In addition to variability in the anatomic sites of stimulation, researchers have experimented with several different approaches with respect to the kind of rTMS delivered. While most studies employed a single rTMS protocol, Chieffo and colleagues (2014) recently administered both excitatory and inhibitory rTMS to the right inferior frontal language areas (and compared these interventions to sham rTMS) to disentangle the role of these areas in language recovery (Chieffo et al. 2014). This study however was not a treatment study because only a single session of each stimulation type was administered. Their findings suggested that excitatory (and not commonly applied inhibitory) stimulation of right homologues can also result in improved language outcomes, which supports theories claiming a compensatory role of these areas to recovery. One other study employed a novel rTMS protocol using two different frequencies within a single rTMS session and demonstrated marked improvement in language performance with this approach (Kakuda et al. 2011; Carey et al. 2010; Iyer et al. 2003); patients were primed with 6 Hz-rTMS for 10 min before the application of low-frequency/1 Hz rTMS for 20 min over the right frontal sites. In another recent study, dual-hemispheric rTMS was delivered in a sequential manner within the same rTMS session. Based on the observation that a bilateral language network is selectively more active during the subacute phase after stroke, first 1 Hz/inhibitory rTMS was applied sequentially over 2 right Broca’s homologues (pars triangularis and pars opercularis), which was then followed by 20 Hz/excitatory rTMS over matching regions of the left hemisphere (Khedr et al. 2014). This approach also led to improved language outcomes.

In a few of the rTMS treatment studies in aphasia, the stimulation sites were localized using cranial landmarks and the 10–20 international system (e.g., Kindler et al. 2012; Kakuda et al. 2011). However, because this method of localization does not adequately address significant differences in normal neuroanatomy or the large differences in anatomy that can be seen in the setting of stroke, application of rTMS across patients can be highly variable using this approach. Therefore, more recent studies have determined sites of stimulation using frameless stereotactic neuronavigation systems that use individual patients’ MRI scans to precisely localize targets for stimulation. This approach minimizes variability across patients and also across multiple sessions of stimulation within subjects (Treister et al. 2013).

Because most treatment studies have been predicated on a specific model of language recovery, a uniform rTMS approach is typically adopted, whereby all patients within a study are stimulated using an identical rTMS protocol. In these studies, as described previously, right PTr within the IFG was most frequently stimulated. Although studies using this approach have reported group-averaged improvements, rTMS applied in this way may not reliably facilitate recovery at the level of individual patients. Correspondingly, to increase the likelihood of therapeutic benefits of rTMS for all patients, there is some effort in this field to establish and validate individualized treatment strategies that use outcome-driven methods for localizing stimulation sites. We (Medina et al. 2012; Hamilton et al. 2010) and our collaborators (Naeser et al. 2011) employed a strategy that involved an optimal site-finding phase as part of the rTMS treatment protocol. In these studies, a single, optimal site was selected on the basis of individual patients’ best response to rTMS, which was first applied over several predefined sites, after which protracted rTMS treatment was delivered to the optimal site (Medina et al. 2012; Hamilton et al. 2010; Naeser et al. 2011; Martin et al. 2009). In the site-finding phase, each patient underwent low-frequency rTMS (1 Hz) in six separate sessions during which he or she was stimulated at (Fig. 7.1), the area in the motor cortex corresponding to the mouth, the pars opercularis (POp; BA44), three sites within the PTr (dorsal posterior, ventral posterior, and anterior PTr), and the pars orbitalis (BA47); the Brainsight® Neuronavigation system (Rogue Research, Montreal) was used to precisely depict these sites and also the TMS coil positions over these sites using individual patients’ own MRI scans. Optimal response to a site was defined as the site that produced the greatest transient increase in picture-naming accuracy. Subsequently, patients were stimulated at their individually determined optimal site, daily over 10 rTMS (1 Hz) sessions. We found that nine out of ten patients responded optimally after inhibition of the right PTr, while only one patient responded optimally to right pars orbitalis stimulation. Importantly, after protracted rTMS treatment, patients who received real stimulation improved in several measures of language production, while patients who received sham stimulation did not improve on any of the measures. Furthermore, the improvement after real rTMS also persisted over at least 2 months after the treatment ended, suggesting long-term efficacy of this approach (Medina et al. 2012).

While we adopted an approach that employed transient rTMS-induced changes in naming performance as a “functional” localizer for treatment, fMRI-driven approaches are also becoming increasingly popular. Using this approach, optimal sites for stimulation are defined on the basis of activation patterns observed on the fMRI in response to specific language tasks (Abo et al. 2012; Szaflarski et al. 2011; Allendorfer et al. 2012). For example, in one study, perilesional stimulation targets were determined in each individual patient as areas that exhibited greater activation during a language task (Eaton et al. 2008). Subsequently, intermittent TBS (iTBS) was delivered to these targets in ten daily sessions. After this treatment, significant improvement in semantic verbal fluency was observed and patients tended to report that they were better in their ability to communicate (Szaflarski et al. 2011). Another study extended this work by defining optimal stimulation sites based on both the fMRI activation patterns and the type of language deficits exhibited by individual patients (Abo et al. 2012). In patients who were categorized as nonfluent patients, inhibitory rTMS was applied to the areas surrounding the IFG, while in patients with fluent aphasia, rTMS was applied to the superior temporal gyrus (STG; Fig. 7.1). Specific stimulation sites within these territories were then defined by the fMRI activation patterns acquired as the patients performed a language task. In fluent patients, improvement after ten daily 1 Hz rTMS sessions (40 min/day) was reported in auditory and reading comprehension and repetition tasks, and in nonfluent patients, spontaneous speech was reported to have improved.

Because optimal site-finding approaches, whether rTMS- or fMRI-driven, account for individual variability across patients, they are likely an improvement over studies wherein stimulation is guided only by cranial landmarks, although the superiority of one site-finding approach to another is yet to be determined (Heiss and Thiel 2006).