedback. However, research into electroencephalography (EEG)-based gait classification remains limited, with no consensus on optimal methodological or classification approaches. This scoping review aimed to identify (i) gait states and phases classified to date, (ii) methodological processes employed in gait and gait phase classification and (iii) classification systems used. A pre-registered scoping review was conducted. SCOPUS, EMBASE, PubMed and Web of Science were searched. Studies investigating EEG-based gait or gait phase classification, with or without robotic assistance were included. From 15,915 unique studies, 62 were included. Ten studies classified gait phases, with most limited to two phases (n = 5). The majority involved treadmill walking (n = 8), with only two investigating overground walking. Only one study addressed gait phase classification (swing versus stance) during robotic-assisted gait training (RAGT), achieving a maximum accuracy of 83.06%. No studies investigated gait phase classification in a neurological population. Almost all gait phase studies used offline analysis (n = 9). One study performed online analysis, achieving an accuracy of 82.3% for swing versus stance classification. The remaining studies classified gait against other forms of movement. Sensorimotor area electrodes were most frequently used for classification and sensorimotor regions consistently exhibited gait-related discriminative EEG features. Collation of methodological processes identified little consensus within the field. Furthermore, classification accuracy was variable. This scoping review summarises EEG-based gait classification systems, highlighting the need for methodological consensus. Knowledge gaps include the classification of more than two gait phases during free overground walking and limited RAGT-based studies. Notably, gait phase classification research in neurological populations is entirely lacking. Addressing these gaps is critical to advance BCIs for gait rehabilitation.
引用
- Ryan C, White C, Lennon O, et al. The classification of walking and phases of gait using EEG: a scoping review. J Neuroeng Rehabil. 2026. PMID: 42380885. doi:10.1186/s12984-026-02076-6
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。