Table 1.
BCI user performance metrics.
| References | Metric | Purpose | Criteria | Remark | |||
|---|---|---|---|---|---|---|---|
| 1 | 2 | 3 | |||||
| Resting | Blankertz et al. (2010) and Sannelli et al. (2019) | SMR predictor | Predict CA | ✓ | Not designed for instantaneous feedback; requires recordings of brain at rest | ||
| Ahn et al. (2013b) | (Pα+Pβ)/(Pθ+Pγ) | Predict CA | ✓ | ||||
| Zhang et al. (2015) | M1 spectral entropy | Predict CA | ✓ | ||||
| Active brain state | Wolpaw et al. (2002) and Wolpaw et al. (2000) | Strength of SMR suppression | Feedback | ✓ | ✓ | No synthesis of current and Previous feedback | |
| Pfurtscheller et al. (2003) and Sannelli et al. (2019) | Classifier-based, predicted task label | Feedback | ✓ | May not reflect changes in user performance; difficult to interpret | |||
| Bamdadian et al. (2014) | Pre-trial onset EEG signal power | Predict CA | ✓ | Pre-trial activity may not reflect task performance | |||
| Shu et al. (2018) | Laterality index based on mean event-related EEG signal power | Predict CA | ✓ | Not conducive to online use—needs user-specific frequencies | |||
| Lotte and Jeunet (2018) | Riemannian interclass discriminability and intraclass consistency | Track user skill | ✓ | Classifier-independent | |||
| Duan et al. (2021) | Riemannian metrics (Lotte and Jeunet, 2018) + diffusion maps | Feedback | ✓ | ✓ | ✓ | Performance represented only visually; user infers performance changes from visualization | |
CA, classifier accuracy; M1, primary motor cortex; SMR, sensory motor rhythm.