EEG microstates are transient, patterned, and quasi-stable states or patterns of an electroencephalography (EEG for short). These brief states, or microstates, tend to last anywhere from milliseconds to seconds. These transient periods are hypothesized to be the most basic initializations of human neurological tasks, and are thus nicknamed "the atoms of thought".[1]. Microstate data is obtained from a person's Alpha wave of their EEG signal.[2] This idea of these microstates being "quasi-stable" means that the "global [EEG] topography is fixed, but strength might vary and polarity invert."[3]
Contents |
Drs. Thomas Koenig (University Hospital of Psychiatry, Switzerland) and Dietricht Lehmann (KEY Institute for Brain-Mind Research, Switzerland)[1] are often credited as the pioneers of EEG Microstate analysis.[2] In their 1999 paper in the European Archives of Psychiatry and Clinical Neuroscience[1], Koenig and Lehmann had been analyzing the EEGs of schizophrenic patients, in order to investigate the potential basic cognitive roots of the disorder. They began to turn their attention to the EEGs on a millisecond scale. They determined that both normal subjects and schizophrenic patients shared these microstates, but they varied in characteristics between the two groups, and concluded that:
Isolating and analyzing one's EEG microstate sequence is a Post-hoc operation that typically utilizes several averaging and filtering steps. When Koenig and Lehman ran their experiment in 1999 they constructed these sequences by starting from a subject's eyes-closed resting state EEG. The first several event-free minutes of the EEG were isolated, then periods of around 2 seconds each were refiltered (Band-pass ≈ 2-20 Hz). Once the epochs were filtered, these microstates were analytically clustered into mean classes via k-means clustering, post hoc. [1]
Since the brain goes through so many transformations in such short time scales, microstate analysis is essentially an analysis of average EEG states. Koenig and Lehmann set the standard for creating classes, or recurrent averaged EEG configurations. Once all the EEG data is collected, a "prototype" EEG segment is chosen, with which to compare all other collected microstates. This is how the averaging process starts. Variance from this "prototype" is computed to either add it to an existing class, or to create a separate class. After similar configurations are "clustered" together, the process of selecting and comparing a "prototype" is repeated several times for accuracy. The process is described in more detail by Koenig & Lehmann:
"Similarity of EEG spatial configuration of each prototype map with each of the 10 maps is computed using the coefficient of determination to omit the maps' polarities. ...Separately for each class the prototype maps are updated combining all assigned maps by computing the first spatial principal component (Principal Component Analysis) of the maps and thereby maximizing the common variance while disregarding the map polarity." This process is repeated several times using different randomly selected prototype maps from among the collected data to use for statistical comparison and variance determination.[1]
Most studies[1] [4] [5] [6] [7] [8] [9] reveal the same 4 classes of microstate topography:
It is the current hypothesis that EEG Microstates represent the basic steps of cognition and neural information processing in the brain, but there is still much research that needs to be done to cement this theory.
Koenig, Lehmann et al. 2002 [10]
This study investigated EEG Microstate variance across normal humans of varying age. It showed a "lawful, complex evolution with age" [10]with spikes in mean microstate duration around ages 12, 16, 18, and 40-60 years, suggesting that there is significant cerebral evolution occurring at those ages. [10]. As for the cause of this, they hypothesized that it was due to the growth and restructure of neural pathways,
Van De Ville, Britz, and Michel, 2010 [3]
In a more recent and groundbreaking study done by researchers in Geneva, the temporal dynamics and possible fractal properties of EEG microstates were analyzed in normal human subjects. Since microstates are a global topography, but occur on such small time scales and change so rapidly, Van De Ville, Britz, and Michel hypothesized that these "atoms of thoughts" are fractal-like in the temporal dimension. That is, whether scaled up or scaled down, an EEG is itself a composition of microstates. This hypothesis was initially illuminated by the strong correlation between the rapid time scale and transience of EEG microstates and the much slower signals of a resting state fMRI.
This scale-invariant dynamic is the strongest characteristic of a fractal, and since microstates are indicative of global neuronal networks, it is justifiable to conclude that these microstates exhibit temporally monofractal (one-dimensional fractal) behavior. From here we can see the possibility that fMRI, which is also a global topography measure, is possibly just a scaled-up manifestation of its microstates, and thus further supports the hypothesis that EEG microstates are the fundamental unit of one's global cognitive processing.
Comparing normal humans' EEG Microstate classes to those of psychotic patients has yielded important results suggesting that the basic resting-state condition of these patients' brains are irregular. This implies that before any information is processed or created, it is bound to the dynamics of the irregular microstate sequencing. [1] [11] [12] [13] [14] [15] [16]. Although microstate analysis has great potential to help understand the basic mechanisms of some neurological diseases, there is still much work and understanding that needs to be developed before it can be a widely accepted diagnostic.[2]
Koenig & Lehmann, in their breakthrough 1999 study, looked at microstates of schizophrenics versus control patients. Schizophrenia is one of the main disorders studied with EEG Microstate analysis. The resting state EEGs were found to have aberrant microstate classes that lasted either too long or too short in comparison to analogous microstate classes of normal humans. [1] Schizophrenics' brains spend too much time in a topographically right-anterior to left-posterior (referred to as "Class A") microstate orientation, which maps to unfocused, frontal lobe-quiescent states, and too little time in focused frontal lobe-strong attentional states. Schizophrenic patients spent an average of 24.7% of the observed time in this Class A microstate, versus the controls who spent an average of 19% of observed time in Class A.[1] These inappropriate microstate durations occur intermittently in an otherwise normal microstate sequence. This further supports the theory that microstates are the basic steps of cognition, for if such a small-scale irregularity causes such a dramatic disorder. it must be a precursor for both basic and higher-level brain function. [1]. It is also noteworthy that schizophrenic-like behavior can be induced by manipulating a normal human's alpha wave frequency. [2]
In July 2011, Dr. Koenig collaborated with researchers from Kanazawa University in Japan, and others from the University of Bern in Switzerland, to do a microstate analysis on patients with Panic Disorder (PD). They found that these patients spent too much time in the same right-anterior to left-posterior microstate as in the schizophrenic studies.[17] This would suggest a temporal lobe malfunction, which has been reported in fMRI studies of patient with PD. The patients spent an average of 9.26 milliseconds longer in this microstate than did control subjects. These aberrant microstate sequences are very similar to those in the schizophrenia study, and as anxiety is the most prevalent symptom of schizophrenia, it may indicate a strong correlation between different severities of neurological pathologies and a patient's microstate sequence.
In 2004, Cantero, Atienza, Salas, and Gómez studied alpha rhythms in normal human subjects during 3 different drowsy/sleep states: eyes closed/relaxing, drowsiness at sleep onset, and REM sleep. They found that the mean determined microstate classes were different amongst consciousness states on 3 different parameters.[18]
This study yet again illuminates the complexity of brain activity and EEG dynamics. The data suggest that "alpha (wave) activity could be indexing different brain information in each arousal state."[18] Furthermore, they suggest that the alpha rhythm could be the "natural resonance frequency of the visual cortex during the waking state, whereas the alpha activity that appears in the drowsiness period at sleep onset could be indexing the hypnagogic imagery self-generated by the sleeping brain, and a phasic event in the case of REM sleep." [18]. Another claim is that longer periods of stable brain activity may be handling smaller amounts of information processing, and thus few changes in microstates, while shorter, less-stable brain activity may reflect large amounts of different information to process, and thus more microstate changes.