Cortical reorganization occurs in the adult central nervous system, especially during motor skill acquisition. Plasticity contributes to various forms of human behavior including skill learning and memory formation, consolidation, reconsolidation and short- and long-term retention. It is very important to understand the role of these different behavioral processes and of the mechanisms underlying these various forms of human plasticity during skill acquisition. Findings this year: Over the past year, we have advanced several research initiatives aimed at characterizing intra- and inter-individual variability in responses to non-invasive brain stimulation. The rationale for this work is that a more nuanced understanding of these sources of variability is crucial for the development of personalized neuromodulatory protocols using non-invasive brain stimulation into effective therapies for neurorehabilitation in individual patients. Human corticospinal excitability show substantial inter-trial variability, which may be partially due to endogenous brain oscillatory activity at the time of TMS delivery. For example, cortical excitability co-varies with frequency-specific oscillatory phase and power in occipital and motor areas (Romei et al. 2008a, 2008b; Sauseng et al. 2009; Berger et al. 2012; Bergmann et al. 2012). Invasive intracortical recordings showed that neuronal spiking is time-locked to the endogenous oscillatory phase, and that this locking is stronger when power is high compared to low (Haegens et al. 2011; Miller et al. 2012). However, how oscillatory power and phase interact to influence corticospinal excitability in intact humans is not well understood. Here, we explored the relationship between TMS-elicited motor evoked potential (MEP) amplitudes, instantaneous phase and instantaneous power of endogenous sensorimotor cortical oscillations at the time of TMS delivery (N=20). 28-channel EEG was recorded during collection of MEPs following TMS to the primary motor cortex (M1) hotspot for the left first dorsal interosseous (FDI) muscle at 120% of resting motor threshold (RMT). Given their involvement in motor control, we focused on alpha (8-12 Hz) and beta (13-30 Hz) oscillations. We determined the instantaneous phase and power of alpha and beta oscillations at the C4 electrode 2ms before each TMS pulse. Trials were separated based on power (high or low) and the phase during which each TMS pulse occurred (peak or trough) for each frequency band. MEPs were larger when TMS was delivered during alpha troughs relative to peaks, but only when alpha power was high (PHASE x POWER interaction: p=0.042). The difference between alpha trough and peak MEP amplitude was 0.165 0.07 and -0.017 0.04 mV for high and low power trials, respectively. For high power trials, this amounted to 12% difference in MEP amplitude between alpha trough versus peak MEPs. We found no clear relationship between oscillatory phase, power, and MEP amplitude for the beta band (PHASE x POWER interaction: p=0.446). The difference between beta trough and peak MEPs was -0.056 0.07 and 0.032 0.07 mV for high and low power trials, respectively. Our results suggest that instantaneous alpha oscillatory phase and power interact to influence corticospinal excitability in intact humans. These results raise the possibility that interactions between alpha power and phase influence TMS effects to an extent not yet recognized in cognitive and interventional neuroscience. The introduction of neuronavigation has refined TMS protocols by providing information on coil position and orientation at the time stimulation is delivered (Sparing et al., 2007, Pellicciari et al., 2016). However, TMS effects remain substantially variable within and across subjects. We also developed a new multivariate deconvolution modeling approach to estimate the impact of TMS coil position and orientation variability on TMS-induced effects. Participants (N=19) were seated in a chair at rest during the experiment. The hotspot for the left FDI was determined, and the resting motor threshold (RMT) was obtained. A total of 600 TMS stimuli at 120% RMT were delivered with concurrent acquisition of 28-channel EEG (Brain Products) and TMS coil position and orientation data using Brainsight 2 (Rogue Research). Deconvolution was performed under the Ordinary Least Squares (OLS) framework, with MEP amplitudes as dependent variables and orthogonal factors combining 6-D coil displacement and orientation information using Singular Value Decomposition (SVD). We then fit a linear or quadratic model depending on the difference between the coefficients of determination. SVD factors across subjects consistently captured how the operator used the online visual feedback provided by the neuronavigation system to correct for positioning and orientation error. Factors 1 and 2 accounted for over 80% of the coil data variance across subjects and loaded predominantly on the coil displacement dimensions, while Factors 3 and 4 loaded mostly on orientation errors. The method was able to decorrelate MEP amplitudes from SVD coil factors while preserving their original distributions. We then tested our deconvolution approach on previously reported brain state-dependent MEP amplitude effects (TMS stimuli delivered at EEG alpha-band phase peaks vs. troughs). Deconvolution led to increased effect sizes (Cohens d MEPtrough - MEPpeak amplitudes) in 11 out of 19 subjects. Inclusion of a regressor to account for drifts in MEP amplitude (Pellicciari et al, 2016) resulted in increased effect size for 13 of 19 subjects. Modeling trial-by-trial variability induced by coil angle, tilt and displacement as experimental factors increased measurement sensitivity of TMS effects on brain-state dependent corticospinal excitability. Thus, reduction of variability introduced by experimental factors could allow more accurate determination of TMS effects in therapeutic interventions. Another promising avenue of research that advances towards minimizing interindividual differences in patients responses to NIBS integrates brain imaging data, particularly functional and structural brain connectivity, to provide personalized, guided stimulation in clinical contexts (Fox et al., 2013; Dunlop et al., 2016; Otal et al., 2016). However, integrating imaging datasets into neuronavigation systems while simultaneously searching for the optimal stimulation site(s) in the clinic is a time-consuming process which requires operator training and may not be well-tolerated by some neurological and neuropsychiatric patient populations. Our group recently proposed a strategy for optimizing operator training and allowing for individualized, pre-intervention planning offline, which does not require the patients participation. Here, we used invividual subject MRI scans to generate 3D printed physical head models. These models and anatomical and functional MRI data were then integrated within a commercially-available neuronavigation system. We demonstrated an approach that will allow clinicians and operators of NIBS technologies to conduct extensive neuronavigation-based, pre-treatment planning sessions as needed for each individual subject. This approach will allow integration of multiple imaging modalities into the neuronavigation environment of an individual subject without prolonging the patients treatment time in the clinic. This technique may also be suitable for training purposes. Neuronavigation sessions in which multiple imaging modalities are integrated and the optimal stimulation trajectory is determined may last 45 min to an hour (and longer in the context of training). Thus, the approach described here may ideally facilitate individualized neuronavigation in frail populations that may not otherwise tolerate lengthy pre-treatment procedures.
Showing the most recent 10 out of 88 publications
