A Review of Online Classification Performance in Motor Imagery-Based Brain–Computer Interfaces for Stroke Neurorehabilitation

Motor imagery (MI)-based brain–computer interfaces (BCI) have shown increased potential for the rehabilitation of stroke patients; nonetheless, their implementation in clinical practice has been restricted due to low accuracy performance. To date, although a lot research carried out benchmarking and highlighting most valuable classification algorithms BCI configurations, them use offline data are not from real performance during closed-loop (or online) sessions. Since training relies on availability an accurate feedback system, we surveyed articles current past EEG-based frameworks who report online movement two upper limbs both healthy volunteers patients. We found that recently developed deep-learning methods do outperform traditional machine-learning algorithms. In addition, patients subjects exhibit similar configurations. Lastly, terms neurofeedback modality, functional electrical stimulation (FES) yielded best compared non-FES systems.