Advertisement

A novel application of an adaptive filter to dissociate the effects of TMS on neural excitability and trial-to-trial latency jitter in event-related potentials

Open AccessPublished:February 12, 2022DOI:https://doi.org/10.1016/j.brs.2022.02.002
      Dear editor
      Causal effects of transcranial magnetic stimulation (TMS) on neurocognitive processes are commonly assessed using event-related potentials (ERPs) [
      • Farzan F.
      • Vernet M.
      • Shafi M.M.
      • Rotenberg A.
      • Daskalakis Z.J.
      • Pascual-Leone A.
      Characterizing and modulating brain circuitry through transcranial magnetic stimulation combined with electroencephalography.
      ]. However, changes in ERP amplitude between any experimental condition can come from differences in the intensity of the neural response to cognitive events (i.e., signal strength), trial-to-trial variation in the timing of neural responses to cognitive events (i.e., latency-jitter), or a combination of both [
      • Luck S.J.
      An introduction to the event-related potential technique.
      ,
      • Unsal A.
      • Segalowitz S.J.
      Sources of P300 attenuation after head injury: single-trial amplitude, latency-jitter, and EEG power.
      ]. For instance, a TMS-induced reduction in ERP amplitude could arise from the following scenarios: (i) TMS decreases the signal strength, (ii) TMS increases the latency-jitter, and (iii) TMS decreases the signal strength and increases the latency-jitter (Fig. 1A and B). While an effect of TMS on cortical excitability (or signal strength) has been the most common interpretation, one cannot make a definite inference without evaluating the impact of TMS on trial-to-trial latency-jitter, which may further provide important insight into the effects of TMS on neural efficiency [
      • Shucard D.W.
      • Covey T.J.
      • Shucard J.L.
      Single trial variability of event-related brain potentials as an index of neural efficiency during working memory.
      ]. To resolve this issue, we applied a signal processing technique, the Adaptive Woody Filter (AWF) [
      • Harris E.K.
      • Woody C.D.
      Use of an adaptive filter to characterize signal-noise relationships.
      ,
      • Woody C.
      Characterization of an adaptive filter for the analysis of variable latency neuroelectric signals.
      ], that measures and corrects for latency-jitter found in ERPs in order to dissociate the effects of TMS on common ERP components (N200, P200, P300) across two TMS studies [
      • Baker T.E.
      • Lesperance P.
      • Tucholka A.
      • Potvin S.
      • Larcher K.
      • Zhang Y.
      Reversing the atypical valuation of drug and nondrug rewards in smokers using multimodal neuroimaging.
      ,
      • Baker T.E.
      • Lin M.-H.
      • Gueth M.
      • Biernacki K.
      • Parikh S.
      Beyond the motor cortex: theta burst stimulation of the anterior midcingulate cortex.
      ]. We hypothesize that if TMS does alter trial-to-trial variation in the timing of neural responses to cognitive events, one would observe changes in the effect size between pre-and-post AWF-correction. Alternatively, if correcting the latency-jitter does not impact the effect size, then the effects of TMS on ERP amplitude likely resulted from a direct change in signal strength.
      Fig. 1
      Fig. 1(A) An illustrative example of a reduction in N200 amplitude following TMS. Parent ERP waveforms for SHAM and active TMS conditions. (B) An illustration of the impact of the effects of TMS on latency-jitter, signal strength, or a combination of both. (CE) Grand-averaged Raw and AWF-corrected ERP waveform at Cz for (C) 10-Hz rTMS, (D) continuous TBS (cTBS), and (E) non-TMS dataset. For the purpose of comparison, an identical analysis on a non-TMS dataset revealed no significant differences in P200 and P300 amplitudes between BLK3 and BLK1 in Raw and AWF-corrected data (ps > .05). (F-G) The bar graphs depict (F) Raw and AWF-corrected ERP amplitudes, CCMax, and CCRaw for N200, and (G) Raw and AWF-corrected ERP amplitudes for P200 and P300. (H) Latency-jitter across 10-Hz rTMS, cTBS, and No TMS datasets. Bar lines denote standard error of the mean. To note: false discovery rate (FDR) was used to control for multiple comparisons.
      In Study 1 (Baker et al., 2017), 1000 pulses of repetitive TMS (rTMS) pulses were delivered offline at 110% of participants' resting motor threshold at 10-Hz over the left dorsal lateral prefrontal cortex (DLPFC) target as participants performed a feedback task (200 trials). Identical parameters were applied to the SHAM session on a separate day except that the coil was flipped (order counterbalanced) [
      • Baker T.E.
      • Lesperance P.
      • Tucholka A.
      • Potvin S.
      • Larcher K.
      • Zhang Y.
      Reversing the atypical valuation of drug and nondrug rewards in smokers using multimodal neuroimaging.
      ]. In Study 2 (Baker et al., 2020), 600 pulses of continuous theta-burst stimulation (cTBS) were applied at 80% of participants' resting motor threshold over the left DLPFC following the first block (100 trials) of the same task. Participants then completed three additional blocks (5–7 mins per block) [
      • Baker T.E.
      • Lin M.-H.
      • Gueth M.
      • Biernacki K.
      • Parikh S.
      Beyond the motor cortex: theta burst stimulation of the anterior midcingulate cortex.
      ]. EEG data were analyzed, and the peaks of pre-and-post AWF-corrected ERPs were measured according to parameters reported in the original studies [
      • Baker T.E.
      • Lesperance P.
      • Tucholka A.
      • Potvin S.
      • Larcher K.
      • Zhang Y.
      Reversing the atypical valuation of drug and nondrug rewards in smokers using multimodal neuroimaging.
      ,
      • Baker T.E.
      • Lin M.-H.
      • Gueth M.
      • Biernacki K.
      • Parikh S.
      Beyond the motor cortex: theta burst stimulation of the anterior midcingulate cortex.
      ]. The AWF-correction was applied to each negative feedback segment after artifact rejection and before averaging in BrainVision Analyzer (Brain Products GmbH, Germany) using the Brainwaves Research Lab ToolBox [
      • Gavin W.J.
      • Davies P.L.
      Adaptive Woody filter. A MATLAB program created in the.
      ]. The AWF is a template-matching procedure consisting of the following steps: (1) pre-AWF corrected ERPs for negative feedback were averaged and served as the template for each subject, (2) the correlation-coefficients between single-trial ERPs and the template in the time window of 150–450 ms post-stimulus (which captures the P200–N200–P300 complex) at channel Cz were computed and averaged, (3) single-trial ERPs were shifted forward and backward one data point at a time to a maximum of 70 ms until the “best” fit (i.e. the maximum correlation-coefficient) was found, and (4) shifted segments were averaged to create AWF-corrected ERPs. Three AWF-related parameters were derived for each subject: (1) the averaged correlation-coefficient between single-trial ERP to the template before AWF-correction (CCRaw), (2) the averaged maximal correlation between single-trial ERP to the template after AWF-correction (CCMax), and (3) the latency-jitter: the standard deviation of the amount of shift across trials [
      • Gavin W.J.
      • Lin M.H.
      • Davies P.L.
      Developmental trends of performance monitoring measures in 7- to 25-year-olds: unraveling the complex nature of brain measures.
      ]. Cohen's d was used to measure the effect size (trivial effect: d ≤ 0.19; small effect: d = 0.20–0.49; medium effect: d = 0.50–0.79; large effect: d ≥ 0.80).
      First, we initially reported that rTMS significantly increased the amplitude of the feedback-locked N200 component (p = .005, Fig. 1C and F) [
      • Baker T.E.
      • Lesperance P.
      • Tucholka A.
      • Potvin S.
      • Larcher K.
      • Zhang Y.
      Reversing the atypical valuation of drug and nondrug rewards in smokers using multimodal neuroimaging.
      ]. Following AWF-correction, this effect remained significant (p = .02), confirming a direct impact on cortical excitability. Interestingly, the effect size was reduced from large to moderate (d: 0.80 → 0.65), and there was a trend for a lower amount of latency-jitter in the rTMS condition relative to SHAM (p = .12, d = 0.40; Fig. 1H). These findings suggest that the previously observed effect of rTMS on N200 was driven by both a decrease in trial-to-trial variability and an increase in signal strength, but more so for the latter (see regression analysis in SOM). Next, we previously demonstrated that relative to Baseline, the P200 amplitude was significantly enhanced following cTBS (p = .04, Fig. 1D and G) [
      • Baker T.E.
      • Lin M.-H.
      • Gueth M.
      • Biernacki K.
      • Parikh S.
      Beyond the motor cortex: theta burst stimulation of the anterior midcingulate cortex.
      ]. Following AWF-correction, the effect remained significant (p = .002) and the effect size was increased from medium to large (d: 0.38 → 0.58), indicating that naturally occurring latency-jitter may have masked the true effects of cTBS on cortical excitability. By contrast, the P300 amplitude was significantly reduced following cTBS relative to Baseline (p = .004). Following AWF-correction, this effect remained significant (p = .03), yet the effect size was reduced (d: 0.55 → 0.39), indicating that latency jitter may have exaggerated the suppressive effect of cTBS on P300 amplitude. No differences in latency-jitter were found between BLK3 and BLK1 (p = .86; Fig. 1H). Regression analyses revealed that the observed effects of cTBS on P200 and P300 amplitude were predominantly driven by changes in signal strength (see regression analysis in SOM). Together, these results revealed that while cTBS can cause a direct impact on cortical excitability, latency-jitter may confound the cTBS effects on the Raw ERP amplitudes, and correcting for it amplified the effects of cTBS on P200 amplitude yet attenuated the effects of cTBS on P300 amplitude.
      Taken together, these findings confirm that both rTMS and cTBS can cause a direct impact on cortical excitability and that removing the latency-jitter to the greatest extent allows for a more accurate evaluation of the contribution of a change in cortical excitability to the effect size of the TMS manipulation. Moreover, by discriminating components of the ERP to cortical excitability from latency-jitter, AWF provides a novel measure of the latter, thereby providing evidence that rTMS not only enhances cortical excitability but may also improve the efficiency of the neural system being targeted [
      • Shucard D.W.
      • Covey T.J.
      • Shucard J.L.
      Single trial variability of event-related brain potentials as an index of neural efficiency during working memory.
      ]. Future studies should consider applying the AWF or techniques that can account for latency-jitter to help identify the true causal effects of TMS on motor, cognitive, and perceptual functioning as measured by ERPs.

      Declaration of competing interests

      The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

      Acknowledgements

      We thank Drs. Bill Gavin and Patricia Davies for the MATLAB code used to conduct the Adaptive Woody Filter analyses. This research was supported by departmental research start-up funds from Rutgers University (to TEB).

      Appendix A. Supplementary data

      The following is the Supplementary data to this article:

      References

        • Farzan F.
        • Vernet M.
        • Shafi M.M.
        • Rotenberg A.
        • Daskalakis Z.J.
        • Pascual-Leone A.
        Characterizing and modulating brain circuitry through transcranial magnetic stimulation combined with electroencephalography.
        Front Neural Circ. 2016; 10: 73https://doi.org/10.3389/fncir.2016.00073
        • Luck S.J.
        An introduction to the event-related potential technique.
        2014
        • Unsal A.
        • Segalowitz S.J.
        Sources of P300 attenuation after head injury: single-trial amplitude, latency-jitter, and EEG power.
        Psychophysiology. 1995; 32: 249-256https://doi.org/10.1111/j.1469-8986.1995.tb02953.x
        • Shucard D.W.
        • Covey T.J.
        • Shucard J.L.
        Single trial variability of event-related brain potentials as an index of neural efficiency during working memory.
        in: Schmorrow D. Fidopiastis C. Foundations of augmented cognition: neuroergonomics and operational neuroscience. AC 2016. Lecture notes in computer science. vol. 9743. Springer, Cham2016https://doi.org/10.1007/978-3-319-39955-3_26
        • Harris E.K.
        • Woody C.D.
        Use of an adaptive filter to characterize signal-noise relationships.
        Comput Biomed Res. 1969; 2: 242-273https://doi.org/10.1016/0010-4809(69)90005-6
        • Woody C.
        Characterization of an adaptive filter for the analysis of variable latency neuroelectric signals.
        Med Biol Eng. 1967; 5: 539-554https://doi.org/10.1007/BF02474247
        • Baker T.E.
        • Lesperance P.
        • Tucholka A.
        • Potvin S.
        • Larcher K.
        • Zhang Y.
        Reversing the atypical valuation of drug and nondrug rewards in smokers using multimodal neuroimaging.
        Biol Psychiatr. 2017; 82: 819-827https://doi.org/10.1016/j.biopsych.2017.01.015
        • Baker T.E.
        • Lin M.-H.
        • Gueth M.
        • Biernacki K.
        • Parikh S.
        Beyond the motor cortex: theta burst stimulation of the anterior midcingulate cortex.
        Biol Psychiatr: Cognitive Neuroscience and Neuroimaging. 2020; 5https://doi.org/10.1016/j.bpsc.2020.06.009
        • Gavin W.J.
        • Davies P.L.
        Adaptive Woody filter. A MATLAB program created in the.
        Brainwaves Research Laboratory at Colorado State University, Fort Collins, Colorado2013
        • Gavin W.J.
        • Lin M.H.
        • Davies P.L.
        Developmental trends of performance monitoring measures in 7- to 25-year-olds: unraveling the complex nature of brain measures.
        Psychophysiology. 2019; 56e13365https://doi.org/10.1111/psyp.13365