Among different brain imaging techniques that have been applied to BCI, electroencephalogram (EEG) is the most commonly used, owing to the minimal risk involved and the relative convenience. In general, power spectrum analysis is typically used for decomposing the EEG signal into different frequency bands. This is useful in that it is accepted that specific bands of power in the EEG spectrum are linked to a broad variety of perceptual, sensorimotor, and cognitive operations. As far as BCI systems are concerned, the most important frequency activity in the EEG spectrum lies below 40 Hz. There are 5, possibly 6, accepted bands of EEG activity below 40 Hz which are often associated with specific states of mind. For instance, the frequencies between 8 Hz and 13 Hz, referred to as alpha rhythms, are usually associated with a state of relaxed wakefulness. Investigations on EEG oscillations have particularly focused on alpha, theta, beta and gamma frequencies, and only recently on very slow activity. These oscillators are active usually in a random way. By application of sensory stimulation these generators may be coupled and act together in a coherent way. This synchronization and enhancement of EEG activity gives rise to ‘‘induced rhythms’’ (or ‘‘evoked’’) and it is crucial to establish whether or not users can obtain power within distinct frequency bands in a voluntary manner. EEG variations, then, need to be noticed in order to be used as control signals (Lance et al., 2012). The application of neurofeedback for regulation purposes has used computer software to show visual and audio feedback according to brainwaves such as alpha, beta, or theta frequencies (Baehr, Rosenfeld & Baehr, 2001; Peniston, 1995). The paradigm typically in use is that of evoked potentials. Evoked potentials or event-related potentials (ERPs) occur from perception to an external stimulus or set of stimuli. VEPs can be generated from visual sensory stimulation to achieve a neuronal synchronization with a similar frequency (or harmonic) to the flash of light presented. With repeated stimulation at short intervals, the brain’s response to each subsequent stimulus is evoked before the response to the prior stimulus has finished. Instead of being permit to return to a baseline state, a so-called steady-state response (SSVEP) is elicited (Middendorf, McMillan, Calhoun & Jones, 2000). Light therapy (Light-Emitting Diodes – LEDs) has also been show to produce beneficial cellular and physiological effects (Walter, Dovey & Shipton, 1946), particularly with the recent discovery of non visual opsins (Pino & La Ragione, 2014; Youssef, 2013).
In order to show meaningful and interactive visual feedback, rather than merely the feedback directly associated with brainwaves, the BCI technique is often considered. Sensory evoked potentials (SEPs) are electrical potentials recorded from the central nervous system; while stimulating sense organs, SEPs are phase-locked to the stimulus. Therefore, they can be enhanced. They can be interpreted as a reorganization of spontaneous brain oscillations in response to a stimulus (Basar, Demilrap, Schurmann, Basar-Eroglu & Ademoglu, 1999). SSVEPs were recently employed to show that patients suffering a major depressive disorder had low activation in the right temporo-parietal cortex when watching arousing stimuli, although they had normal occipital activation (Moratti, Rubio, Campo, Keil & Ortiz, 2008), probably for a deficit in the arousal of related brain structures, along with intact basic visual stimuli processing in patients with major depressive disorder. SSVEPs were typically used as a marker of anticipative anxiety (Gray, Kemp, Silberstein & Nathan 2003). Regarding the emphasis on the alphas, it was proposed an evolutionary-based perspective of brain oscillations relevant for research of EEG correlates of personality, suggesting that delta, theta, and alpha oscillations reflect activities of three hierarchical phylogenetic brain systems. Alpha is the dominant frequency in adult humans, theta dominates in the EEG of lower mammals, and delta in the reptilian EEG. Some investigations indicated that the associations among psychometric measures of anxiety and depression and electroencephalogram (EEG) spectral power measures were positively correlated to alpha and negatively correlated to delta, independently from cortical area (Knyazev, Savostyanov & Levin, 2004). In high-anxiety subjects, alpha2 sub-band seems to be the most reactive while low-anxiety subjects tend to adjust to environmental changes by alpha3 power modification.
Brain-computer interface-based therapy might provide a useful complement to standard treatment methods and might lower cost by reducing the need for the presence of a therapist. Neuro-Upper (NU) is the BCI developed in our laboratory (by the second author) settled on brain responses to exogenous visual and auditory stimuli. There are several headsets with scalp sensors used in connection with a computer to create a system in order to intensify attention and meditation via EEG-based neurofeedback (Shangkai, Yijun, Xiaorong & Bo, 2014). Synchronization of oscillatory activities in distributed neural assemblies is a well-studied mechanism. It can be understood as a reflection of the cooperative and synchronous activity of neural assemblies with different EEG frequencies revealing synchronies related to different perceptual, motor or cognitive states. One property of oscillating elements is that they can be perturbed by an external periodic force becoming synchronized to this periodic event; in other words, the oscillating component starts to cycle with the same period as the external force. So, synchronization, entrainment, and locking are considered synonymous. The notion of driving brain oscillations by directly stimulating neuronal elements with rhythmic stimulation protocols has become increasingly popular (Thut, Schyns & Gross, 2011; Will & Berg, 2007). Controlled entrainment of brain rhythms may therefore prove highly advantageous for the study of human brain oscillation, and further research in this area may well lead to applications in rehabilitation or treatment of diseases. The ideal scenario to achieve this form of entrainment is to tune the frequency of the periodic force to the natural frequency of the to-be-entrained neuronal elements. Evoked potential generation is also of great interest for the study of entrainment.
Music with certain rhythmic parameters is firstly capable of triggering specific brain waves and physiological responses (Zatorre, 2003). Baroque music (60 bpm), for example, induce alpha rhythms and slow down heart and respiration rates, while music with driving rhythms or fast tempos (e.g., rock) stimulate beta waves and speed up heart and respiratory rate. When specific musical stimuli activate corresponding brain mechanisms, global extra-musical responses are generated, such as enhanced cortical synchrony (Juslin, Liljeström, Västfjäll, & Lundqvist, 2010). Respect to neuromediators, endorphins, endocannabinoids, dopamine and nitric oxide are altered in the musical experience (Boso, Emanuele, Minazzi, Abbamonte & Politi, 2006). Different forms of music activate distributed brain regions in unique ways, and it was showed that brain responds predictably to different styles or pieces of music (Salimpoor, Benovoy, Longo, Cooperstock & Zatorre, 2009) or voices (Loui, Bachori, Li, & Schlaug, 2013). Entrainment is most likely involved in periodic stimuli perception in primary auditory areas. It was suggested that it is through synchronization that the perception of metrical structure occurs and that perception, attention and expectation are all rhythmic processes subject to entrainment (Large & Kolen, 1994).
Most of studies indicated that such rhythmic stimulation may change attention and perception by modifying communication in oscillatory networks through their entrainment. How stimulation parameters and ongoing oscillations interact to give rise to entrainment will need to be studied in detail in future research. In addition, the duration of entrainment effects after stimulation will determine subsequent uses beyond basic research on brain oscillations. In our work we investigated real-time interactions with a BCI. The main aim of the study is to shown the effect of a device (NU), realized at our laboratory, in changing brainwaves pattern in real-time through audio-video entrainment regulating emotional states of individual with anxiety and depressive disorders. A group of 7 participants (4 females) all right handed with mean age of 47.29 years old (SD=14.98) were recruited through advice in local newspaper. All subjects declared the absence of neurological or mental illnesses, and were screened against the photosensitive epilepsy. Informed, written consent was obtained from all of the subjects. We included participants that met the Diagnostic Statistical Manual of Mental Disorders-Fourth Edition Text Revision (DSM-IV-TR) criteria for almost depressive and anxiety disorders.
2.2. Material and device
Each participant received a structured interview (Structured Clinical Interview for DSM-IV-R Axis I Disorders (SCID-4-RV; APA, 2000) and prior to the inclusion into the study, all individuals underwent a comprehensive clinical and neuropsychological assessment with the following tests: Spielberger State Trait Anxiety Inventory (STAI; Spielberger et al., 1983), Hamilton Rating Scale for Depression (HAMS; Hamilton, 1960), Wechsler Adult Intelligence Scale Revised (WAIS-R; Wechsler, 1981), Raven’s Progressive Matrices (SPM; Raven, Raven & Court, 2003) and Mini Mental State Examination (MMSE; Folstein Folstein & McHugh, 1975).
Neuro-Upper used NeuroSky Mindwave headset (Fig. 2, b) measuring raw data of the EEG activity at a 512 Hz sampling rate. To reduce electrical interference generated by the human body, the device has a “base” contact, which is attached to the earlobe and allows filtering electrical waves produced by the body (noise). The apparatus detects MindWave brain electrical activity and decomposes the signal into eight outputs according to their frequency. Values are assessed at a rate of 1Hz. A array of colored lamps producing flickering light has been used with a direct mapping to the monitor of a personal computer. Microsoft Visual C # software is utilized to design the BCI, to mainly measure the brainwave signals when subjects listening to music, and the data for different frequency bands can be observe in real-time from the interface.
Fig. 1. a) Emotiv EPOC neuroheadset, b) Neurosky Mindwave
The frequency of the light sources able to emit modulated light (eight PAR 56 Omnilux lamps, 300 Watts, 26 x 23,5 x 22 cm each covered by coloured gelatin sheets, Fig. 2) is controlled by electronic circuits, which internally used micro-controllers providing frequency of each brainwave. The lamps array is placed about at 210 cm from the participants’ seat. The challenges of this software are the real time frequency generation to present visual repetitive stimuli on the computer screen to the user (e.g., feedback that shows real time brainwaves activity). This application receives eight commands from the BCI analysis software and was based on the self-regulation of the SSVEP amplitude. BCI feedback task is basically the user interface that translate data coming from the signal processing unit of the BCI into a visual representation on the screen to provide a second visual feedback to the user.
Figure 2: Neuro-Upper: hardware system and effector device
2.3. Procedure
Basing on previous studies, different play-lists for the audio stimulation are selected for different disorders and presented for each subject in sessions of 30 minutes for four consecutive months. Subjects were seated on a comfortable chair in a dim room wearing the Mindwave headset with eyes open and listening to the stimuli via headphones. We recorded EEGs of the subjects whose had been asked to listen their musics observing the screen with their changing brain rhythms. Visual stimulation was delivered using a BCI stimulator, placed at 210 cm in front of the chair. Waves graphics patterns are rendered on the computer screen in form of histograms bar showing real-time frequency for each brainwave (Fig. 3).
Figure 3: Graphics of brainwaves’ real-time frequency
Indeed, the participant received feedback from the effector device highlighting each lamp, and the visual one in form of dynamic vertical bars on the computer screen. Data were recorded continuously and saved to a file. The procedure for each session took approximately 40 minutes per subject.
3. Results
The main question of our laboratory study was whether there was any significant difference in outcome measures between Pre and Post-treatment. Wilcoxon rank-sum test was used to determine these differences. P values <.05 were considered significant. Figure 4 addresses this question by showing the results for the Hamilton Depression Rating Scale score where significant differences were found (HAMS 19.71 versus 7.714, P = 0.022). State Trait Anxiety Inventory reported increasing levels of severity of anxiety (STAI-Y1 42.14 versus 53, P = 0.14; STAI-Y2 40.71 versus 48.14, P = 0.15, respectively). However, no difference is showed for Mini Mental Status Examination (29.71 versus 29.57, P = 0.85) in which participants reached maximum score also at the baseline. Significant improvements were noted in outcome measures for cognitive function, specifically in Wechsler Adult Intelligence Scale III (QI total 117.9 versus 143.9, P = 0.016), for Verbal (VIQ 104.1 versus 130.6, P = 0.016) and Performance (PIQ 104.1 versus 145.6, P = 0.016) respectively, and Ravens Progressive Matrices scores (RPM 45.57 versus 48.14, P = 0.209).
On average, forty-eight sessions of EEG were recorded. A frequency-domain-based method was implemented (Welch, 1967). Namely we estimated the power spectra of M seconds of data, and the power values at the eight base frequencies were extracted. The power spectrum analysis is a kind of analysis technique that is used when the time-series signals changing according to time is transformed into the frequency field and the signal aspect is evaluated by the change degree of the frequency. A time series analysis on median scores of spectral power for each sub-bands was carried out. Statistical analysis showed that some quantitative (median) values are rather stable, precisely delta and theta bands, while the others showed a descending trend.
Figure 4. Hamilton Depression Rating Scale Pre-Post Treatment mean scores
4. Conclusions
Interest in the BCI field is expected to increase, and BCI design and development will in all probability continue to bring benefits to the daily lives of people. Furthermore, recent results suggests that BCI device may find useful applications in the general population, and not just for people with severe disabilities. The present paper has indicated preliminary evidence from the intervention through a methodology in which a combination of audio-visual entrainment and BCI is used in order to treat anxiety and depressive symptoms. Despite methodological limitations (the absence of a control group) and heterogeneity of the results (the increase of anxiety scores), our findings supports the idea that a dis-regulation of brainwaves patterns and abnormal connectivity may account for a wide range of symptoms in these disorders, including emotional disturbances, and cognitive dysfunction, and that at least some of these disturbances can be alleviated by external manipulation or audiovisual entrainment. Neurofeedback real-time allows simultaneous acquisition, analysis, visualization, and feedback of brainwaves. The behavioral change produced by NU was demonstrated in depressive symptomatology remission, probably due to a modulation of the insular activity, that plays a central role in sensory integration, emotion, and cognition, or to generalized brain activation, as evidenced by the resulting improvement in cognitive abilities (WAIS and RPM scores).
A main limitation of the present study was the real nature of the signals recorded, because it is probable that Mindwave record non-brain signals, such as electromyographic signals from cranial, or electro-oculographic signals from eye movements and blinks. So, in the next future we aimed to use Emotiv Epoch (Fig. 1, a) for the new adaptation of the device. Furthermore, although BCI training implies learning through operant conditioning, little is known of the deeper mechanism of learning using this methodology. Other studies have to explore whether brain self-regulation persists if feedback is removed and whether the capability for self-regulation will persist with people without psychological disorders showing the same pattern of increase in cognitive ability. Although our results are still empirical (given the limited number of participants and measurement available thus far) and the solid confirmation of our claims requires further investigation and more data analysis, we believe that the presented results open the door for future applications of BCI as a research tool and therapeutic approach.