gms | German Medical Science

In the surgical management of acoustic neuromas the facial nerve can be preserved anatomically in most cases [Ref. 5], [Ref. 9]. The fact that 20–70% of the patients suffer from at least transient deficits concerning facial nerve function suggests that functional integrity of the nerve can be harmed without noticeable surgical damage [Ref. 3]. To keep functional integrity under surveillance, intraoperative neuromonitoring using electromyography has been introduced [Ref. 1].

Although a wide variety of EMG-patterns have been described, all descriptions were based on acoustic characteristics and occasional screenshots [Ref. 4], [Ref. 6], [Ref. 7]. The resulting classification systems were rather unsystematic. In 2000, Romstöck and co-workers developed a classification system of EMG-activity with respect to waveform characteristics, frequencies and amplitudes [Ref. 8]. The “A-train”, described as a sinusoidal pattern of high frequency and homogenous appearance, was demonstrated to be the sole pattern clearly indicating postoperative paresis with high sensitivity and specificity [Ref. 8].

In the study presented, a system for automated detection of the A-train is used to calculate a quantitative parameter correlating with postoperative deterioration of facial nerve function. This parameter was called “traintime” and was measured in seconds.

Data obtained from 40 patients (18 women, 22 men) who underwent surgery for acoustic neuroma removal via the lateral suboccipital approach between 1994 and 2003 in our institution were examined in a retrospective study. Neurological examination to determine facial paresis was performed before surgery and ten days after the operative procedure, using the House-Brackmann (H. B.) grading system [Ref. 2]. 26 patients displayed normal function of the facial nerve before surgery (group A), while in 14 patients facial weakness was already found preoperatively (H. B. grade II or III) (group B).

A-trains were assumed to be composed of single “elements”, being mono- to triphasic potentials with amplitudes of 20–2000 µV. All elements in any given A-train appear to share geometrical characteristics; so all elements of a whole recording were divided into certain groups of similarity, called “entities”. All elements of any given A-train belong to the same entity.

Based on visual observation, A-trains are required to be composed of at least four elements. Additionally, there is a frequency-threshold. With automated measuring, the typical frequency range of A-trains was set to 100–200 Hz and, as the A-train is a homogenous, “clean” pattern, the frequency was required to be steady over the course of any single A-train.

A quantitative parameter was acquired by automatically adding up the time intervals during which A-trains were detected. This parameter was named “traintime” and was measured in seconds. Traintime was assessed from all three channels of facial nerve EMG in all 40 sets of data and subsequently correlated with facial nerve function following surgery.

The software program used a range of customized algorithms to pre-process the EMG-data of the three EMG-channels monitoring the facial nerve by needle electrodes, placed transcutaneously into the lateral orbicular oculi muscle, the nasal muscle and the orbicular oris muscle. Artefacts and noise were excluded by specialized filtering. This was followed by the application of a set of specifically developed algorithms in order to detect A-trains. Traintime was then calculated automatically from the resulting data.

Group A consisted of 26 patients without preoperative facial paresis. In group A, six patients showed no A-trains at all. In the remaining 20 patients, traintime ranged from 0.08 sec to 85.49 sec.

13 of the patients in group A showed no or less than 0.5 sec of traintime. Only three of them (23%) experienced postoperative deterioration of facial nerve function of varying degree (postoperative H.B. grade III, IV and V respectively). The other ten patients with less than 0.5 sec of traintime (77%) kept an intact facial nerve function.

The remaining 13 patients in group A presented with more than 0.5 sec of traintime. Without exception they suffered from postoperative paresis, deteriorating to H. B. grade II (7 patients), III (3 patients) and V (3 patients).

This group of 13 patients was subdivided further. As a first subgroup, eight patients showed traintime of 0.5 sec to maximally 10 sec. Six of them (75%) suffered from deterioration of facial nerve function by one grade only; the remaining two proceeded to grade V.

The second subgroup consisted of five patients presenting with more than 10 sec of traintime. Four of them (80%) suffered from deterioration of their facial nerve function by at least two grades: three patients to grade III, one patient to grade V. The fifth patient’s facial nerve function deteriorated by one grade only.

The 14 patients in group B presented with varying degrees of preoperative paresis before surgery. All 14 patients showed traintime, ranging from 0,10 sec to 209.48 sec (mean traintime in group B was 27.96 sec). No cases without traintime were observed, and all 14 patients suffered from deterioration of their facial nerve function postoperatively. This clearly differs from group A, in which most patients with traintime of less than 0.5 sec did not deteriorate.

Six of the 14 patients presented with less than 10 sec of traintime. As in the other group, this moderate amount of traintime was associated with comparatively lesser paresis; facial nerve function deteriorated by one grade only in all six patients.

The other eight patients presented with more than 10 sec of traintime. Six of them showed deterioration of facial nerve function postoperatively by at least two grades. The remaining two patients deteriorated by one grade.

In group B, a good relation between traintime and postoperative facial nerve function was seen just like in group A. Increasing traintime indicated increasing postoperative paresis in both groups.

The two thresholds of 0.5 sec, respectively 10 sec were defined empirically in order to allow for a first quantification of the relation between traintime and postoperative deterioration of facial nerve function. The threshold of 0.5 sec is relevant in the group of patients without preoperative paresis only. A strong correlation between traintime exceeding this threshold and impaired postoperative facial nerve function could be demonstrated. Based on zero false-positive and three false negative results, a sensitivity of 81.3% and a specificity of 100% were achieved for the threshold of 0.5 sec in group A as an indicator of postoperative deterioration of facial nerve function. Results of the Chi-square-test were highly significant (χ²=16.25; p<0,001), suggesting that traintime of more than 0.5 sec reliably indicates deteriorated postoperative facial nerve function in patients without pre-existing paresis of the facial nerve. Thus, traintime of 0–0,5 sec can be described as the “safe zone”; staying below this threshold indicated a high chance of completely preserved facial nerve function, but only in patients without preoperative deficits. For group B, no such “safe zone” could be determined, since all patients with pre-existing paresis of the facial nerve suffered from some deterioration of nerve function postoperatively.

The threshold of 10 sec was basically applicable to both groups. Exceeding 10 sec of traintime held a high chance of more severely impaired postoperative facial nerve function, meaning deterioration by at least two grades of the H. B. grading system. Based on three false-positive and five false-negative results, sensitivity was 66.7% and specificity 88% for the 10 sec-threshold with both groups united. The highly significant result in the Chi-square-test (χ²=12.77; p<0,001) indicates that exceeding traintime of 10 sec is correlated with deterioration of postoperative facial nerve function by at least two grades in both groups of patients.

With all patients from both groups, a good correlation between traintime and postoperative deterioration of facial nerve function was seen (kendall's tau = 0.43, one-sided significance <0.00005).

Figure 1 [Fig. 1]

Although the results of using traintime as a warning criterion were highly significant, a number of false positive and false negative results were observed. In group B, there were two cases presenting with considerable amounts of traintime, but only slight deterioration of postoperative facial nerve function (both patients proceeded from grade II to grade III H. B.). With 27.74 sec and 209.48 sec of traintime (the latter being the largest amount of traintime measured in all 40 patients), postoperative paresis of a greater extent had to be expected in these patients.

For the 10 sec-threshold, an overall sensitivity (both groups united) of 66.7% was calculated, as there were five false-negative results. Altogether, 15 patients suffered from severe deterioration of facial nerve function; five of them showed traintime of less than 10 sec. Actually, traintime was much less than 10 sec in theses cases, being 0 sec (tumour size 2.5 cm), 0.1 sec (tumour size 2.8 cm), 0.12 sec (tumour size 2 cm), 0.57 sec (tumour size 2.5 cm) and 0,94 sec (tumour size 2 cm) respectively. All these patients belonged to group A.

A possible explanation for these five cases is based on the mechanism and pathophysiology of facial nerve damage. While the exact pathophysiological mechanism resulting in A-trains is still unknown, they do exclusively occur in the course of direct manipulation of the facial nerve by stretching and compression during the preparation, thus exerting moderate trauma. It has to be emphasized that massive and rapid trauma to the facial nerve, e.g. by direct bipolar coagulation, anatomical disruption or complete dissection of the nerve is not covered by EMG-monitoring, resulting in an electrophysiological gap.

Traintime is a quantitative parameter calculated by a system for fully automated analysis of intraoperative facial nerve EMG. Analysis was done off-line for 40 sets of data. The parameter was demonstrated to reliably correlate with deterioration of facial nerve function. Two thresholds were defined, indicating the extent of postoperative deterioration of facial nerve function.

Differences in the impact of traintime on patients with, respectively without preoperative paresis were observed. Patients without preoperative paresis were shown to be able to tolerate up to 0.5 sec of traintime without deterioration of facial nerve function, while this was not true for patients with preoperative paresis. In these cases, facial nerve function was always further impaired. Up to the point of 10 sec of traintime, this impairment was limited to a deterioration by one grade H. B. in patients with pre-existing paresis. The transgression of this 10 sec-border lead to a more severe deterioration of facial nerve function by at least two H. B. grades. This was also true for patients without preoperative paresis of the facial nerve.

A system using traintime as a parameter for intraoperative real-time monitoring of the facial nerve can now be implemented, based on the basics outlined here. Automated monitoring of the facial nerve in the operating room has a considerable potential of being helpful in the surgical treatment of acoustic neuroma. Practicability of this parameter in real-time analysis during the operative procedure can be assumed.