An automated method to determine the transcranial magnetic stimulation-induced contralateral silent period

Abstract

Background: The transcranial magnetic stimulation (TMS)-induced contralateral silent period (CSP) refers to a period of interruption of voluntary muscle activity measured in tonically active muscles. The length of the CSP is generally interpreted to reflect cortical inhibition. The determination of the return of voluntary motor activity is typically accomplished via visual inspection of the electromyography (EMG) waveform and may be subject to inaccuracy on the part of the rater.

Objective: To present and evaluate an automated method (AM) to determine the CSP.

Methods: The CSP of 11 healthy controls was recorded using stimulus intensities 20 and 50% above the resting motor threshold (RMT). The mean CSP duration obtained by the two raters using visual inspection and our automated approach were compared.

Results: The interclass correlation coefficient (ICC) between the two raters and the AM was 0.99 at 150% of RMT and was 0.97 at 120% of RMT. The level of pre-stimulus EMG amplitude and sampling rate did not affect agreement between the AM and more conventional visually guided methods.

Conclusions: Our study demonstrates that this AM is a simple, objective and reliable approach for CSP determination.

Significance: The CSP is an important neurophysiological measure of cortical inhibition and its determination by our AM provides a more objective and automated approach compared to visually guided methods.

Introduction

Transcranial magnetic stimulation (TMS) of the contralateral motor cortex during voluntary muscle activity produces a motor evoked potential (MEP) followed by a period of cessation of electromyographic (EMG) activity that is termed the contralateral silent period (CSP) (Cantello et al., 1992). The early part of this CSP is at least partly related to decreased spinal motor neuron excitability, while the late part (after about 50 ms) is due to intracortical inhibitory mechanisms (Chen et al., 1999, Fuhr et al., 1991, Triggs et al., 1993). The CSP has been shown to be abnormal in many neurological (Cantello et al., 1992) and psychiatric disorders (Daskalakis et al., 2002, Ziemann, 1997) reflecting altered cortical inhibition and may be a key to understanding the pathophysiology of these disorders.

How to demarcate both the start and end of the CSP is often the subject of debate. Some investigators regarded the start of the CSP as the MEP onset, while others used MEP offset or the time of delivery of TMS. Measuring the end of the CSP is even more difficult. The return of voluntary EMG activity, marking the end of the CSP, is frequently imprecise. Terms such as the absolute CSP and relative CSP have been introduced to provide some uniformity when deriving this measure. The end of the absolute CSP is defined as the point in which any EMG activity returns. In contrast, the end of the relative CSP is defined as time when the EMG activity approaches the pre-stimulus state (Tergau et al., 1999). With either method, extraneous signals originating from a variety of sources such as movement artifacts can often cause deflections in the waveform that may be inaccurately interpreted as a return of voluntary motor activity. Such signals may remain even after waveforms are averaged, particularly if they are large. As such, determining the CSP can be subject to an interpretative arbitrariness and may lead to discrepancies between raters. Given that CSP determination lacks the objectivity that is conventionally used with other neurophysiologic measures (e.g. MEP size, cortical inhibition measured by paired pulse methods), being able to reliably automate this procedure may help simplify and standardize its determination.

Nilsson et al. (1997) endeavored to measure the CSP duration through a computer-automated approach. In this method, multiple unpaired student's t tests were performed comparing 2–4 ms epochs of the post-stimulus EMG amplitude to the pre-stimulus EMG amplitude. The end of the CSP was defined as the first epoch in which there was no significant difference compared to baseline conditions. However, several problems exist with these methods of CSP determination. First, multiple comparisons are made without appropriate adjustments such as Bonferonni correction. Thus, it is difficult to ascertain whether the first significant difference is a bona fide one or simply secondary to type I error. Second, this approach fails to take into consideration the magnitude of difference. Hence, a given epoch may not necessarily differ statistically from baseline conditions yet still represent a period of partial cessation of voluntary motor activity and, therefore, be part of the CSP. Third, as Garvey et al. (2001) point out, the data points used in this approach are not independent from each other, violating basic assumptions of the unpaired t test.

Recently, Garvey et al. (2001) reported an alternative mathematical approach for determining the CSP duration based on the mean consecutive difference (MCD) of the data points in the pre-stimulus EMG. The offset of the CSP is defined as the first data point of a 5 ms window in which at least 50% of the data points are higher than (mean −2.66×MCD) for the pre-stimulus EMG. The constant 2.66 was chosen because it represents approximately 3 standard deviations (Garvey et al., 2001). Using this approach they found that the CSP measured using their graphical method closely approximated measurements obtained using the manual method (Intraclass correlation coefficient (ICC)=0.96). Several problems also exist, however, with this approach. First, despite high ICC between two raters and this approach in adult subjects, there were still considerable differences when measured in children (approximately 10 ms). This is, in part, due to the fact that the CSP in children is often short or non-existent making it difficult to demarcate the return of voluntary motor activity. Second, there is no attempt to filter extraneous, non-muscle artifacts. Non-muscular artifact of extraneous sources may remain in the waveform and confound the proportion of data points within the 3 standard deviation range. Third, incipient motor activity beyond 3 standard deviations would not be recognized by their method. As Tergau et al. (1999) defined the absolute CSP as the return of any motor activity; the return of very low intensity motor activity would be missed. Fourth, the CSP duration calculated using their approach varied with different sampling frequencies. This is because at higher sampling frequencies the MCD is decreased, moving the variation limits for a given EMG epoch closer to the mean EMG amplitude level and results in lengthening of the CSP.

In the current study we use a combination of filtering, squaring and threshold detection to measure the CSP through an automated method (AM). AM has the advantage that it is objective and avoids the potential biases and interpretation discrepancies inherent in the more conventional visually guided CSP measuring techniques. The aim of this study is to determine whether our AM is reliable compared to standard methods. We compared the results generated by this AM to visually guided measurements obtained from two experienced raters.