Neurofeedback technology and scientific references

A very useful summary of the principles of neurofeedback, taken from a review published in 2021 (full reference below)


Neurofeedback Systems

Neurofeedback protocols aim to train users to achieve self-regulation of specific neural substrates through real-time feedback (Kim and Birbaumer, 2014; Sitaram et al., 2017). This learning process is grounded in operant conditioning (or reinforcement learning) such that desired brain activity is rewarded (Ros et al., 2014). Neurofeedback systems consist of three main components: an imaging modality (e.g., functional magnetic resonance imaging), a series of signal processing steps to extract and filter relevant (i.e., ideally neural) information, and a feedback presentation of this information to the user.



Neuroimaging techniques

Two different groups of neuroimaging techniques are used for neurofeedback studies: Electrical or magnetic signals (that result from dipole sources of electrical, neural activity) (Min et al., 2010) are the basis for electroencephalography (EEG) (Enriquez-Geppert et al., 2017) and magnetoencephalography (MEG) (Parkkonen, 2015). These neuroimaging techniques possess high temporal resolution (of up to 100 kHz in modern equipment), allowing for high sampling rates and thus frequent updates of presented neurofeedback. Blood oxygenation level (i.e., hemodynamic responses) forms the basis for functional magnetic resonance imaging (fMRI) (Weiskopf, 2012; Paret et al., 2019) and functional near-infrared spectroscopy (fNIRS) (Kohl et al., 2020). These imaging techniques thus provide an indirect measure of neural activity, which results from the metabolism of brain cells (Min et al., 2010). fMRI and fNIRS have lower temporal but higher spatial resolution compared to EEG, allowing for more specific targeting of brain structures.


EEG data processing during Neurofeedback training

During real-time data processing, recorded signals are converted into an output of the closed-loop system (Sitaram et al., 2017). Ideally, noise-reduction and feature extraction approaches are used to remove artifacts and convert the original time series into standardized and informative measures of neural activity (Gruzelier, 2014). Processing algorithms used to achieve better data quality vary between imaging techniques and regions of interest and remain an active field for methodological research. For example, EEG-based protocols usually involve self-regulation of frequencies or electrical potentials of specific EEG channels (Enriquez-Geppert et al., 2017). On the other hand, fNIRS- and fMRI-based neurofeedback systems focus on the up- or down-regulation of the hemodynamic signal in specific brain areas. Noteworthy, while fMRI allows recording from and thus targeting subcortical areas directly, the spatial resolution of fNIRS is limited to the cortical surface (Kohl et al., 2020).


Feedback system of neurofeedback training

Feedback presentation constitutes another relevant component of neurofeedback protocols. The goal of most currently employed paradigms is to provide users with real-time about the targeted neural activity, allowing them to adapt their control strategy to achieve a desired level of proficiency (Curran and Stokes, 2003; Birbaumer et al., 2013). Different perceptual modalities can be stimulated, e.g., via auditory, visual, vibrotactile, electrical or proprioceptive feedback systems (Sitaram et al., 2017). The choice and configuration of the feedback modality should be carefully planned because it can interfere negatively with self-regulation performance and learning curves of participants (McFarland et al., 1998; Birbaumer et al., 2013).



Trambaiolli, L. R., Cassani, R., Mehler, D., & Falk, T. H. (2021). Neurofeedback and the Aging Brain: A Systematic Review of Training Protocols for Dementia and Mild Cognitive Impairment. Frontiers in aging neuroscience, 13, 682683. Click HERE for full article