Brain responsivity provides an individual readout for motor recovery after stroke

Abstract

Promoting the recovery of motor function and optimizing rehabilitation strategies for stroke patients is closely associated with the challenge of individual prediction. To date, stroke research has identified critical pathophysiological neural underpinnings at the cellular level as well as with regard to network reorganization. However, in order to generate reliable readouts at the level of individual patients and thereby realize translation from bench to bedside, we are still in a need for innovative methods. The combined use of transcranial magnetic stimulation (TMS) and EEG has proven powerful to record both local and network responses at an individual's level. To elucidate the potential of TMS-EEG to assess motor recovery after stroke, we used neuronavigated TMS-EEG over ipsilesional primary motor cortex (M1) in 28 stroke patients in the first days after stroke. Twenty-five of these patients were reassessed after >3 months post-stroke. In the early post-stroke phase (6.7 ± 2.5 days), the TMS-evoked EEG responses featured two markedly different response morphologies upon TMS to ipsilesional M1. In the first group of patients, TMS elicited a differentiated and sustained EEG response with a series of deflections sequentially involving both hemispheres. This response type resembled the patterns of bilateral activation as observed in the healthy comparison group. By contrast, in a subgroup of severely affected patients, TMS evoked a slow and simplified local response. Quantifying the TMS-EEG responses in the time and time-frequency domain revealed that stroke patients exhibited slower and simple responses with higher amplitudes compared to healthy controls. Importantly, these patterns of activity changes after stroke were not only linked to the initial motor deficit, but also to motor recovery after >3 months post-stroke. Thus, the data revealed a substantial impairment of local effects as well as causal interactions within the motor network early after stroke. Additionally, for severely affected patients with absent motor evoked potentials and identical clinical phenotype, TMS-EEG provided differential response patterns indicative of the individual potential for recovery of function. Thereby, TMS-EEG extends the methodological repertoire in stroke research by allowing the assessment of individual response profiles.

Keywords: TMS-EEG; neurorehabilitation; plasticity; slow waves.

Figure 1

TMS-EEG responses to M1 stimulation in representative subjects. (A) Healthy subject. (B) Stroke patients. Top row: Lesion location. Middle row: Butterfly plot representing all 62 EEG electrodes (bold line: stimulation electrode/C3). Bottom row: Topographic plots of the TMS-evoked responses.

Figure 2

Differential TMS-EEG responses in the subgroup of patients with no evocable MEP. Individual TMS-EEG responses of the stimulated ipsilesional motor cortex for all patients (n = 16) without an MEP in the early subacute phase post-stroke. Grey bars indicate the 99% confidence interval derived by bootstrap statistics. Next to the TEP plots, the corresponding motor scores of the individual patients are shown. In addition, coronal slices of the individual diffusion-weighted MRIs are depicted, showing the acute ischaemic lesion. Please also note that bilateral hyperintensities at the temporal lobes results from susceptibility artefacts.

Figure 3

Comparison of TMS-evoked EEG potentials of the ipsilesional motor region using LMFP. (A) TMS-evoked EEG responses for one representative stroke patient and one healthy control subject are shown as butterfly plots of all channels (bold channels represent the region of interest), and (B) the corresponding LMFP with the two respective time intervals (10–100 ms; 100–200 ms). (C) For each time interval analysed, the LMFP values are shown in the bar chart for the entire group of healthy subjects and stroke patients (**P < 0.001; error bars indicate the standard error). (D) Grand-average LMFP for all stroke patients and healthy subjects. Thick traces represent the grand-average across subjects and shaded regions the standard error. Note that the dashed line indicates the timing of the TMS pulse.

Figure 4

Comparison of natural frequencies using ERSP. TMS-evoked EEG responses for a representative stroke patient and one healthy control subject as butterfly plots. Red channels in the butterfly plots highlight the region of interest, i.e. prefrontal, motor, and parietal. The corresponding ERSP patterns (between 5 and 50 Hz) are shown below. Note that the red crosses indicate the region of interest for ERSP analysis and not the stimulation site, which was always the ipsilesional motor cortex. The greyscale graph plotted to the right of each ERSP reveals the power spectrum profile during 20–200 ms after TMS onset. The dotted lines indicate the frequency with maximum power, i.e. the natural frequency. The dashed line indicates the timing of the TMS pulse. The rightpanel shows a bar graph of averaged natural frequencies for each region analysed (*P < 0.05; error bars indicate the standard error).

Figure 5

Correlation analyses. (A) Correlation analyses between initial deficit and TMS-EEG parameters. (B) Correlation analyses between recovery after >3 months post-stroke and TMS-EEG parameters. (C) Correlation analyses between recovery after >3 months post-stroke and TMS-EEG parameters only for the subset of severely affected patients (group cluster no. 1) with no detectable MEP in the first few days post-stroke. TMS-EEG parameters, as well as the recovery scores in B and C, are controlled for the initial deficit using Pearson partial correlations. Note that due to partial correlations, accounting for the initial deficit, axes show residuals of parameters.

Figure 6

Lesion overlap and VLSM. (A) Stroke patients showed the maximum overlap at the level of basal ganglia, including the crus posterius of the capsula interna, the putamen and parts of the thalamus. (B) Lesions associated with enhanced early EEG-activity (LMFP) evoked by TMS responses. (C) Lesions related to the deterioration of the numbers of deflections of the TMS-EEG responses. (D) The bottom panel represents lesions associated with changes in natural frequencies. The colour bars represent the corresponding t-values of the VLSM analysis.