ReviewUsing Recent BCI Literature to Deepen our Understanding of Clinical Neurofeedback: A Short Review
Introduction
Through their recent paper, Alkoby et al. (2017) provide the readership with an extensive and very insightful review of the factors influencing NeuroFeedback (NF) performance. These factors are drawn from both the NF literature and the Brain–computer interface (BCI) literature. Our short review aims to complement the review of Alkoby et al. by depicting some additional recent BCI literature. The object is to highlight this literature and discuss its potential relevance and usefulness to better understand the processes underlying NF and further improve the design of clinical trials assessing NF efficacy. Indeed, we are convinced that while they have fundamental differences, by building upon one another both the BCI and NF communities could reach compelling achievements.
As extensively described by Alkoby et al. (2017), the efficacy of clinical NeuroFeedback (NF) is subject to significant between-patient and between-study variability. The clinical efficacy of NF is heavily debated, particularly regarding psychiatric disorders. For this reason, this paper is devoted specifically to clinical NF. Some researchers indeed suggest that the clinical efficacy of NF is underlain by a placebo effect (Thibault et al., 2017). We agree that the level of evidence is still weak concerning the clinical efficacy of NF, and that a placebo effect may be involved to some extent. However, it is unlikely that this lack of evidence is due to the fact that NF is fully underlain by a placebo effect. Rather, we hypothesize that it may be due to the lack of randomized controlled trials (RCT) assessing NF learning effects. Yet, in order to rigorously estimate these learning effects, and provide a higher level of evidence for the clinical efficacy of NF, the variables influencing these effects should first be identified. In this paper, we argue that recent BCI results could be relevant and useful to identify such variables and help us deepen our understanding of the clinical efficacy of NF.
As stated by Sitaram et al. (2016) “much remains to be investigated, including the integration of the vast knowledge of training and learning psychology into NF protocols”. Thus, a human-factor-centered standpoint, considering the influence of the technology and the way it was designed on patients’ achievements (Sanders and McCormick, 1993) is required. A human-factor-centered standpoint would take into account the interaction between the patient and the system during the NF procedure. Such an approach could help us understand how various factors affect the ability of patients to learn to modulate the target neurophysiological pattern – i.e., the EEG feature(s) that the patient is learning to self-regulate (e.g., alpha rhythm power, the theta/beta power ratio, etc.) – during NF training (Micoulaud-Franchi et al., 2015, Arns et al., 2017). These include factors such as the design of the NF training protocol (e.g., type of feedback), the neurophysiological features, and the states (e.g., motivation) and traits (e.g., self-reliance) of the users (Jeunet et al., 2015b). This human-factor-centered standpoint was adopted in the review by Alkoby et al. (2017), in which the authors depict many factors that affect NF efficacy. Their goal in doing so was to promote the use of these factors to adapt NF training protocols to the user’s personality, and to their cognitive and neurophysiological profiles. In order to adapt these training protocols, the authors propose to focus on three aspects: neurophysiological features, feedback and mental strategies.
Although their review is already extensive and very instructive, further insight can be gained by studying the recent literature on training and learning in the field of BCIs, and more specifically in the field of Mental-Imagery-based BCIs (MI-BCIs) (Wolpaw and Wolpaw, 2012, Jeunet et al., 2016, Jeunet, 2016). MI-BCIs differ from NF in that the goal of MI-BCIs is to control an application without moving, by modulating specific brain rhythms through the completion of Mental-Imagery (MI) tasks. These tasks can be motor-imagery tasks, such as imagining moving one’s hands (Pfurtscheller and Neuper, 2001), or non-motor-imagery tasks, such as mental calculation or mental rotation (Friedrich et al., 2012, Jeunet et al., 2015b), all these mental tasks being detectable in EEG signals. The rationale for this approach is that performing each of these mental-imagery tasks will induce modulations of different brain rhythms, which are theoretically specific to each task. Each task is associated with a specific control command, such as “imagine left-hand movements to turn the wheelchair towards the left” and “imagine right-hand movements to turn the wheelchair towards the right” (Clerc et al., 2016). Thus, the system is able to detect modulations of the user’s brain activity and determine which command the BCI user intended to send. For instance, a decrease in mu amplitude over the left sensorimotor cortex should occur when users imagine a right-hand movement, i.e., when they want to turn right (Pfurtscheller and Neuper, 2001).
Both NF and MI-BCI users need to learn to regulate their neurophysiological EEG activity, using the feedback they are provided with, in order to produce specific EEG patterns (Neuper and Pfurtscheller, 2009, Sherlin et al., 2011, Lotte et al., 2013, Strehl, 2014). The objective is either to reach a target EEG pattern in NF (Sherlin et al., 2011, Strehl, 2014, Gruzelier, 2014a) or to produce a given EEG pattern that can be translated into a given command for an application in BCI (Neuper and Pfurtscheller, 2009, Lotte et al., 2013, Clerc et al., 2016). Consequently, similar cognitive and neurophysiological processes are likely to be triggered during both BCI and NF training procedures. Thus, we advocate considering the literature on BCI training to deepen our understanding of NF efficacy – in the same way that the BCI community should avail themselves of the NF literature.
First, we attempt to give a brief review of the relevant BCI literature, in order to complement the review by Alkoby et al. (2017). Indeed, the BCI community is also currently investigating the factors that influence user performance, training and learning. Notably, three main categories of factors were identified based on a review of the literature (Jeunet et al., 2016): task-specific factors, cognitive and motivational factors and technology-acceptance†-related factors. We suggest that these factors could be relevant for clinical NF training as well. Next, we elaborate on the potential implications of this research for improving the design of NF sessions and clinical NF efficacy, i.e., to reduce the clinical symptoms to which the target neurophysiological patterns are associated. We conclude with a summary and a diagram that outlines a framework (in Fig. 1), which takes into account the different factors identified in the review, in order to deepen our understanding of EEG signal self-regulation during NF, thereby potentially improving the clinical efficacy of NF.‡
Section snippets
Adapting the neurofeedback training protocol using a human-factor-centered standpoint
In the coming section, we provide information from the BCI training literature that could be relevant to adapt NF procedures to each patient, following the structure used by Alkoby et al. (2017), namely: (1) adapted neurophysiological features, (2) adapted feedback and (3) adapted mental strategy. Each of these three sections covers the factors belonging to each of the three categories mentioned in the introduction: task-specific factors, cognitive/motivational factors and
Task-specific and task-unspecific variables influencing the efficacy of clinical neurofeedback
The literature reviewed in the previous sections suggests that adapting training protocols to each user based on human-factors may help to improve BCI learning efficacy. It seems relevant to reinforce this practice in RCT evaluating clinical NF procedures, with the goal of optimizing the level of evidence concerning clinical NF efficacy in patients with mental/brain disorders.
As emphasized by Arns et al. (2014) many double-blind RCT are very well designed regarding the clinical trial itself,
Conclusion
As highlighted in this paper, more and more effort is devoted to the understanding of between-patient variability and to the control of between-study variability. The paper by Alkoby et al. (2017) and this paper offer complementary reviews of the literature regarding the factors, which have been experimentally shown to correlate with NF/BCI performance.
The different papers reviewed here suggest that in order to better understand the processes underlying clinical NF efficacy (and potentially
Acknowledgments
This work was supported by the French National Research Agency with the REBEL project and grant ANR-15-CE23-0013-01, the European Research Council with the BrainConquest project (grant ERC-2016-STG-714567), the EPFL/Inria International Lab and the Swiss National Foundation (grant IZSEZ0_179349/1).
References (90)
- et al.
Computer usage: The impact of computer anxiety and computer self-efficacy
Procedia Soc Behav Sci
(2015) - et al.
Mutual information-based selection of optimal spatial–temporal patterns for single-trial EEG-based BCIs
Pattern Recogn
(2012) - et al.
Evaluation of neurofeedback in ADHD: The long and winding road
Biol Psychol
(2014) - et al.
Neurofeedback: one of today's techniques in psychiatry?
L'Encéphale
(2017) - et al.
Closed-loop adaptation of neurofeedback based on mental effort facilitates reinforcement learning of brain self-regulation
Clin Neurophysiol
(2016) - et al.
Neurophysiological predictor of SMR-based BCI performance
NeuroImage
(2010) - et al.
Computer anxiety and its correlates: a meta-analysis
Comput Hum Behav
(1999) - et al.
The “sense of agency” and its underlying cognitive and neural mechanisms
Conscious Cogn
(2008) - et al.
Mindfulness-based cognitive therapy for generalized anxiety disorder
J Anxiety Disord
(2008) - et al.
The effect of distinct mental strategies on classification performance for brain–computer interfaces
Int J Psychophysiol
(2012)
A psychoengineering paradigm for the neurocognitive mechanisms of biofeedback and neurofeedback
Neurosci Biobehav Rev
Causal influence of gamma oscillations on the sensorimotor rhythm
NeuroImage
EEG-neurofeedback for optimising performance. I: a review of cognitive and affective outcome in healthy participants
Neurosci Biobehav Rev
EEG-neurofeedback for optimising performance. III: A review of methodological and theoretical considerations
Neurosci Biobehav Rev
Brain–computer communication: self-regulation of slow cortical potentials for verbal communication
Arch Phys Med Rehabil
Should the parameters of a BCI translation algorithm be continually adapted?
J Neurosci Methods
Neurofeedback: time needed for a promising non-pharmacological therapeutic method
Lancet Psychiatry
Electroencephalographic neurofeedback: Level of evidence in mental and brain disorders and suggestions for good clinical practice
Neurophysiol Clin
Machine learning for real-time single-trial EEG-analysis: from brain–computer interfacing to mental state monitoring
J Neurosci Methods
Imagery of motor actions: differential effects of kinesthetic and visualmotor mode of imagery in single-trial EEG
Brain Res Cogn Brain Res
An auditory brain–computer interface
J Neurosci Methods
Effect of mindfulness meditation on brain–computer interface performance
Conscious Cogn
When can neurofeedback join the clinical armamentarium?
Lancet Psychiatry
Can we predict who will respond to neurofeedback? A review of the inefficacy problem and existing predictors for successful EEG neurofeedback learning
Neuroscience
Evidence for efficacy of neurofeedback in ADHD?
Am J Psychiatry
Optimizing the channel selection and classification accuracy in EEG-based BCI
IEEE Trans Biomed Eng
Biased feedback in brain–computer interfaces
J NeuroEng Rehabil
Individual EEG alpha activity analysis for enhancement neurofeedback efficiency: two case studies
J Neurother
The Berlin brain–computer interface: machine learning based detection of user specific brain states
J Univ Comput Sci
Optimizing spatial filters for robust EEG single-trial analysis
IEEE Signal Process Mag
Two brains, one game: design and evaluation of a multiuser BCI video game based on motor imagery
IEEE Trans Comput Intell AI Games
Meditation and neurofeedback
Front Psychol
Embodied neurofeedback with an anthropomorphic robotic hand
Sci Rep
A brain–computer interface with vibrotactile biofeedback for haptic information
J NeuroEng Rehabil
Vibrotactile feedback for brain–computer interface operation
Comput Intell Neurosci
Multivoxel neurofeedback selectively modulates confidence without changing perceptual performance
Nat Commun
Optimal experience in work and leisure
J Pers Soc Psychol
Generalized methods for user-centered brain–computer interfacing
EEG-neurofeedback as a tool to modulate cognition and behavior: a review tutorial
Front Human Neurosci
Automatic motor task selection via a bandit algorithm for a brain-controlled button
J Neural Eng
Investigating the role of combined acoustic-visual feedback in onedimensional synchronous brain computer interfaces, a preliminary study
Med Dev Evid Res
Neurofeedback in children with ADHD: validation and challenges
Expert Rev Neurother
Cited by (40)
EEG sensor driven assistive device for elbow and finger rehabilitation using deep learning
2024, Expert Systems with ApplicationsNeurofeedback and epilepsy: Renaissance of an old self-regulation method?
2024, Revue NeurologiqueITACA: An open-source framework for Neurofeedback based on Brain–Computer Interfaces
2023, Computers in Biology and MedicinePhysical principles of brain–computer interfaces and their applications for rehabilitation, robotics and control of human brain states
2021, Physics ReportsCitation Excerpt :The functional unit of biological feedback of the passive BCI is bounded on the diagram in Fig. 9 by a dashed line. The BCI can affect the user by means of biological feedback or neurofeedback [153,154] implemented in various ways, for example, by a visual stimulus in the form of a message on the monitor screen, sound signal, vibration, etc. [152,155]. On the other hand, the feedback can be used and affect directly the neural ensemble in the brain, bypassing standard channels of information transmission, for example, using the transcranial magnetic stimulation (TMS) [156] or transcranial electrical stimulation (tES) [157].
The learning effects and curves during high beta down-training neurofeedback for patients with major depressive disorder
2020, Journal of Affective DisordersCitation Excerpt :This psycho-engineering model included five properties: perceptibility, autonomy, mastery, motivation, and learnability. Thus, Batail et al. (2019) suggested NFB researchers assessing how well clients can self-regulate their brain activities at a time period, and evaluating their behavior improvements after NFB training, such as changes in performance, attention networks, cognitive flexibility, mental rotation tasks, semantic and phonological fluency, and clinical symptoms (Alkoby et al., 2018; Batail et al., 2019; Choi et al., 2011; Jeunet et al., 2018). There were also debates about how NFB generated the behavioral and neural effects, with a series of articles and correspondences on few journals, including Journal of Attention Disorder, Lancet Psychiatry, and Brain.