Pretreatment, Intratreatment, and Posttreatment EEG Imaging of EMDR: Methodology and Preliminary Results From a Single Case

Neuroimaging Studies of Posttraumatic Stress Disorder

PTSD research has extensively used autobiographical script-driven brain activation to identify the regions in which the reliving of the traumatic experience causes CBF and metabolic changes (Francati, Vermetten, & Bremner, 2007). This has been achieved by either comparing patients to a control group or measuring the activation parameters in different conditions (i.e., resting vs. traumatic script or traumatic script before and after therapy). Neuropsychological and neuropsychiatric tests have been used to assess the clinical status under different conditions. For example, neuropsychological tests and self-evaluation scales used by Högberg et al. (2008) showed immediate changes toward normality following eye movement desensitization and reprocessing (EMDR) treatment. These changes were still present at 3-year follow-up.

Studies performed with both SPECT and PET have reported regional CBF (rCBF) changes during trauma recall; both increases and decreases in rCBF were described (see Francati et al., 2007). Altered rCBF were mostly found in hippocampus, amygdala, orbitofrontal and frontal cortex, middle prefrontal and middle temporal cortex, anterior and posterior cingulated cortex, temporal poles, Broca’s area, and caudate. The critical involvement of the limbic system has been hypothesized to be connected to the fear-related stimuli and to the emotional responsiveness to the retrieved traumatic experience elicited by symptom provocation.

Neuroimaging Studies of EMDR Effects

Recently, the neurophysiological response to EMDR therapy has been investigated with consistent findings of an increased parasympathetic tone during bilateral stimulation in three different studies (Elofsson, von Schèele, Theorell, & Sondergaard, 2008; Sack, Lempa, & Lamprecht, 2007; Sack, Lempa, Steinmetz, Lamprecht, & Hofmann, 2008). In addition, some promising neuroimaging studies used SPECT (Lansing, Amen, Hanks, & Rudy 2005; Oh & Choi, 2007; Pagani et al., 2007) and MRI scans (Nardo et al., 2010) to investigate the neurobiological substrate of EMDR therapy in clinical practice. These studies were conducted to examine the effects of EMDR on brain pathophysiology and have provided some preliminary evidence that change in CBF patterns and in gray matter density may be linked to effective treatment. Generally, in PTSD and in anxiety disorders, functional deactivation of sensitive cortical areas is associated with decreased hyperreactivity to emotional and memory disturbances, resulting in symptom relief.

The first functional MRI (fMRI) investigation of real-time response of brain neurons during a single EMDR session was conducted by Richardson et al. in 2009, assessing regional blood oxygenation changes during therapy. The authors found consistent prefrontal cortex activation during auditory alternating bilateral stimulation. This study paved the way for a more exhaustive description of the dynamics of regional and hemispheric neuronal firing during psychotherapy.

Online monitoring of the human brain is limited by time resolution. For example, previous functional neuroimaging investigations of EMDR detected CBF changes over a range of minutes. The most advanced PET technology (which has not yet been used to investigate EMDR processing) can detect changes that occur over a range of several seconds. One of the tools that overcome such time limitations is electroencephalography (EEG) in which electrical activity resulting from neuronal activation is recorded, with a time resolution of milliseconds.

EEG was recently implemented in EMDR research by Harper, Rasolkhani-Kaòhorn, and Drozd (2009). They used a 19-channel unit and reported indirect evidence of depotentiation of the amygdala during therapy. Although they used an EMDR procedure with no therapist-directed eye movements, Harper and colleagues were not able to rule out possible spontaneous eye movements during the stimulation session in their EEG recordings.

Participant

The participant, a 43-year-old right-handed woman, referred spontaneously with symptoms of generalized anxiety disorder, which she had developed when her adolescent daughter started going out in the evening. She reported that her daughter’s behavior was triggering memories of her own childhood sexual abuse and presented with posttraumatic emotional disturbance, with significant impairment in daily activities. She also disclosed that she had experienced other multiple and cumulative traumas.

Study Design

After the initial history taking and preparation session, three more EMDR therapy sessions were provided. Pretreatment measures were taken during the second session and included five psychological self-report tests, pretreatment EEG recordings (script), and EEG recordings made during lateral eye movements in the desensitization phase. Posttreatment measures after successful processing of the targeted trauma included during the third (last) treatment session the readministration of psychological tests as well as EEG recordings during both script and lateral eye movements in the desensitization phase. The targeted trauma was the participant’s memory of childhood sexual abuse.

Assessment

EEG Procedure

EEG recordings were taken at pretreatment (Time 0, T0) at the beginning of the second session, and at posttreatment (Time 1, T1) at the beginning of the fourth (last) session, using the following sequence: (a) at rest with eyes open, (b) at rest with eyes closed, (c) listening to the script with eyes closed, and (d) at rest with eyes closed. EEG recordings were also taken in Sessions 2 and 4, during lateral eye movements in the desensitization phase of EMDR (see Figure 1 for experimental design).

FIGURE 1

Study design.

Preprocessing

Digital data recording was exported to European data format (EDF) from the native format using NPX Lab 2010 (available at: http://www.brainterface.com). Whereas the script recording was fully exported, in the EMDR recordings, we selected and exported only the eye movement periods of BS sessions (eliminating, arbitrarily, the first four and the last two eye movements), creating files with concatenated/merged periods of BSs. Data were analyzed in the EEGLAB environment ( a available at: http://www.sccn.ucsd.edu/eeglab/index.html), collection of scripts running under Matlab 7.7.0 R2010a (MathWorks, Inc., Natick, MA), digitally band-pass filtered between 1 and 45 Hz (with an IIR filter), and re-referenced to average reference. After visual inspection and manual elimination of paroxysmal artifact periods, artifactual noncerebral source activities (eye blinks and movements, cardiac and muscle/electromyographic activity) were identified and rejected using a semiautomatic procedure based on ICA (Barbati, Porcaro, Zappasodi, Rossini, & Tecchio, 2004; Porcaro et al., 2009). Briefly, the EEG signal was decomposed into independent components (ICs) using FastICA 2.5 (available at: http://www.cis.hut.fi/projects/ica/fastica; Hyvärinen & Oja, 2000). After removal of artifactual noncerebral ICs, the “cleaned” signal was reconstructed by retroprojecting all the ICs, except for artifactual ones. After downsampling to 128 Hz, we selected 128 consecutive seconds in script recordings and 180 consecutive seconds in EMDR recordings for next step of analysis.

Electrical Source Imaging

To compute the intracerebral electrical sources underlying EEG activity recorded at the scalp, we used the exact low resolution brain electromagnetic tomography (eLORETA) software (available at: http://www.uzh.ch/keyinst/loreta.htm). From the scalp-recorded electric potential distribution, eLORETA computes the cortical three-dimensional (3D) distribution of current density (i.e., the amount of electrical current flowing through a solid), without assuming any number of active sources. The eLORETA method is a discrete, 3D, distributed, linear, weighted minimum norm inverse solution (Pascual-Marqui, 2007). The particular weights used in eLORETA endow the tomography with the property of exact zero-error localization to test point sources, yielding images of current density with exact localization, although with low-spatial resolution (i.e., neighboring neuronal sources will be highly correlated). In the current implementation of eLORETA, computations were made in a realistic head model (Fuchs, Kastner, Wagner, Hawes, & Ebersole, 2002), using the MNI152 template with the 3D solution space (i.e., the locations in which sources can be found) restricted to cortical gray matter and hippocampi, as determined by the probabilistic Talairach atlas (Lancaster et al., 2000). The intracerebral volume (eLORETA inverse solution space) is partitioned in 6239 voxels at 5-mm spatial resolution (i.e., cubic elements of 5 × 5 × 5 mm). Thus, eLORETA images represent the electric activity at each voxel in neuroanatomical Montreal Neurological Institute (MNI; Montreal, Quebec, Canada) space as the exact magnitude of the estimated current density. Anatomical labels such as Brodmann areas (BAs), are also reported using MNI space, with correction to Talairach space (Brett, Johnsrude, & Owen, 2002). We calculated eLORETA images corresponding to the estimated neuronal generators of brain activity within each band (for details of LORETA frequency band analysis, see Frei et al., 2001). Frequency band ranges were delta (1.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta1 (12–20 Hz), beta2 (20–30 Hz), and gamma (30–45 Hz). Because all eLORETA inverse spatial solution voxels have a certain current density and for exploratory nature of the current single case analysis, we adopted two criteria to select maximum cortical activations: The first restriction was a Z-score voxel values >1.5 (i.e., only the values greater than 1.5 times the standard deviation of the standardized data [in the LORETA spatial solution] were accepted as being activated); the second strict constraint was a minimum number of 27 voxels (an intracerebral volume cube with an edge of 15 mm) with Z-score >1.5 for single BA in a hemisphere.

Session One

In the current study, in Phase 1, the therapist determined that the participant’s symptom presentation did not meet diagnostic criteria for any mental disorder, although she had symptoms of anxiety, depression, and posttraumatic stress. It was apparent that the core issue was the sexual abuse that the participant had experienced in childhood, and it was decided that this incident would have been targeted in EMDR treatment.

In Phase 2—preparation—EMDR was explained, and the participant gave informed consent. She also agreed to make, on site, a digital recording of the autobiographical script of her traumatic experience (childhood sexual abuse), for use with the EEG recordings. The safe place exercise was introduced to show the bilateral stimulation procedure, while urging her to focus on more positive recollections, feelings, emotions, and physical sensations.

Session Two

The targeted memory was of childhood sexual abuse by her brother’s friend. During processing, the participant reexperienced physical sensations such as the taste of paper, which the abuser had used to gag and silence her. Another important and highly distressing fragment of the experience, recalled during processing, was her mother’s disbelief when the participant had tried to tell her of the abuse. After processing reduced the abreaction elicited by that fragment, she recalled many memories of physical violence at the hands of her father. After 45 minutes, the participant was able to think of the targeted sexual abuse memory with less emotional and physical discomfort, although the processing was still incomplete. In this session, the therapist carried out 19 sets of bilateral visual stimulation.

Session Three

In the following session, processing about the sexual abuse continued until the participant reported no distress with a SUD score of 0, VOC score of 7, and a clear body scan. The rest of the session focused on processing other memories related to her father.

Session Four

In Session 4, the target was current triggers, such as meeting the perpetrator and her fear that her daughter will be hurt when she goes out in the evening (she fears her daughter could be hurt). A future template was done on her daughter’s going out downtown in the evening. These issues were successfully resolved. In this session, 15 sets of bilateral visual stimulation of about 30 seconds were conducted.

Psychometric Measures

There was a clear decrease in the participant’s IES score, as assessed after EMDR treatment. In fact, the client moved from a clinical pretreatment score of 43 (at the high end of the moderate range) to a nonclinical posttreatment score of 7. This was caused by a dramatic reduction, both within the Intrusion (from 20 to 4) and within the Avoidance (from 23 to 3) subscales. These changes indicate that the participant no longer experienced the targeted event as disturbing. Neuropsychologically speaking, this result is suggestive of a thorough effect of EMDR treatment on both components of posttraumatic lasting consequences.

The participant’s BDI scores did not change with treatment, remaining in a clinical range with a pretreatment level of 27 and a posttreatment level of 28 (BDI scores higher than 18 indicate moderate-to-severe depressive symptoms). There were also no changes with regard to symptoms of psychopathology, as measured by the SCL-90 R. Scores on the global severity (from 1.05 to 1.11), positive symptom total (from 63 to 61), and positive symptom distress (from 2.03 to 1.95) indexes indicate no change in this respect. In the same line, moderate PTGI overall scores remained quite stable (from 47 to 41), showing no significant changes in the client self-perception of posttraumatic growth.

EEG Recordings

EEG recordings were made at pretreatment in Session 2 (T0) and at posttreatment in Session 4 (T1). After having averaged the EEG signals and considering the whole spectrum (1.5–44 Hz), the analysis of the maximally activated regions during the script at T0 highlighted the bilateral prefrontal cortex, and regions of the left parieto-occipital cortex (Table 1, Figure 2) as the ones dominantly activated during trauma reliving. This pattern was quite similar to the one observed in the delta (1.5–4 Hz), beta1 (12–20 Hz), and beta2 (20–30 Hz) bands.

FIGURE 2

During script at pretreatment presentation: Cortical representation of the clusters of voxels in which the EEG signal exceeds a Z-Score of 1.5.

Note. Z-score increases are depicted by shifts of the color scale from the red range (small increase) toward the yellow range (large increase).

The same analysis at T1 showed significantly activated clusters in the left occipital cortex and in the right temporal cortex (Table 2, Figure 3). This latter finding was present in all bands with the exception of delta and theta (4–8 Hz).

FIGURE 3

During script at pretreatment presentation: Cortical representation of the clusters of voxels in which the EEG signal exceeds a Z-Score of 1.5.

Note. Z-score increases are depicted by shifts of the color scale from the red range (small increase) toward the yellow range (large increase).

When analyzing all bands individually, and the whole spectrum, there was a concordant significant dominant cortical activation during the desensitization phase at first EMDR session (T0) in the bilateral limbic prefrontal cortex and in the lateral orbitofrontal cortex on the right (Table 3, Figure 4). During the last EMDR session (T1), the activation in the bilateral prefrontal cortex was still present but largely extended to the left temporo-occipital cortex. This pattern occurred consistently in beta1, beta2, and gamma (30–44 Hz) bands (Table 4, Figure 5). The specific activation in the lateral orbitofrontal cortex shifted from the right to the left hemisphere in beta1, beta2, and gamma bands.

FIGURE 4

In first EMDR session, during bilateral visual stimulation in the desensitization phase: Cortical representation of the clusters of voxels in which the EEG signal exceeds a Z-Score of 1.5.

Note. Z-score increases are depicted by shifts of the color scale from the red range (small increase) toward the yellow range (large increase).

FIGURE 5

In last EMDR session, during bilateral visual stimulation in the desensitization phase: Cortical representation of the clusters of voxels in which the EEG signal exceeds a Z-Score of 1.5.

Note. Z-score increases are depicted by shifts of the color scale from the red range (small increase) toward the yellow range (large increase).

Limitations

In this preliminary investigation, we have limited the spectral analyses to the localization of the areas exhibiting field powers exceeding the critical threshold of Z-score voxel values >1.5. Investigating a single case prevented deeper and more sophisticated analyses of the EEG signal because large interindividual differences might exist in signal amplitudes and in the dominant topographic location of each band (Chen, Feng, Zhao, Yin, & Wang, 2008). Furthermore, because of the poor statistical power of investigating a single case, the correlation analyses between bands power would not be reliable, increasing the risk of random findings. Finer and more exhaustive statistical analyses and relative inferences will be possible when the numerousness of the experimental sample becomes larger.

It has been argued that the EEG signal can be the result of either an inhibitory or an excitatory neuronal signal and might not be directly related to an increase or decrease of metabolism or blood flow, especially for the low-frequencies bands (i.e., theta and alpha1; Oakes et al., 2004). However, the preliminary results presented in this study show a fairly homogeneous behavior of all bands in response to the psychological stimulation elicited by both script listening and reprocessing during EMDR’s desensitization phase (see Tables 1–TABLE 2TABLE 34). More reliable and deep analyses of amplitude, distribution, and coherence within and between single bands in both states will be possible when the numerousness of experimental subjects increases.

Theories Related to Our Findings

This study showed that the processing of the traumatic event was followed by a clear dominant electrical activity of the lateral prefrontal cortex and less activation in the prefrontal cortex limbic system. This neurobiological change provides support for several theories and future research is needed to better determine which theory best accounts for these findings.

Considerations Regarding Frequency Bands

In this study, fast frequency bands (beta2 and gamma, see Tables 2 and 4) contributed significantly in determining the whole spectrum pattern in both script listening and EMDR’s desensitization phase. Their dominant activation in the temporo-occipital cortex following successful therapy is in accordance with their prevalence in enhanced arousal states.

Selective attention is processed in areas prone to fast waves, resulting in beta rhythms being often enhanced throughout the brain, with the engagement of “cognitive-type” cortical circuitry. Consistently, Tallon-Baudry, Mandon, Freiwald, and Kreiter (2004) observed an increased activity in the beta-frequency range in inferior temporal cortex during a working memory task. Working memory effects have also been strongly implicated as one of the mechanisms associated with the dual-attention task of eye movements in EMDR (see van den Hout et al., 2010). These mechanisms may underlie the dominant activity of beta waves in associative areas after trauma elaboration. It is our opinion that one of the markers of successful EMDR therapy is that the experience of reliving changes from being emotional to being cognitive in nature, and we hypothesize that this is associated with increased fast bands (beta and gamma) activity in the left hemisphere—an effect which we found in this study.

Furthermore, beta oscillations support functional cortical networking (Lopes da Silva & Pfurtscheller, 1999; Neuper & Pfurtscheller, 2001) as well as cognitive functions coordinating distributed neural responses, such as attention-dependent stimulus selection, working memory, and consciousness. Schnitzler and Gross (2005), revising the neural networks underlying attention control in visual processing, concluded that communication within the fronto-parieto-temporal attentional network proceeds via transient long-range phase synchronization in the beta-band. In a recent review, Engel and Fries (2010) correlated beta band activity to the maintenance of the current cognitive control and to the strong endogenous top–down influence overriding the effect of unexpected external events. All such evidence underscores the role of beta rhythms in the neurophysiology of cognitive control and networking, possibly making them as one of the major determinants of the efficacy of EMDR in brain processing normalization.