In patient 1, initial DCS approved vPrG to be a language-positive site, whereas DCS during the second awake surgery determined the vPrG and opIFG to be language eloquent spots. In contrast, rTMS elicited the highest error rates with regard to the combined group of no responses and performance errors by stimulation to mPrG during the first and second mapping (Figures 4 and 5).
In patient 2, initial DCS determined vPrG to be a language eloquent spot, whereas vPrG and trIFG were language-positive sites during the second surgery’s DCS. These findings correlate well with the results of rTMS prior to the first surgery, which showed a high error rate by stimulation to vPrG. In addition, there is a partially good correlation between DCS and rTMS prior to re-resection, when both methods elicited errors by stimulation to trIFG. During rTMS, vPrG was not prone to errors this time (Figures 4 and 5).
In patient 3, it turns out that the opIFG and pMFG were sites with clear language errors during DCS in both awake surgeries. With regard to the combined group of no responses and performance errors, rTMS elicited high error rates by stimulation to pMFG in the first and opIFG in the second mapping (Figures 4 and 5).
Looking at patient 1 and 2, it becomes obvious that initial DCS results and findings during the second awake surgery only correlate partially. In comparison with initial findings, DCS approved an additional CPS region to be language eloquent during re-resection in both cases (Figure 5). This aspect can be suggested to be an expression of brain plasticity for language tasks as reported in other studies [4, 22–25]. Additionally, as a result of brain plasticity, rTMS prior to and DCS results during initial surgery only partially corresponded with rTMS and DCS findings of the second examination (Figures 4 and 5). However, rTMS was able to reveal changes in language organization in both cases.
Regarding the rTMS results of those sites included in the craniotomy and undergoing DCS, it appears that stimulations in the precentral gyrus are associated with a high error rate in general which has to be considered as dysarthria by temporarily disturbed function of the motor component of speech rather than actual language impairment [20, 21, 26].
Moreover, we observed in this cohort, that correlation of rTMS and DCS is much better in anterior language areas compared to posterior sites such as aSMG, pSMG, anG, and temporal sites, which is in accordance with a recently published study . However, we have to keep in mind that the observations in this study cannot be based on sufficient statistics due to the small number of patients.
Comparing the results of all initial rTMS sessions with the corresponding remappings, the error rate was higher during the second mapping in two cases (Figure 4, Table 1). In fact, this primarily outlines the progression of aphasia caused by the growth of the tumor and increasing perifocal edema. With regard to the total amount of stimulations, the rTMS mapping session before initial surgery was performed with a higher number of stimulations than the second one in all three cases (Table 1). This is most likely caused by the learning effect of the examiner but also a sign for the reduced ability to focus of our brain tumor patients when suffering from recurrent glioma.
Moreover, with regard to the extent of resection (Figure 1) many rTMS-positive CPS regions were part of the resection cavity with no long-term impairment of language function. This observation of false-positive rTMS sites can be explained by two facts: on the one hand, language was shown to be organized in a complex network, which can undergo substantial reorganization [3, 4, 22–24, 27]. On the other hand, the induced current density and direction by rTMS differs from that induced by DCS. As it was already described for motor mapping, DCS activates cortical axons directly whereas rTMS activates neurons mainly through indirect intracortical pathways [28–31]. Such unspecific activation or inhibition of these intracortical pathways might identify sites rTMS-positive, which do not carry really essential language function. These differences have to be considered when analyzing rTMS mapping results and the correlation between both methods. To face such potential limitations further studies have to be performed on the optimal rTMS intensity, frequency, and duration to improve our current protocol. Moreover, the immediate effect of rTMS on the cortical excitability requires further profound investigation. The current protocol was used in this study because sufficient results were reported in the past and were also able to induce language errors in all patients [7, 10, 11]. Moreover, the (by our definition) “false” positive rTMS sites may also identify language sites, which are not intraoperatively defined and in this case may also be regarded as potentially dangerous areas for resection. The CPS areas, although relatively small, still exceed the size of the 10-mm error margin of the DCS, and thus some false positive results may be due to less dense spatial sampling by the DCS [1, 32].
Pain during rTMS sessions was measured for each patient as mentioned before. In general, language mapping was tolerated well by all subjects, which is shown by the fact that there was no abandonment due to stimulation-related discomfort like in other studies published before (Table 1) [7, 10].
Concerning the value of this new tool, we also have to keep in mind that additional preoperative information on the distribution of language eloquent cortical regions would also enable tailored craniotomies for a more targeted intraoperative DCS. However, our results also show that with the current protocol rTMS is more applicable to show language reorganization instead of language-eloquent cortical sites per se.
Concerning fMRI as another non-invasive method, a recently published case report shows that fMRI failed to provide adequate accuracy compared to DCS and rTMS . However, fMRI cannot be principally regarded as less accurate than DCS or rTMS but there are many studies at hand, which proved insufficient accordance of fMRI to DCS [5, 10, 33, 34]. Yet, when comparing lesion-based investigations by rTMS with hemodynamic studies such as fMRI, one should remember that fMRI does not show neurological activation per se but changed oxygenation levels within the brain which can also be based on impaired tissue oxygenation by the adjacent intracerebral lesions operated on in this cohort [5, 10, 33–35].
Future impact of nTMS on neurosurgery
Preoperative nTMS for motor mapping allows us today to inform each patient individually of possible transient postoperative paresis, as we know exactly how close the rolandic region is to the intended resection border in every single case. Thus, we are able to assess operative risks for permanent paresis more precisely and we can use these data to prepare the patient preoperatively.
Hence, rTMS data for language mapping have also influence that cannot be measured by simple outcome studies but may lead to better prepared patients and thus improve patients’ satisfaction. Likewise, rTMS language mapping could allow us to consider indication for surgery by outlining language negative regions quite reliably. Thus, if we are able to further improve the precision and reliability of this method, we might even be able to operate some of these patients without awake surgery.
However, with regard to the low specificity and PPV of this method, we need to improve rTMS language mapping significantly before thinking about further applications.