Transcranial alternate current stimulation approaches (tACS)
Transcranial Current Stimulation (tCS) is achieved by circulating a low intensity current (1-2 mA, ~0.06 mA/cm2) between at least two electrodes (an anode and a cathode) placed on specific regions of the human scalp (Fig. 3A & B). A substantial portion of the circulating current is generally shunted through the scalp skin (Vöröslakos et al., 2018). Nonetheless part of it will penetrate across the different tissue layers between the skin and the cortical surface (i.e., bone outer and inner tables, and the cerebrospinal fluid cumulated in the epidural and subdural spaces) to reach the pia-mater and spread across rather large cortical areas located between both electrodes (Miranda et al. 2006).
The current gradients will polarize electrical charges in the extracellular space in a polarity dependent manner, shifting the resting membrane potential of exposed neurons closer (anodal stimulation) or away (cathodal stimulation) from their firing thresholds, hence increasing or decreasing their probability to generate an action potential when receiving physiological dendritic inputs of sufficient intensity.
If instead of a constant current (tCS modality know as transcranial direct current stimulation or tDCS), an alternating current (AC) is applied, the resting membrane potential and consequently the firing rate probability of neurons influenced by the current field will also fluctuate periodically, following the frequency of the AC signal. This specific modality of tCS is referred to as transcranial Alternating Current Stimulation (tACS) and has been used to non-invasively entrain oscillations in cortical regions (Fröhlich & McCormick, 2010; Herrmann et al., 2013; Merlet et al., 2013).
Although tCS devices delivering either tDCS or tACS are recognized as being portable and highly affordable compared to TMS (Fig. 3A & C), these technologies possess a rather poor spatial resolution. Given the diversity of possible electrode montages (particularly when density tCS approaches based on combination of several return electrodes in complex configurations are used) and interindividual differences in head anatomical features, it is not easy to predict how currents applied to the scalp will diffuse transcranially to reach the cortical surface. Indeed, it is generally accepted that induced brain currents will not remain restricted to cortical areas beneath the electrodes but will spread (Bikson et al., 2010; Datta et al., 2012).
Rhythmic transcranial magnetic stimulation approaches (TMS)
TMS is currently the most established non-invasive technology used to activate clusters of neurons responsible for specific behaviors within a rather circumscribed cortical area (estimated ~12-15 mm radius) in healthy humans and patients.
TMS equipment consists in capacitators which charge and store electrical current, which is then briefly circulated (120 to 250 µs) through a stimulation coil (the most commonly used are figure-of-eight coils) made of two contiguous loops of copper wire encapsulated in butterfly shape protective case (Fig. 3C). Following the principles of electromagnetic induction discovered in 1831 by Michael Faraday, the circulation of the high-intensity current generates a brief and rapidly changing magnetic field, called a pulse, which distributes perpendicular to the surface of the TMS coil lying flat on the scalp. Thanks to the electromagnetic induction phenomenon, the magnetic field penetrates painlessly, and with very little distortion, the skull bone and the epidural and subdural spaces filled with CSF to reach the cortex under the coil and induce a current intracranially which will cause the depolarization of clusters of excitable neurons (Hallett, 2007; Kobayashi & Pascual-Leone, 2003) hosted within a focal area of 12-15 mm radius (see Valero-Cabré et al., 2005 for an estimation in animals models). To achieve its effect, the TMS coil is placed on the scalp region most closely overlying a given cortical target (i.e. the one enabling the shortest straight path to cortical target) using a frameless stereotaxic MRI-based neuronavigation system customized to the anatomy of each healthy participant or patient (Fig. 3D).
Moreover, thanks to its excellent temporal resolution (Hallett, 2007), TMS allows single pulses or multi-pulse bursts arranged in a great variety of patterns to be used in online trial-by-trial designs to impact specific time windows during the performance of behavioral tasks (for recent reviews see Polanía et al., 2018 or Valero-Cabré et al., 2017). Likewise, long patterns of so called repetitive TMS (or rTMS) can induce, depending on stimulation parameters (essentially, stimulation frequency, pattern duration and number of pulses, magnetic field intensity and length of inter-burst intervals), excitatory or inhibitory offline modulations of neural activity and associated behaviors, which remain transiently active beyond the discontinuation of pulses.
More interesting for the experimental work presented in this dissertation, either single pulses or, more efficiently, short episodes of the so-called rhythmic TMS (a modality of rTMS delivering short bursts of 4-5 regularly spaced TMS pulses) are being used to manipulate cortical oscillations within a targeted region. The first published precedent using TMS to manipulate ongoing oscillations used the ability of single isolated TMS pulses to phase-reset and synchronize local oscillators operating at the so called ‘natural frequency’ of the region. Such an approach has been applied to induce transient increases of oscillation amplitude in several cortical regions (Paus et al. 2001; Rosanova et al. 2009; Van Der Werf and Paus 2006). Some years thereafter, Thut and colleagues (2011a) put forward the notion that cortical populations of neurons consist in several oscillators, all fluctuating independently at an identical frequency but with a random phase (Fig. 4A). Given their rather natural desynchronized state in awake individuals, their summed spatio-temporal activity patterns tend to cancel off, and scalp EEG or MEG recordings prove unable to reveal clear signs of oscillations with a meaningful amplitude or increases of oscillatory power density in time-frequency analyses.
Rhythmic Transcranial Magnetic Stimulation in attentional and visual behaviors
The emergence of rhythmic TMS patterns to manipulate cortical oscillations set the stage to add causal evidence to correlational outcomes linking oscillations, attentional orienting and perceptual modulations (see Section I of this introduction). Rhythmic TMS bursts delivered online during visual detection tasks confirmed a well-documented (Dugué et al., 201; Mathewson et al., 2009; Thut et al., 2006; van Dijk et al., 2008) role of occipital and parietal alpha oscillations (Romei et al. 2010) and the preferred posterior parietal alpha phase (Jaegle & Ro, 2014) in the modulation of visual detection. This same approach revealed a double dissociation between the role of theta and beta frequencies over the intraparietal sulcus for the perception of global vs. local object features (Romei et al. 2011) previously identified in a correlational study (Smith et al. 2006).
Extending a prior study using single-pulse TMS to prove the causal role of the right FEF in conscious visual detection (Chanes et al., 2012), our team used rhythmic TMS in humans to explore prior correlational intracranial EEG evidence from monkeys highlighting the multiplexing of high-beta vs. gamma rhythms across the same fronto-parietal network to engage endogenous vs. exogenous orienting of attention (Buschman & Miller, 2007) or other correlational evidence in the human brain for a role of fronto-parietal high-beta synchronization in attention and conscious visual perception (Gross et al., 2004; Hipp et al., 2011; Phillips & Takeda, 2009).
Work by our lab employed trial-by-trial bursts of rhythmic TMS at two distinct frequencies (30 Hz vs. 50 Hz) delivered on the right FEF while participants performed a near-threshold visual detection task (Chanes et al., 2013). The results confirmed in humans previous correlational monkey work (Buschman & Miller, 2007) exploring the roles for high-beta and gamma band activity in an homologue frontal cortical region, the right FEF, and revealed that the episodic entrainment of these frequencies prior to target onset modulated different aspect of a conscious perception paradigm; gamma entrainment (active vs sham 4 TMS pulses at 50 Hz, compared to a non-uniform fixed pattern) decreased response bias (rendering participants less conservative in indicating they had seen a target when in doubt) whereas high-beta oscillations (active vs sham 4 TMS pulses at 30 Hz compared to a non-uniform fixed pattern) increased visual sensitivity (i.e., boosted the capacity to differentiate the presence of a visual target compared to a no-target noise condition).
Given the long-proven network distribution of focally applied TMS effects (Chouinard et al., 2003; Paus et al., 1997; Valero-Cabré et al., 2005), modulations of cortical activity by TMS cannot be reasonably expected to stay confined to the targeted region. As no brain region works in isolation, but rather as nodes linked to complex systems, TMS induced activity spreads to other associated network sites depending on the richness and strength of the connectivity. Consequently, a single pulse delivered focally on a cortical region will phase-reset local oscillators and increase the amplitude of rhythmic activity in sites distant from the TMS target region (Rosanova et al., 2009). Similarly, rhythmic TMS delivered to frontal regions such as the FEF could have an effect on the topography of synchronization/desynchronization of alpha activity recorded in parieto-occipital areas (Capotosto et al., 2009).
Therefore, local rhythmic TMS can be used to probe a causal contribution of frequency specific cortical oscillations to a task, nonetheless, their use with concurrent mapping technologies sensitive to neural spatio-temporal dynamics such as EEG can serve to monitor its influence on extended neural systems the stimulated region is part of. To this regard, two TMS behavioral studies analyzed individual diffusion imaging tractography datasets (Quentin et al., 2014, 2015) and reported significant correlations between the facilitatory impact of high-beta right frontal rhythmic TMS stimulation on conscious visual detection and white matter connectivity estimates of the 1st branch of the right superior Longitudinal Fasciculus (SLF), linking the FEF and the Intraparietal Sulcus (IPS) and subtending the dorsal attentional network (Thiebaut de Schotten et al., 2011). In conclusion, converging with correlational evidence in monkeys and humans, EEG experiments and theoretical models of oscillatory perturbations, rhythmic TMS has demonstrated an ability to manipulate cortical oscillations. Such effects have been proven to be TMS-frequency dependent, to primarily enhance power of intrinsic frequencies in the tested area, and active during pulse delivery but short lasting thereafter (no longer than two cycles). Moreover, in the attentional and visual domain, TMS alone or in conjunction with concurrent EEG recordings has been paramount to explore the causal role of episodic high-beta and gamma right frontal rhythms to enable visuo-spatial orienting leading to visual detection improvements. Further research using similar causal interventional approach will help build a more comprehensive picture of how other brain regions and frequency bands might contribute to the modulation of attention and visual perception in healthy humans. In parallel, an emerging research field is attempting the translation of oscillatory manipulation principles to effective treatments for the rehabilitation of cognition in human neurological patients.
Rhythmic Transcranial Magnetic Stimulation in neuropsychiatric rehabilitation
Beyond their use in experimental science to probe causal relationships in the brain, non-invasive stimulation approaches have shown promise as treatments for patients with altered cognition. Brief rhythmic TMS bursts applied for exploratory purposes induced short-lasting effects that are essentially restricted to the duration of the stimulation train (and one to two cycles beyond for the frequency of interest). This limitation makes them suited for trial-by-trial exploratory studies in cognition, operating episodically in relatively shorty time windows prior or during task events, on the other hand it does not enable longer-lasting modulations of neural rhythms that could be used for therapeutic purposes. It remains, however, controversial if the use of long TMS stimulation patterns may be able to operate beyond the discontinuation of the stimulation trains and modulate ongoing oscillations or network synchrony in a way that can be predicted according to input parameters (reviewed in Polanía et al., 2018; Valero-Cabré et al., 2017). Such longer lasting effects are paramount to support future uses of rhythmic TMS stimulation to correct cognitive symptoms linked to abnormal oscillations or synchrony patterns in the context of neuropsychiatric diseases.
Repetitive TMS (or rTMS), consisting in series of pulses or bursts tested in a wide range of low or high frequencies (usually conventional rTMS at 1, 3, 5, 10 or 20 Hz, or patterned TMS such as continuous (cTBS) or intermittent (iTBS) theta burst, which consists in bursts of 3 pulses at 50 Hz repeated every 200 ms, i.e. at 5 Hz, for longer periods of time, from 30 seconds to 30 min) have been shown to lastingly modulate measures of motor cortico-spinal excitability (Gangitano et al., 2002; Huang et al., 2005; Maeda et al., 2000; Pascual-Leone et al., 1994), visual evoked potentials (Aydin-Abidin et al., 2006; Thut et al., 2003) and, most relevant for this thesis, offline cortical oscillatory activity (Chen et al., 2003; Schindler et al., 2008; Strens et al., 2002; Thut et al., 2003; Woźniak-Kwaśniewska et al., 2014).
Generally speaking, neurophysiological offline or after-effects have been proven to last for up to 30 min (see Thut & Pascual-Leone, 2009 for a review), depending on stimulation site and TMS pattern specification (frequency, type, number of pulses or bursts, duration etc.). For conventional rTMS (pure frequencies between 1 to 20 Hz), a commonly applied rule of thumb has established that after-effects are effective for a period of time which is ~50% of the pattern duration. Patterned TMS (cTBS and iTBS) have been found to induce long lasting after-effects of 20 to 60 minutes, after a pattern lasting 20-190 seconds in the cortico-spinal tract (Huang et al., 2005). The accrual of rTMS daily sessions repeated at intervals of <24 hours showed a potential to induce even longer-lasting effects (Bäumer et al., 2003; Maeda et al., 2000) and backed up promises of therapeutically meaningful outcomes in neuropsychiatric diseases.
Stimulation regimes based on daily rTMS sessions protocols tested in multicentric clinical trials have been approved by the US Food and Drug Administration (FDA) to treat medication-resistant depression (George et al., 2010; O’Reardon et al., 2007), which is one of the most clinically established application for rTMS. Likewise, pre-clinical and clinical rTMS studies are currently being conducted for the treatment of Obsessive-Compulsive Disorder (Dunlop et al., 2016) or positive symptoms of schizophrenia such as auditory verbal hallucinations (reviewed in Thomas et al., 2016, see Thomas et al., 2019).
Neural noise, stochastic resonance and the modulation of visual perception
In prior sections of this introduction, we have extensively reviewed existing evidence supporting a behavior-specific role of local oscillatory activity and interregional synchrony; with a special focus on anatomical systems and frequency-specific coding strategies subtending attentional orienting and perception. We have highlighted (and shown experimental evidence of) how the emergence of frequency-specific oscillations, a highly predictable, regular and synchronous fluctuation of activity, enabled a most efficient processing of top-down attentional allocation or bottom-up saliency, facilitating perception (Fries, 2005, 2009; Singer & Gray, 1995; Tallon-Baudry & Bertrand, 1999). Despite such a strong focus in brain oscillations and network synchrony as one of the fundamental pillars of neural coding, research has also inquired about the potential role of neural noise (i.e. a highly random and unpredictable neural signal) in the modulation of cognitive processing leading to specific behaviors. One of the most influent concepts to this regard lies at the core of the so-called Stochastic Resonance Theory. This well-established framework in signal processing has theorized and shown experimentally that the addition of an optimal level of sensory noise (i.e., not too high, not too low) to a weak signal can unexpectedly enhance stimulus saliency hence its ability to be perceived (detected or discriminated) instead of blurring it.
To this regard, regular frequency-specific oscillations and neural noise, two different and in some ways opposite patterns of cortical activity produced by joint or segregated neural systems, may both be called to show a complementary role in neural coding strategies, through different mechanisms, and contribute together to the modulation of local and network synchronization events. Some indirect hints challenging the notion that increases of oscillation power or synchrony is necessarily favorable to efficient coding of cognitive operations and ultimately lead to better performance can be found in neuropsychiatric diseases. This evidence would leave some room for neural models of cognitive coding in which systems generating controlled levels of random neural noise would be as important as local and long-range neural networks able to generate episodic brain rhythms.
Cognitive impairments associated to abnormal oscillations and synchrony
The evidence reviewed in the preceding sections of this introduction seems to strongly support the notion that the lack of cortical oscillations are in general considered to cause pathological states and explain cognitive impairment (Grice et al., 2001; Uhlhaas et al., 2006).
Nonetheless, neurophysiological evidence in well-known neuropsychiatric diseases also suggest a detrimental role for cortical oscillations or excessive frequency-specific synchrony. For example, in Parkinson patients an excess of high-beta synchronization in cortical motor system and basal ganglia loops slows movements and increases rigidity (Witcher et al., 2014). In epilepsy, abnormally high levels of local gamma synchronization often precedes the onset of a seizure, whereas generalized epilepsy can be characterized by very high levels of gamma synchronization throughout large cortical regions, leading to impaired behavior and loss of consciousness (Uhlhaas & Singer, 2006). Therefore, although synchronization subtends the activation of cognitive operations and behavioral facilitation, the desynchronization of neural activity (i.e. preventing the build-up of temporally synchronized activity in specific frequency bands) can also be the necessary condition to avoid signs of pathology.
In hemispatial neglect, an attentional awareness disorder impairing patient’s ability to orient attention hence consciously detect, localize or discriminates perceptual events occurring the left hemispace or hemibody (Bartolomeo, 2007), strongly synchronized left frontal beta oscillations (~13-14 Hz) prior to target onset correlates with trials in which this right hemisphere stroke patients omit visual targets displayed in the left visual hemifield (Rastelli et al., 2013). Conversely, in healthy participants, a pre-target onset desynchronization of left frontal beta activity has been associated to successful ability to anticipate the appearance of a target in a contingent negative variation paradigm (Gómez et al., 2006). Similarly, the levels of left frontal beta-band desynchronization predicted the detection of supra-threshold somatosensory stimuli during a backward masking paradigm (Schubert et al., 2009). A desynchronization of alpha rhythms is also observed in occipito-parietal cortex contralateral to the attended visual hemifield in spatial cueing paradigms (Sauseng et al., 2005; Thut et al., 2006; Worden et al., 2000) and the level of pre-stimulus alpha power is negatively correlated with visual detection (Mathewson et al., 2009) and discrimination performances (Hanslmayr et al., 2007). Such findings have led to the hypothesis that alpha oscillations suppress processing of sensory stimuli (reviewed in Foxe & Snyder, 2011).
Table of contents :
TABLE OF ABBREVIATIONS
I – Brain oscillations, local and network synchronization and orienting spatial attention .
I.1 – Oscillations and synchronization in network communication and information transfer
I.2 – Network synchronization subtending visuo-spatial attention and visual perception
II – Manipulation of brain oscillations subtending attentional and visual behaviors
II.1 – Non-invasive stimulation techniques to manipulate brain oscillations and synchrony
II.1.1 – Rhythmic peripheral sensory stimulation for oscillatory entrainment
II.1.2 – Transcranial brain stimulation technologies for oscillatory entrainment
II.2 – Rhythmic Transcranial Magnetic Stimulation in attentional and visual behaviors
II.3 – Rhythmic Transcranial Magnetic Stimulation in neuropsychiatric rehabilitation
III – Neural noise, stochastic resonance and the modulation of visual perception
III.1 – Cognitive impairments associated to abnormal oscillations and synchrony
III.2 – Stochastic Resonance Theory, modulation of neural coding and information processing
III.3 – Neural noise and Stochastic Resonance in the modulation of perception
I – Behavioral paradigm to assess visual performance
I.1 – Near-threshold lateralized visual detection paradigm
I.2 – Visual target properties, features and titration procedures
I.3 – Experimental blocks and session organization
I.4 – Subjective and objective measures of perception
I.5 – Signal Detection Theory and visual performance outcome measures
II – Transcranial Magnetic Stimulation
II.1 – Stimulation parameters
II.2 – Design of rhythmic and random TMS patterns
II.3 – Cortical target selection and MRI-based frameless neuronavigation
III – Concurrent TMS-EEG recordings of brain activity
III.1 – Electromagnetic TMS-EEG artifact removal and data cleaning procedures
III.2 – Concurrent TMS-EEG recordings and EEG data pre-processing
III.3 – Control analysis on the TMS-EEG artifact removal and data cleaning procedures
III.4 – Outcome measures to assess the impact of TMS on oscillatory activity
III.4.1 – Outcome measures for local oscillatory activity
III.4.2 – Outcome measures for inter-regional network synchronization
III.5 Outcome measures to quantify and characterize noise in EEG datasets
III.5.1. Measures to characterize noise in the time-frequency domain
III.5.2. Measures to characterize noise in the time domain
III.6 – Cluster-based permutation tests for the correction of multiple comparisons
PROJECT 1: Causal role of high-beta oscillations in the right fronto-parietal network for conscious visual detection
I – Entrainment of local synchrony reveals a causal role for high-beta right frontal oscillations in human visual consciousness
II – Causal role of high-beta right fronto-parietal synchrony in the modulation of human conscious visual perception
PROJECT 2: Exploring unexpected contributions of left frontal neural noise to the modulation of conscious visual perception in the human brain: a combined TMS-EEG study
PROJECT 3: Non-specific effects of auditory stimulation generated by transcranial magnetic stimulation (TMS) on cortical oscillations and visual detection performances
I – Summary of the main results
II – Frontal and fronto-parietal contributions to the modulation of visual perception
II.1 – Interhemispheric asymmetries in top-down systems for the facilitation of visual performance
II.2 – Methodological limitations of our datasets and experimental approaches
II.3 – Modulating visuo-spatial attention and recording conscious visual perception
III- Pending questions and some future directions
III.1 – Towards an oscillatory model of attentional orienting and perceptual modulation
III.2 – Contributions of parietal and occipital cortices to conscious perception
IV- Further considerations
IV.1 – Unexpected impact of ‘control’ TMS patterns on EEG activity
IV.2 – Network impact and state dependency of frequency-tailored TMS effects
V- Conclusion and final remarks