3.5.1. Sixty-One-Year-Old Male with Subarachnoid/Subdural Hemorrhage

A 61-year-old male presenting with left-sided subdural and subarachnoid hemorrhage after a fall was transferred to ICU. For clinical management, left-sided hemicraniectomy and sedation were required, the latter of which kept the patient in a comatose state. Due to asymmetric reaction to painful stimuli (left limbs moving against gravity, no movement on right side), motor mapping by nTMS was performed on the left hemisphere to test MCS function (right hemisphere was blocked due to cranial bolt). In contrast to the clinical examination, we were able to demonstrate intact MEPs in the adductor digiti minimi muscle (rMT: 36% of maximum stimulator output). Furthermore, by increasing mapping intensity to 125% of rMT, we observed additional MEPs in biceps brachii and flexor carpi radialis muscles, indicating that corticospinal tract innervation of those muscle groups was intact. As a result of this map, additional workup (CT angiography, EEG, MRI) of the patient's asymmetric examination was deferred. A follow-up motor mapping performed three days later showed a decrease in rMT to 28% of maximum stimulator output, indicating normalization of MCS threshold. Additionally, MEPs for tibialis anterior and gastrocnemius muscles were now detectable. Volitional movement started to return in the clinical examination six days after the initial mapping with slight withdrawals from painful stimuli. At discharge, 14 days after initial motor mapping, the patient displayed symptoms of hemineglect, but was able to move all extremities spontaneously.

### 3.5.2. Sixty-Seven-Year-Old Male with C1 Ring Fracture

A 67-year-old male patient presenting with comminuted/displaced C1 ring fracture after assault was treated on the ICU, presenting without any initial neurological symptoms. The patient underwent stabilization surgery, after which he developed aspiration pneumonia with acute respiratory distress syndrome (ARDS), necessitating prolonged sedation and ventilation under medically induced paralysis. Seventeen days after admission, the patient's ARDS improved and the paralysis was stopped. After the paralytics had worn off (as determined by intact train-of-four), the patient would open their eyes and track movement but remained unresponsive to central or peripheral stimulation. Motor mapping by nTMS was performed for both hemispheres. Despite the lack of movement in the clinical examination and a concurrent propofol infusion (10 mcg/kg/min), we were able to demonstrate MEPs in at least one muscle for every limb, albeit with very high rMT (81% and 99% of maximum stimulator output for left and right hemisphere, respectively). These findings effectively ruled out an occult spinal cord injury, and as a result, further workup with MRI and EEG was deferred. The next day, in a follow-up study with increased propofol dosage (40 mcg/kg/min), MEPs disappeared from both upper limbs and the left lower limb. The right lower limb still demonstrated MEPs but required a higher stimulation intensity of 100% versus 90% on the day before. These changes were thought to be a result of the increased propofol dose. In the subsequent days, the patient level of consciousness improved but he remained unable to follow commands or move volitionally, which was attributed to myoneuropathy of critical illness. Shortly thereafter, the patient developed a bowel perforation, and died of septic shock.

#### 3.5.3. Seventy-Seven-Year-Old Male with Subdural Hematoma

A 77-year-old male was referred to the ICU with left holohemispheric subdural hematoma and midpontine hemorrhage after a fall. The patient remained non-verbal during his entire treatment. At admission, the patient did not follow commands and displayed hemiparesis on his right side but showed flexion to painful stimulation with his left side. The patient's family declined immediate surgical intervention, instead opting to wait for potential improvement over time. Four days after admission, the patient started following commands from his wife, but continued to only move his left side. Eight days after admission, the neurological state of the patient deteriorated, and he demonstrated left-sided decerebrate posturing. CT imaging demonstrated an increase in the size of the subdural hematoma and subfalcine as well as uncal herniation. After consulting with the patient's family, hemicraniectomy and evacuation of the hematoma were performed over the left hemisphere. Two days after the procedure, nTMS motor mapping was performed due to continued absence of movement on the right side. Contrary to the clinical presentation, MEPs were present bilaterally for both upper and lower extremities. Notably, the rMT on the left hemisphere was higher (42% of maximum stimulator output) than the rMT on the right hemisphere (34% of maximum stimulator output), despite the left sided hemicraniectomy. The results were therefore interpreted as pointing towards a damaged but still functional MCS, with hemineglect as a potential explanation for the present symptoms. Physical therapy, therefore, emphasized maintaining right-sided range of motion and targeted therapy to address right hemi-neglect. During this time, patient began to regain motor function on his right side. A follow-up mapping took place 8 days after the inital map. At this point, the patient had received cranioplasty and demonstrated an rMT of 37% of maximum stimulator output for the right hemisphere and 62% of maximum stimulator output on the left hemisphere. The patient's rehabilitation continued, his neglect improved, and he eventually demonstrated 4/5 muscle strength throughout his right side. Upon his discharge 26 days after admission, he was intermittently following commands, moving all his extremities purposefully, and was able to continue his physical and cognitive therapy in an acute rehabiliation institution.

#### **4. Discussion**

We performed nTMS motor mappings in 21 patients being treated in the neurological ICU for a variety of neurological injuries. All mappings were successfully and safely completed. In so doing, we identified common problems that occur in the ICU setting and, where possible, we arrived at solutions for these problems. This feasibility study will enable future studies to focus on establishing the clinical value nTMS-based motor mappings in the ICU and, more broadly, in the inpatient setting.

#### *4.1. Safety*

As with any new clinical test, particularly when dealing with the critically ill, patient safety is of utmost importance. nTMS in general is widely considered a safe modality, especially when single-pulse protocols are used [5,21]. The most serious adverse effect associated with TMS is the occurrence of epileptic seizures. These, however, are exceedingly rare, limited to isolated case reports often in individuals with prior history of epilepsy [27,28]. While we did exclude from our study any patient with uncontrolled epilepsy, none of our patients showed any adverse effects during or immediately after single-pulse nTMS motor mapping. No acute worsening of clinical state was attributable to motor mapping. This was also true for the nine patients who underwent mapping following prior hemicraniectomy, in whom the magnetic field incident upon the cortex is likely stronger than for patients with an intact skull. It should be noted again, however, that we capped the stimulation intensity for hemicraniectomy hemispheres to 75% of maximum stimulator output for precautionary reasons. Although these data are gathered from a limited number of cases, it is highly encouraging and indicates that nTMS is likely safe enough to be performed in patients with severe acute neurological damage and altered cerebral anatomy.

#### *4.2. Neuronavigation Based on CT*

A prerequisite for nTMS motor mapping is preexisting cranial imaging used for neuronavigation. In the conventional workflow, MRI offers unparalleled anatomical imaging of brain tissue with a high soft tissue contrast, and it is, therefore, the gold standard for navigational imaging [8]. MRI, however, is oftentimes not a viable option for ICU patients due to practical considerations or contraindications for scanning. Specifically, critically ill patients are often unable to tolerate lengthy MRI studies. Furthermore, intracranial monitoring devices such as cranial bolts and indwelling electrodes can be incompatible with the scanning environment [29]. Although it offers less detailed imaging of intracranial anatomy than MRI, CT is by far the more prevalent modality for patients in the ICU due to its rapid acquisition time and compatibility with clinical equipment [29,30]. Therefore, to enable

widespread use of nTMS motor mapping in the ICU, we realized that CT must be adapted and validated as the basis for neuronavigation. Using CT for this purpose has previously only been described in a singular recent case which did not address possible navigation discrepancies between CT and MRI [31]. By preprocessing CT scans with a slice thickness of 2 mm, we were able to generate usable head models suited for patient co-registration (Figure 1). Comparing the calculated maximum e-field at corresponding points in CT head models to that in MRI head models, we found no significant difference in calculated e-field values. While the maximum difference between two points was 18 V/m, we believe that this discrepancy was likely an artifact due to the inherent difficulty of co-registering a given head model to a physical dummy with different measurements. Our results indicate that both CT and MRI can yield comparable neuronavigation in the employed nTMS system. Following our process enables targeted, replicable stimulation of given brain loci in patients lacking MRI data. This application of CT-based neuronavigation not only makes motor mapping in the ICU an option, but also extends nTMS usage to any patient unable to undergo MRI (because of retained metal fragments, implanted devices, etc.).

#### *4.3. Compatibility with ICU Workflows*

In our cases, translation of the outpatient workflow into the ICU setting did, for the most part, not pose significant problems. Nursing routines or monitoring were not impeded. In this regard, the flexibility of nTMS motor mappings is a relevant asset, as the examination takes place at the bedside and can be paused at any moment. Pulse oximetry, while being influenced by navigational equipment (Figure 2), can easily be maintained by physically covering the oximetry probe. The only major limitation was imposed by immovable devices connected to the patient's skull (i.e., cranial bolts placed for anchoring invasive ICP monitoring), as they tended to physically obstruct free movement of the coil. Overall, our experience demonstrated that nTMS motor mapping is compatible with the monitoring setup and clinical routine in ICU cases.

#### *4.4. Optimization of EMG*

For reliable and accurate motor maps, clear EMG readings are necessary to detect even MEPs with small amplitudes. Any active electrical device in proximity to the recording site (e.g., ventilators, perfusors, and monitoring) is a potential source of noise [23]. It is, therefore, unsurprising that noise levels in the ICU setting are far higher than in the normal outpatient setting. Usually, electronic noise presents as a uniform waveform with frequency of about 60 Hz (corresponding to the 60 Hz alternating current of standard US power outlets). Elimination of noise by deactivation of electrical devices is for the most part impossible since their continued functioning is essential for patient care. While this noise generally allows for MEP detection via disruption of the regular noise pattern (Figure 5), maximum signal-to-noise ratio is required to ensure validity and comparability of individual mappings. To minimize noise, general guiding principles of EMG should be followed, such as avoiding contacts between individual cables, cleaning the skin prior to electrode placement, and ensuring full surface contact of electrodes with skin.

The most important factor in our study, however, seemed to be the location of electrode placement, particularly regarding reference and grounding electrodes (Figure 3). Established anatomical landmarks that work in the outpatient setting may not work in the intensive care case. Anatomical positions often proved to be successful neutral electrode targets in this study (Figure 4). In our cases, we were able to achieve an average noise amplitude of 64 ± 58 µV. This distribution is slightly skewed by a few cases with persistent high-amplitude noise levels, which is reflected in the median level of 43.5 µV. Analyzing data at this level of noise is possible, since any valid MEP (threshold of 50 µV) will peak through the background noise. In higher noise levels, however, analysis gradually loses sensitivity and it becomes difficult to detect low-amplitude MEPs. Although this noise lowers the resolution of motor mappings, useful information regarding binary questions of motor system integrity (such as "is any activation present?") can still be obtained. One heuristic for detecting valid MEPs amongst

regular noise is to subtract noise amplitude from the potential MEP. If the amplitude is still >50 µV, the MEP may be considered valid (Figure 5). Placement of multiple stimuli at a given location to verify replicability is also advisable in cases of questionable MEPs.

**Figure 5.** Motor evoked potential (MEP) disrupts 60 Hz noise. This figure shows regular 60 Hz noise disrupted by MEP occurrence. Measurements on the right correspond to amplitude between yellow measurement bars. One heuristic for detecting valid MEPs amongst regular noise is to subtract the noise amplitude from the potential MEP. If the amplitude is still >50 µV, the MEP may be considered valid (pictured). Additionally, placement of multiple stimuli at a given location can aid when MEPs are questionable. Note the two individual peaks within the MEP formed by the underlying noise reaching its minimum.

### *4.5. Pharmacological and Physiological Confounders*

Patients in the ICU are subject to a variety of factors that most likely affect MEP elicitability, such as elevated ICP, hemicraniectomy, and propofol sedation. Propofol has previously been shown to decrease MEP amplitude, yet without impacting MEP latency [32–34]. One study compared the influence of different anesthetic agents on MEPs in animals and humans. The results indicated that even in high concentrations, propofol was unable to fully inhibit MEPs elicited by higher-than-threshold stimulation [34]. Additionally, it has been demonstrated that intravenous agents such as propofol impact MEPs less than volatile anesthetics [35]. The highest sedation levels in our study were propofol infusion of 80 mcg/kg/min and 70 mcg/kg/min. These doses did not prevent elicitation of MEPs in either case. While other patients may require higher doses, the current findings fit to the aforementioned literature, which indicates that MEP monitoring is still possible under propofol sedation. Based on the literature and our experience, we would recommend that patients under sedation showing no MEPs at 105% rMT should be tested with progressively higher stimulation intensities. While not tested in our approach, the literature indicates that TMS with repetitive pulse patterns might also be an option to elicit reliable MEPs under propofol sedation [34,36]. The dose-dependent input/output relationship between stimulus intensity and MEP amplitude might be exploited for individualized functional anesthesia monitoring. For nTMS applications in the ICU that are not specifically related to the MCS, it should be kept in mind that higher-order neuronal function such as network complexity are influenced by propofol as well [37]. Of note, opioids seem to only possess a small effect on MEPs, and we found no dampening of MEP due to opioids in our series [35].

Furthermore, ICP is another factor that could plausibly influence MEPs. Persistently elevated ICP is a known predictor of poor neurological outcome, as it lowers cerebral perfusion pressure resulting in cerebral hypoxia [38,39]. These effects are primarily observed at sustained ICPs of 25 mmHg and higher [38]. All of our cases were performed with an ICP lower than 20 mmHg, as patients are treated aggressively to maintain an ICP below this value. Therefore, while we can state that MEPs were still elicitable in the highest observed ICP of 18 mmHg, we can make no claims as to the influence of ICPs in higher ranges. Hemicraniectomy is a procedure often utilized to combat elevated ICP [40,41], and nine of our patients had undergone this procedure. While our main concern in mapping patients after hemicraniectomy was patient safety, it is worth reporting that MEPs could still be elicited in this abnormal anatomical state.

#### *4.6. Implications and Applications*

We believe that our reported findings serve as valid evidence for the general feasibility of nTMS motor mappings in the intensive care setting. In our series, we were able to map successfully 18 out of 21 patients, yielding reproducible MEPs, and resulting in no complications. These findings demonstrate that, for the critically ill patient with neurological injury, nTMS can deliver quantitative analysis of MCS functionality in a safe and reliable manner. nTMS motor mapping allows for objective investigations into the integrity of the motor system along its entire path, from upper motor neurons to lower motor neurons, and thereafter to the neuromuscular junction. Importantly, it does not rely on patient consciousness or cooperation. Furthermore, rMT specifically has been described to reflect neuronal excitability, which may be a metric of interest in a variety of acute neurological problems [42]. One example is traumatic brain injury, which has been linked to rises in rMT [43].

The potential applications and benefits of nTMS motor mapping on the ICU are numerous. Examinations can be performed on conscious and unconscious patients alike and yield detailed information about the function of individual corticospinal tracts. The procedure is painless, noninvasive and does not require (yet is possible during) sedation. Additionally, any trained member of the clinical team can perform a mapping without the need for additional personnel.

Compared to other, more established neurophysiological modalities such as EEG or somatosensory evoked potentials, nTMS offers the unique capability of linking effects to a specific, repeatedly targetable cortical site [16]. This enables detailed tracking of functional MCS changes over time, which is not present with non-navigated TMS (due to targeting inaccuracy) or EEG methods. Earlier diagnosis of specific motor deficits and prediction of motor recovery may be possible, since corticospinal integrity can be objectively demonstrated for any EMG-suited target muscle. For example, MEP recovery has recently been shown to predict neurological outcome in endovascular thrombectomy [44]. Muscular recruitment curves could serve as the basis of a standardized measure of cortical excitability. In a similar vein, individualized functional drug monitoring specifically for antiepileptic drugs or anesthesia seems plausible [33,45–47]. Protocols for specific types of monitoring could potentially be used for recognition and monitoring of transient complications, such as vasospasm in subarachnoid hemorrhage.

Of note, in all three clinical cases introduced above, nTMS motor mapping yielded information on the acute state of the MCS that was not visible in the clinical examination by demonstrating functional corticospinal motor connection despite lack of movement.

#### *4.7. Limitations*

While our findings confirm the general feasibility of nTMS motor mappings in the ICU setting and demonstrate some use cases, there are several limitations that need to be addressed. First, this study includes only 21 unique patients, which may be considered a small sample size. However, as a pilot feasibility study, we are confident that this sample is adequate to provide general guiding principles that may be built upon in future studies. Second, our imaging testing was limited both by number of cases and by the margin of error in registering a head model to a dummy head of different size. We believe, however, that with 104 total targets per modality, we would have been able to discover any meaningful systematic difference in e-field calculation if it had existed. A separate aspect of interest would be to investigate the comparability of CT and MRI not only based on calculated e-field, but also on physical coil position, which would require some form of system-separate coordinate measurement. For a more extensive comparison, a large set of cases with CT and MRI datasets acquired within a small timeframe would be necessary. Since CT is inherently connected to radiation exposure, a study such as this could hardly be done with healthy volunteers. Even with patients, however, indications for both imaging modalities in short intervals remain rare, which unfortunately limits the pool of usable data. In conclusion, while the presented evidence may be lacking in robustness, it serves as a basis for more extensive systematic testing and, more importantly, demonstrates that repeatable targeting of specific cortical stimulation sites is possible with CT imaging. Third, this study took place in a single ICU at a single institution. Other centers may encounter other challenges based on respective local workflow or electrical noise environment. Fourth, in terms of clinical factors influencing MEPs, we were unable to collect data on high-level ICPs as well as sedation other than propofol. It is difficult to see a simple resolution to this limitation, since optimal treatment of the patient in terms of ICP control and best practice sedation is naturally of the highest priority. Data collection is, however, ongoing, and respective cases may appear to further our understanding of how clinical factors act upon MEPs. Specifically, a standardized I/O curve for varying levels of propofol could be very useful in interpreting mapping results achieved under varying levels of sedation. Fifth and lastly, while the navigational aspect of nTMS enables reproducible targeting of specific cortical sites, the resulting MEPs are also influenced by aspects aside from cortical stimulation site. This could complicate longitudinal patient-specific monitoring of nTMS based parameters such as rMT. Examples for these factors include underlying EEG activity, prior activity of the target muscle or limb posture [48–51]. While these factors are known confounders of MEP replicability, other studies have shown that by sampling 20 or more trials, good inter-session reliability can be achieved regarding measures of motor cortex excitability [48,49]. This should be kept in mind for any clinical application involving longitudinal measurement of MCS excitability.

### **5. Conclusions**

This study demonstrated the feasibility of nTMS motor mappings in the ICU setting. In so doing, we have shown how a variety of confounding factors inherent to the ICU setting may be mitigated. We have demonstrated that MEPs can be obtained in spite of propofol and opioid sedation, that electrical interference due to ICU equipment can be reduced to the point where mapping is possible, and that CT may be used successfully for nTMS navigation. Moreover, we have demonstrated how nTMS can be successfully integrated into routine workflow and intensive patient care. Our data further suggest that single-pulse motor mapping is safe in a range of different diagnoses commonly encountered in the neurological ICU. We have also presented three case studies in which nTMS mapping provided valuable insight into the neurological state of the patient, and two cases changed the clinical management. These results demonstrate a need for future studies to evaluate systematically the clinical benefits provided by the use of nTMS in patients with acute neurological injury.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2076-3425/10/12/1005/s1, Document S1: Using cranial CT to construct a headmodel in Nexstim NBS.

**Author Contributions:** Conceptualization, S.M.K. and P.E.T.; Data curation, S.S.; Formal analysis, S.S.; Investigation, S.S. and A.F.H.; Methodology, S.S. and P.E.T.; Project administration, P.E.T.; Resources, L.C. and P.E.T.; Supervision, L.C., S.M.K. and P.E.T.; Validation, S.S.; Visualization, S.S.; Writing—original draft, S.S.; Writing—review & editing, S.S., S.M.K., N.S. and P.E.T.; All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** We would like to thank the UCSF Biomagnetic Imaging Laboratory for their support of this study.

**Conflicts of Interest:** S.M.K. is a consultant for Brainlab AG (Munich, Germany) and for Nexstim Plc (Helsinki, Finland). N.S. received honoraria from Nexstim Plc (Helsinki, Finland).
