Objectivation of laryngeal electromyography (LEMG) data: turn number vs. qualitative analysis

Electromyography (EMG) is a diagnostic procedure to assess the status of muscles and their innervating motor neurons. It is used to discriminate nerve vs. muscle dysfunctions and to detect defects in the nerve-to-muscle signal transmission [1]. The first laryngeal EMG (LEMG) was performed by Weddell [2]. Since then, while recognized as promising diagnostic tool in the worldwide ENT community [3], LEMG has been struggling to become a routinely used diagnostic technique. The major drawbacks of LEMG are on one side the lack of a standardized procedural protocol, and on the other side the fact that its results are qualitative rather than quantitative, completely based on the subjective interpretation of experts in the field [4]. These characteristics make this type of analysis suitable only for a limited group of ENT doctors with decades of experience with this procedure and its interpretation. To overcome these difficulties, in 2008, a group of European expert neurolaryngologists founded a working group aiming to provide systematic guidelines for LEMG execution and analysis. In 2012, the group published a detailed guideline on (1) the minimum equipment requirements for performing and recording LEMGs; (2) the procedural steps to guarantee reproducible results; (3) the essential LEMG interpretation criteria to ensure result standardization, and (4) launched the website https://​www.​lemg.​org to encourage experience exchange among the sites regularly performing LEMG and those still struggling with its implementation in their routine protocol. The European Laryngological Society emphasized the need of standardized training and regular practice [5] particularly considering the initial results of the international LEMG registry conducted by this working group between 2012 and 2016, aiming to evaluate the implementation of the LEMG across Central and Eastern Europe. The registry results clearly showed the limit of qualitative analysis. To our best knowledge, there is no standardized protocol available to effectively quantify the LEMG results.

The results presented in this work aim to evaluate a preliminary protocol for the quantification of the LEMG results and to compare them with a typical qualitative analysis. This latter type of analysis is mainly performed by following the guideline published by Volk et al. [4]. On the contrary, the quantification of LEMG results requires a good understanding of the signal characteristics. In short, any EMG signal can be evaluated in time (i.e. variations of amplitude with time) or in frequency (i.e. number of times an event occurs in the total period of observation). A quantification of the signal can be achieved through the evaluation of “turns” per second, where “turns” are defined as any 100 µV signal amplitude change in the rising or dropping slope of the LEMG signal. Provided that the slope extremities do not overlap, this type of analysis allows an easy exclusion of false positives linked to the presence of interferences or background noises. In 1991, McGill et al. [6] first described a significant correlation between pathologically increased [7] morphologic Motor Unit Action Potentials (MUAPs) and decreased number of turns in neuropathic lesions. Later studies confirmed that lesions of the lower motor neurons innervating voluntary muscles severely impair the motor unit recruitment pattern while increasing the MUAP [7]. Later studies were able to roughly identify the typical number of turns/s expected for various healthy muscles, including the laryngeal ones [8‐10]. Based on all these previous results, we developed a protocol to count the turn number from the respective LEMG for the healthy and ailing TA muscle of patients suffering from unilateral vocal fold immobility (UVFI). We then compared these results with a qualitative analysis of the volitional activity of the same muscle to determine a correlation between the two types of analyses and determine whether their combination would provide a more accurate diagnosis, prognosis, and consequent improved therapeutic approach for the patients.

Datasets of 27 patients were considered compatible with turns/s calculation. 16/27 (59.3%) patients were male and 11/27 (40.7%) were female. The average age was 57.2 ± 15.6 with patients between 18 and 80 years. Details about the UVFI etiology and duration until the time the EMG was performed are provided in Tables 1 and 2.

The average number of turn/s characterizing the ailing VFs was 210 ± 264, while on average the turn number of the healthy VFs was 720 ± 297. The Wilcoxon Signed Rank Test showed that this difference in this pattern of turns/s distribution between the ailing and the healthy VF was significant (p ≤ 0.0001). The qualitative analysis performed according to [4] showed that the volitional activity was found normal (i.e. dense) in the majority of the healthy (75%), while it was absent (30%) or decreased (70%) in all the analyzed ailing VFs.

The Kruskal–Wallis Test showed that there is a statistically significant difference between the volitional activity rating and the turn number assessment performed on the ailing (p = 0.001), but not on the healthy TA muscles (p = 0.226). The Kendall rank correlation showed a strong, positive relationship between the volitional activity rating and the turn number assessment on the ailing (r = 0.853,p ≤ 0.0001), but not on the healthy TA muscles (r = 0.268,p = 0.089). The relationship between the turn number and qualitative results of the volitional activity evaluation is depicted in detail in Fig. 3a, b, respectively.

In 20/27 (74.1%) of the evaluated subjects both the ailing and the healthy VFs were evaluated in the same session. In the remaining 7/27 (25.9%) subjects, the resulting amplitude resolution of the EMG signal of the ailing VF was below the threshold and therefore could not be used for analysis. The results of the qualitative and quantitative analysis are summarized in Table 3. While, as expected the complete lack of volitional activity correlated 100% with no turns, the classification of the volitional activity as normal (or dense) correlated with a turn number comprised between 441 and 1234 with a median of 673 and IQR of 432. Mildly decreased volitional activity correlated with a median number of turns/s of 553. Strongly decreased volitional activity correlated with a turn number comprised between 198 and 1524 with a median of 420 and an IQR of 260. Single fiber volitional activity correlated with a turn number comprised between 62 and 208 with a median of 136 and an IQR of 99.

In 1/20 (5%) case for which the UVFI had an idiopathic case, the qualitative analysis rated the healthy VF volitional activity mildly decreased. In this case, these results were confirmed by the turns/s analysis.

Considering only the ailing VFs, single fiber volitional activity correlated with a very low number of turns/s comprised between 62 and 208, while strongly decreased volitional activity correlated with an increased number of turns/s comprised between 198 and 501 (if excluding the stenosis case).

As expected, the percentage decrease in turn numbers when comparing the healthy with the ailing VF is 100% when the latter is completely paralyzed. Single fiber volitional activity correlates with a turn number decrease comprised between 60 and 88%. Strongly decreased volitional activity associates with a turn number decrease comprised between 51 and 72% and mildly decreased volitional activity with a turn number decrease of 35%.

Our preliminary data confirmed previous works describing the possibility to use turn number count to objectify the LEMG results. Statham et al. [11] already showed a statistically significant turn number decrease (290 vs. 450) in ailing VF of UVFI patients compared to the healthy one, strongly suggesting that neurogenic damages are well characterized by a low number of turns. Indeed, when we examined a patient with ankylosis and no neurogenic damage, the turn number evaluation rated both the ailing and the healthy VF as normal. Accordingly, in 4/20 (20%) cases, the qualitative analysis rated the healthy VF volitional activity strongly decreased. In 3/4 cases (75%), these results were confirmed by the turn number calculation. In all the 3 cases the etiology (laryngeal tumor, long-term tracheostomy and aortic arch surgery) was compatible with bilateral damage of both VFs. In the other case, the patient was diagnosed with UVFI due to glottic stenosis. In this case, it is expected that while the voice quality is compromised due to impaired glottic closure, the turn number remains unaffected and thus scores as healthy (1524).

Smith et al. [12] reported an average turn number ≤ 400 as pathological. Statham [8] as well as Lindestad [9] suggested to use a lower threshold comprised between 200 and 300 turns/s to discriminate healthy from ailing VFs. While these studies focused on the mere distinction between healthy and ailing, we tried to determine whether it would be possible to use the turn number to categorize objectively and in more detail the grade of damage suffered by the ailing VF, using a similar setting to a previously published LEMG qualitative analysis [4]. Our results showed that both the evaluation of the absolute number of turns or the evaluation of the percentage decreased compared to the contralateral healthy VF, are accurate methods to objectify the LEMG outcomes. The preliminary results of our study, although based on a limited sample of data sets, seem to suggest that dense volitional activity associates with a number of turns/s ≥ 600; mildly decreased volitional activity with a number of turns/s between 400 and 600; strongly decreased volitional activity with a number of turns/s between 200 and 400; and a single fiber volitional activity with a number of turns/s below 200. In a similar way, the decreased turn number for mildly decreased volitional activity is ≤ 35%, for strongly decreased volitional activity is between 50 and 70%, for single fiber volitional activity between 60 and 90%. Both phonation segments shorter than 3 s, which may not be sufficiently long to represent a real situation, and spontaneous motor unit activity were excluded from analysis. Fibrillation potentials with amplitudes around 100 µV were not recorded during phonation. These strict criteria for results analysis were important to increase the internal validity of the study but may have impacted the generalizability of the findings. In all cases, result verification in larger cohort studies are planned.

Ho et al. [13] showed acceptable interrater agreement between 7 experienced raters on LEMG datasets, but 14.29% of rater pairs achieved only fair agreement in the category “strongly decreased recruitment pattern”. The interrater reliability might be improved by providing the number of turns/s in addition to the LEMG signal traces.

Another clinically relevant aspect linked to the turn number calculation, is the better understanding and interpretation of qualitative LEMG results. In particular, in 6/27 cases (22%), the qualitative volitional activity of the healthy VF was rated dense, i.e. normal, while the turn number calculated was comprised between 441 and 529 suggesting a mild decrease in the VF functionality. The VFI etiology for these cases was total thyroidectomy, mediastinoscopy or lung transplantation, i.e. an iatrogenic cause that in principle could have affected both VFs, with the less affected VF being asymptomatic at the time of the LEMG. Since 4/6 cases (66.7%) were evaluated between 1 and 4 months upon UVFI onset, this piece of information would help the medical staff to schedule the upcoming control visits and plan the conservative therapy for the patient as well as to formulate a more accurate prognosis. In this study, we did not focus on the MUAP morphology, which has been previously introduced by other authors to discriminate between healthy and ailing VFs [14, 15]. While this evaluation could become the subject of future studies, its current setting often associates with the detection of false positive results. In particular, it has been reported that the detection of increasing polyphasic MUAPs in patients suffering from not better identified axonal polyneuropathy may be accompanied by an increased number of turns/s, similar to that observed in healthy VFs [16]. Also synkinesis can associate with an almost normal EMG pattern and number of turns/s [11]. Accordingly, its evaluation would require larger sample sizes than that featured by this preliminary study.

There are some limitations that could have affected the results of our study. In the first place, it was conducted monocentric and on a limited number of patients. In the second place, it is possible that although the turn number was calculated on the TA, it may include some values generated on the lateral cricoarytenoid (LCA) instead, due to the extreme proximity of the two muscles. Since our results are in line with previously published works, we expect that such potential error did not constitute a significant bias for our result analysis.

Our study is one of the first attempts to compare the results of qualitative vs. semi-quantitative analysis applied to LEMG.

Our initial results on a limited sample of data sets seem to confirm that the implementation of turn calculation could become a useful tool to confirm volitional activity assessment and help discriminating UVFI due to dysfunctionalities of the thyroarytenoid muscle from those caused by glottic gap reduction due to mechanical causes such as stenosis.

Future studies in larger cohorts and including standard voice assessments and/or laryngostroboscopy should be considered confirming these results.