With the development of anesthesia technique, awake craniotomy surgery has been widely used for eloquent cortex mapping (Bennett et al., 1997). During the surgery, surgeons are able to locate the brain functional area using direct electrocortical stimulation (DECS) in order to accurately resect the lesion while minimizing post-surgical functional impairments. DECS is widely considered as the “golden standard” for intraoperatively delineating eloquent cortex, but it is a misconception that DECS allows us to draw unequivocal conclusions about the role of stimulated brain areas (Borchers et al., 2011). The DECS meets its burden because of time-consuming and high risk of inducing seizures (Nader et al., 2004). To further provide a high-quality brain surgery and overcome those unavoidable disadvantages, electrocorticographic (ECoG) signal changes associated with the function of specific cortical area has received substantial interest as an alternative approach for the eloquent cortex localization (Ikeda et al., 2002).

The recorded ECoG electrical potential at a broad range of frequencies is illustrated to be strongly correlated with brain activity. The frequency component normally comprises of slow cortical potential (e.g., movement-related cortical potentials [0.05–3] Hz), low frequency oscillation (mu [8–12] Hz, beta [18–26] Hz), and high frequency oscillation (>30 Hz). Each of these components has the potential for cortical function mapping as they can be used to quantify brain activity from a different point-of-view (Ikeda et al., 2002; Leuthardt et al., 2007; Miller et al., 2007b). Low frequency oscillations are produced by thalamo-cortical circuits and their amplitude decreases when brain is activated, such as when motor tasks are performed (Pfurtscheller and Silva, 1999). In contrast, high frequency changes are found to have an increase in amplitude with motor tasks and are more correlated with the activity of local neuronal populations (Manning et al., 2009). Specifically, high-frequency tends to have more concentrated spatial origin than low-frequency (Leuthardt et al., 2007). Within high frequency bands, several studies have shown that it was highly specific to cortical processing in motor tasks (Miller et al., 2007b). Clinically, Leuthardt et al. first used the term of electrocorticographic frequency alteration mapping (EFAM), including both low frequency band (LFB) and high frequency band (HFB) alteration, for the purpose of motor cortex regions mapping (Leuthardt et al., 2007). It has shown that HFB sites were more task specific, whereas LFB sites more sensitive in identifying functional cortex defined by DECS (Miller et al., 2007a; Breshears et al., 2010; Roland et al., 2010), and a combination of HFB and LFB is necessary for specific function localization, such as Broca localization (Wu et al., 2010).

However, high frequency activation is task specific and different tasks generate different activity patterns, e.g., power increases in gamma band was difficult to detect when performing only simple wrist movement (Aoki et al., 1999). Generally, in intraoperative condition, it is hard and time-consuming for patients to perform complex motor task, while a simple movement task, such as wrist extension and flexion, would be an ideal choice for corresponding motor cortex mapping. As no significant gamma activity was induced with simple wrist movement in our study, movement-related cortical potentials (MRCPs) were used for motor function localization.

Cortical potentials associated with voluntary movements are identified as MRCPs since the pioneering work of Bates (1951) and Vaughan et al. (1968). Components of MRCPs are related to the early preparation (Bereithschaftspotential or readiness potential), late preparation (negative slope, NS), initiation (motor potential, MP), and execution of the movement (movement-related response 1, MRR1) (Babiloni et al., 1999). Careful analysis of MRCPs enables us to evaluate the efferent function of the brain and higher brain functions controlling voluntary movements. MRCPs mainly depend on cerebellar-thalamus-cortical circuitry (Shibasaki et al., 1986; Tarkka et al., 1993), and can be used for primary and supplementary motor area function mapping (Ikeda et al., 2002). For different types of movements (e.g., hand, finger, foot, tongue protrusions, and vocalizations), MRCPs have a specific topographic distribution on the motor cortex, moreover, the distribution is consistent with the electrical simulation results in humans (Fried et al., 1981; Neshige et al., 1988; Ikeda et al., 1992, 1995a,b; Yazawa et al., 1997). These studies indicate MRCPs would be a clinically viable method for precise motor cortical mapping.

The MRCP is regarded as the amplitude changing of the [0.05–3] Hz slow cortical potential within the temporal domain. In this study, we defined a term, i.e., electrocorticographic temporal alteration mapping (ETAM), when applying the temporal information for eloquent cortex mapping. Although, MRCPs were selected as the ETAM signal feature in this paper, other researchers could apply specific event-related potentials as the ETAM signal feature for sensory, cognitive, or language cortex mapping.

ECoG-based mapping predominantly occurs in the epilepsy monitoring unit, remaining as an extraoperative endeavor. However, only a few attempts tried to use ECoG in the intraoperative settings instead of after-surgery monitoring for motor and speech/language cortex localization (Breshears et al., 2010; Roland et al., 2010; Ogawa et al., 2014; Taplin et al., 2016). The practicality and potential value of ECoG-based mapping in the operating room is still largely unexplored. In this study, the proposed ETAM method was tested and validated on patients undergoing intraoperative brain surgery, and compared with the results from clinically-proofed DECS method and the state-of-the-art EFAM method.

Contralateral wrist motion showed well-defined MRCP in all five subjects (Figure 2). In proximity to the EMG onset, we got a distinct focalization of the motor potential (MP) nearby the central sulcus, which overlaid the hand motor representation area localized by DECS. MRCP activity peaked at ~100 ms after the EMG onset. Every subject in this study had DECS-induced hand movements. Altogether, 11 DECS-positive electrodes were found as shown in (Table 3).

We analyzed the correlation among the three modalities for cortical mapping, i.e., DECS, ETAM, and EFAM. Electrodes were considered to be EFAM positive (EFAM+) if powers were significantly changed in the LFB, HFB, or both. The total number of electrodes, electrodes with functional responses, sensitivity, and specificity of each methods are summarized in Tables 4, 5. In general, no significant difference in sensitivity was found between ETAM and LFB methods, with 9 and 10 true positive electrodes, respectively. However, the ETAM was more specific compared with LFB. Explicitly, LFB for DECS-positive sites was 90.91% in sensitivity and 86.08% in specificity, while for the ETAM it was 81.82 and 95.36%, respectively. No significant high frequency alterations were found in sub_1, sub_3, and sub_4. There were just three channels having high frequency alterations in all five subjects. HFB for DECS-positive sites was 18.18% in sensitivity and 99.48% in specificity. Taking the LFB and HFB together, the sensitivity and specificity of EFAM positive sites for identifying DECS-positive sites was 90.91 and 86.08%, respectively.

Figure 3 (1–5) shows the activation maps of five subjects. A highly generalized response was found for motor tasks across individual subjects. Decreases were found in LFB while increases were found in HFB. Topographic distribution of ETAM results were more concentrated than the distribution of LFB power changes. HFB showed no somatotopical information for all subjects except sub_5. Further, analyzing the MRCP, LFB, and HFB for each single channel, results are shown in Figure 3 (6–10). Gamma band changes were not significant among subjects, and only chan_33 of sub_2 and chan_25, chan_17 of sub_5 were found with significant gamma changes. All MRCP significant channels showed significant low frequency band changes after the motion onset.

The chi-square independence test results showed that the ETAM, LFB, and HFB electrode distributions were significantly overlapped with the DECS-positive electrode distributions. Specifically, ETAM χ² = 68.08, LFB χ² = 36.68, HFB χ² = 11.94 (α = 0.001, χ² = 10.83). Cross-validation for ETAM was done for all possible combinations of the test-template and test-data (i.e., to test one subject, the test-template was the mean of the representative grand average MRCP of the other four subjects). Cross-validation results showed that ETAM for DECS-positive sites was 81.82% in sensitivity and 94.33% in specificity.

In this work, a novel ETAM method targeting at intraoperative brain surgery was proposed and validated. Neurosurgeons could apply ETAM as an alternative technique to identify the motor cortex during awake craniotomy. Over the past decade, researches pay great attention to understand how brain electrical signals (i.e., EEG and ECoG) correlate to motor, sensory, and language functions. Generally, three territories were investigated for function localization. The first one is the utilization of spectral alteration, such as ERD/ERS (Miller et al., 2014), LFB/HFB (Miller et al., 2007b), and most recently, the broadband spectral changes (Miller et al., 2009, 2014); The second one is the utilization of temporal alteration, which is disclosed by averaging EEG/ECoG data with respect to the task onset from the background oscillatory, such as movement-related cortical potentials which are proved as a promising feature for function localization (Ikeda et al., 2002); the third one is the utilization of signal coherence, which depends on the power and phase dynamics among different frequency components from one electrode (e.g., amplitude phase coupling Fries, 2005). For the clinical application, Leuthardt et al. firstly used the term of electrocorticographic frequency alteration mapping (EFAM) for the purpose of motor cortex localization (Leuthardt et al., 2007), which belongs to the first territory. In this study, we gave the proposed MRCP-based mapping method as ETAM which belongs to the second territory. ETAM results were consistent among different subjects, with sensitivity varying from 66.67 to 100% and specificity varying from 88.89 to 100%. Our results indicated that ETAM would be a clinical and technical useful and reliable method.

This study confirmed the previously proposed notion that ECoG MRCPs can be a useful tool for eloquent cortex mapping. Motor cortex regions defined through ETAM in our study correlated well with regions that defined through DECS. MRCPs reflect discharges from large populations of neurons linking specific aspects of sensory and cognitive processing. Scalp MRCPs are susceptible to the spatially integrating properties of cerebrospinal fluid, skull and skin. Compared with scalp MRCPs, intracranial MRCPs (iMRCPs) extracted from ECoG have much larger (up to 150 microvolt) voltage and higher spatial resolution. iMRCP components are often visible in very few adjacent electrode contacts and span a few square centimeter with steep amplitude gradients, suggesting close proximity to the cortical generators (Breshears et al., 2010; Roland et al., 2010; Ogawa et al., 2014; Taplin et al., 2016). With a systematical analysis of MRCPs, researchers can understand models of these cognitive processes and locate the brain regions that implicate specific cognitive processes. We extracted large MRCP amplitudes (up to 200 uV, Figure 1) directly from the cortex surface, and had a good cross-validation sensitivity (81.82%) and specificity (94.33%) for identifying sites with positive DECS hand motor responses. These anatomic data were acquired in several minutes without the risks of seizure and after discharges caused by DECS. Especially, all experiments were performed intraoperatively which confirmed its clinical viable possibility. ETAM method proposed in this study provided a enormous amount of useful and low risk information. Regarding to the close correlation of the ETAM with DECS maps and short time requirements for information accrual, ETAM could be a powerful adjunct method for delineating the cortex in conjunction with DECS.

Many studies show that complex actions are associated with more robust brain responses than simple actions. Functional near-infrared spectroscopy (fNIRS) and functional magnetic resonance imaging (fMRI) studies indicate that complex tasks are associated with greater hemodynamic changes than simple tasks (Kuhtz-Buschbeck et al., 2003; Wei and Luo, 2010; Lisa Holper, 2011). Although, the relationship between the neural activity types and the oxygen level is complicated, cerebral local oxygenation are closely related to the neural activity (Lenkov et al., 2013; Chernov et al., 2016). By using transcranial magnetic stimulation (TMS), higher motor-evoked potential (MEP) amplitude is found with more complex task, indicating an increased corticospinal excitability with increased task-complexity (Kuhtz-Buschbeck et al., 2003; Roosink and Zijdewind, 2010). Electroencephalography (EEG) studies show that sensorimotor rhythms (SMR) elicited during complex actions can be more reliably detected comparing with simple actions (Gibson et al., 2014). ECoG studies have found no evidence for increases in gamma power in subjects discriminating the strength of a somatosensory stimulus (Menon et al., 1996). Aoki et al. (1999) indicated that different tasks generate different ECoG patterns, and negligible increases in gamma power occur with simple wrist movement. Similarly, this task-dependent changes phenomenon in gamma activity has been investigated in non-human primates (Sanes and Donoghue, 1993; Baker et al., 1997; Donoghue et al., 1998). We further confirmed this task-dependent changes phenomenon intraoperatively. With simple wrist extension movement, gamma changes were rarely found (further analyzing these data, only chan_33 of sub_2 and chan_25, chan_17 of sub_5 were found with significant gamma changes), and the gamma sensitivity for identify DECS-positive sites was only 18.18% for all five subjects. Thus, in specific activation, i.e., simple wrist movement, using gamma changes for motor cortex location was found to be less valid.

The relation between low frequency ([8–30] Hz) ERD and MRCPs is controversial. ECoG recordings from primary sensorimotor (M1-S1) show that MRCP and ERD responses originate in similar cortical regions and share some common timing features, but the magnitude and spatial distribution of the two responses appear to be independent (Toro et al., 1994). In an EEG study, alpha ERD reflects changes in wide cortical sensorimotor areas, whereas MRCPs represent mainly task-specific responses of the supplementary motor area (SMA) and contralateral M1-S1 (Bittar et al., 1999). Results of this study indicated that LFB and MRCPs responded in similar cortical regions, but LFB seemed to have a broader distribution than MRCPs. Sensitivity of the MRCPs and LFB for identifying DECS-positive sites had no obvious difference, but MRCPs tended to be more specific (94.33%) than LFB (86.08%).

The intraoperative condition is stricter than the extraoperative condition. Patients are always anxious and frail, and the time is limited for safety reasons. More complexity in tasks means more conscious and active involvement from patients. But more conscious means more time for awaking from anesthesia. Generally, complex tasks can only be performed with well-cognitive ones. Elderly person, children, and patients with cognition disorders would be extremely difficult to perform the required complex tasks. To reduce the cognitive complexity, patients were instructed to execute single wrist extension cued by the surgeon many times without other tasks between trials, and all patients could easily and successfully perform this intraoperative wrist motion task.

In order to further verify clinical applicability of ETAM, cross-validation was done for all possible combinations of the test-template and test-data, and the results was 81.82% in sensitivity and 94.33% in specificity. Comparing with the results using the template which was defined by averaging the representative MRCP chosen from all five subjects (Figure 1), the cross-validation results were the same in sensitivity (81.82 vs. 81.82%) while slightly decreased in specificity (94.33 vs. 95.36%). Thus, we only needed to calculate the template once, which was clinically practicable and was a promising way to shorten the operation time. The sensitivity for identifying DECS-positive sites may be improved when the template is calculated from the same subject, but it will need additional trials and may not be the optimal choice when taking limited operation time into consideration. The movement onset was visually defined in this study in order to exclude the noisy trials, but it could be automatically defined by estimating the onset when the rectified EMG signal amplitude crossed a certain threshold (e.g., set threshold as the one tenth of maximum EMG amplitude) (Niazi et al., 2011). However, using the EMG amplitude for template onset definition was an indirect method. There may have significant EMG changes without significant MRCP because of additional cognitive process or unknown noise. For rapid and reliable function mapping, a movement detector based on MRCP is needed for real-time detection of movement initiation. Prior studies of our groups showed that the single trial MRCP induced by upper and lower limbs could be detected in real-time, which were based on EEG study and aimed for a closed-loop brain-computer interface application. In these studies LPP-LDA (Jochumsen et al., 2015) and classic matched filter (Jiang et al., 2015) were applied as MRCP detector, and we could use similar strategies in a further work toward a real time function mapping based on ETAM method.

To obtain a friendly interactive interface and present the ETAM results, we applied a weighted spherical kernel to each electrode, and projected the summed results on a grid map with electrode number [Figure 3 (1–5)]. This procedure allowed surgeons to intuitively appreciate the eloquent cortex associated with the relevant function.

Recording reliable ECoG signal is not easy and is one of the main limitation hindering its clinical application, especially in the operation room where lot of surgical instrument produce unknown noise. For example, the spectral power and MRCP amplitude of sub_4 was much smaller than other subjects [Figure 3 (9)]. Further analyzing the raw data, we found an extraordinary large 50 Hz power-line interference which even led to the amplifier saturation some times. However, some clear MRCPs trials could still be found in sub_4, and the contaminated trials were visually excluded. By analyzing 19 uncontaminated trials, the ETAM for DECS-positive sites was 66.67% in sensitivity and 97.67% in specificity, while for the LFB it was 66.67 and 90.7% respectively. ETAM had better specificity than LFB, indicating that ETAM may be more robust in noisy environment.

Dalal et al. holds the view that it is not wise to restrict our understanding of brain dynamics exclusively to higher frequency content, and the brain dynamics need to be examined across several frequency bands (Dalal et al., 2011). In specific condition of this study, gamma band alteration was negligible. The temporal and spectral alteration indicates different brain physiological mechanisms, which may represent different aspects of motor cortex activation. In this study, ETAM method contained [0.05–3] Hz information while EFAM method contained LFB ([8–32] Hz) and HFB ([66–90] Hz) information. For motor cortex mapping, the ETAM and EFAM methods shared some common positive electrode results while still had some opposite results, e.g., chan_2 of sub_1 was ETAM-positive but EFAM-negative. These opposite results may provide additional information to the neurosurgeons. Clinically, both ETAM and EFAM positive electrodes might be regarded as functional-convinced sites, and electrodes with opposite results might be regarded as functional-high-risk sites, as shown in Figure 4. The neurosurgeons need to protect the functional-convinced sites while be careful about the functional-high-risk sites, and additional DECS procedure might be performed on these functional-high-risk sites for function confirmation. As mentioned by Towle et al. that a formidable challenge will be to discriminate between necessary and ancillary areas of cortical function, which are probably not discrete entities, but rather points along a continuum (Towle et al., 2016). When combining the results of ETAM and EFAM together, the sensitivity and specificity for identifying DECS-positive sites were 90.90 and 84.53%, respectively. The combining sensitivity was the same comparing with EFAM (90.90%), but the specificity was lower than EFAM (86.08%) and ETAM (94.33%). Thus, combining the two methods might not improve the sensitivity and specificity, but might be a promising way to separate the functional area into necessary and ancillary areas. Technically speaking, when setting the signal sampling amplifier with a DC analog cut off, researchers can obtain both fine MRCP and LFB/HFB signal.

In this work, a novel ETAM method based on MRCP was proposed for intraoperative mapping. The feasibility and resolution of the proposed ETAM mapping was compared with clinically available DECS approach which requires active cortical stimulation. Due to the passive mapping feature of the ETAM methodology, the proposed method would significantly reduce the risk of inducing seizures during surgery. Moreover, by comparison with the state-of-the-art EFAM method, ETAM demonstrated a better performance in specificity while showed a comparable performance in sensitivity in condition of simple task. For clinical application, as the temporal and spectral alteration indicates different brain physiological mechanisms, combination of the ETAM with EFAM would provide neurosurgeons with more reliable information for a better operation planning.

This study was carried out in accordance with the recommendations of Ethics Committee of Huashan Hospital, Fudan University with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Ethics Committee of Huashan Hospital, Fudan University.

LC and YM conceived the study. LY, DZ, XS, and LC designed the experiment. ZW, TX, and LC conducted the experiments. ZW and TX analyzed the data. ZW, TX, and LY co-drafted the paper. DF, YM, and XZ coordinated the work.