Next Article in Journal
OCTA and Doppler Ultrasound in Primary Open-Angle Glaucoma and Normal-Tension Glaucoma
Next Article in Special Issue
Quantitative Evaluation of Oxygen Extraction Fraction Changes in the Monkey Brain during Acute Stroke by Using Quantitative Susceptibility Mapping
Previous Article in Journal
Alteration in Levels of Specific miRNAs and Their Potential Protein Targets between Human Pancreatic Cancer Samples, Adjacent Normal Tissue, and Xenografts Derived from These Tumors
Previous Article in Special Issue
Altered Language-Related Effective Connectivity in Patients with Benign Childhood Epilepsy with Centrotemporal Spikes
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Life
Volume 13
Issue 3
10.3390/life13030609
life-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Review

Pre-Operative Functional Mapping in Patients with Brain Tumors by fMRI and MEG: Advantages and Disadvantages in the Use of One Technique over the Other

by
Elisa Cargnelutti
* and
Barbara Tomasino
Scientific Institute, IRCCS E. Medea, Dipartimento/Unità Operativa Pasian di Prato, 33037 Udine, Italy
*
Author to whom correspondence should be addressed.
Life 2023, 13(3), 609; https://doi.org/10.3390/life13030609
Submission received: 6 February 2023 / Revised: 17 February 2023 / Accepted: 20 February 2023 / Published: 22 February 2023
(This article belongs to the Special Issue New Advances in Neuroimaging and Brain Functions)
Browse Figures
Versions Notes

Abstract

:

Simple Summary

Providing an accurate localization of the brain functions before proceeding with neurosurgery is of vital importance to spare the removal of crucial brain areas. In this review paper, we investigated how two techniques, namely functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG), can serve this purpose, with specific reference to pre-operative assessment of sensorimotor and language functions in patients with brain tumors. First, we explored the methodological strengths and issues of each technique from data acquisition, to data processing and result display, also providing a temporal overview of these technical aspects and potential strategies developed to overcome the main issues. Then, we identified 16 studies that made pre-operative assessments by both techniques and accurately scrutinized them. Overall, despite the potential limitations associated with one technique or the other, the majority of these studies underlined the reliable and complementary use of fMRI and MEG. For these reasons, with the aim of the most reliable pre-operative assessment as possible, we recommend the combined use of both techniques, favored by the recent technological advances making MEG use more feasible.

Abstract

Pre-operative mapping of brain functions is crucial to plan neurosurgery and investigate potential plasticity processes. Due to its availability, functional magnetic resonance imaging (fMRI) is widely used for this purpose; on the other hand, the demanding cost and maintenance limit the use of magnetoencephalography (MEG), despite several studies reporting its accuracy in localizing brain functions of interest in patient populations. In this review paper, we discuss the strengths and weaknesses of both techniques, from a methodological perspective first; then, we scrutinized and commented on the findings from 16 studies, identified by a database search, that made pre-operative assessments using both techniques in patients with brain tumors. We commented on the results by accounting for study limitations associated with small sample sizes and variability in the used tasks. Overall, we found that, although some studies reported the superiority for MEG, the majority of them underlined the complementary use of these techniques and suggested assessment using both. Indeed, both fMRI and MEG present some disadvantages, although the development of novel devices and processing procedures has enabled ever more accurate assessments. In particular, the development of new, more feasible MEG devices will allow widespread availability of this technique and its routinely combined use with fMRI.

1. Introduction

The accurate identification of the brain areas subserving a given brain function is a crucial process in neurosurgery planning. In recent decades, studies have adopted different non-invasive techniques for this purpose to primarily identify potential lesion proximity to eloquent areas of the language and sensorimotor systems. Additionally, due to extensive availability of magnetic resonance imaging (MRI) instrumentation for diagnostic purposes, functional MRI (fMRI) probably represents the most widely used technique. Besides fMRI, electroencephalography (EEG) and magnetoencephalography (MEG) have been employed too, although traditionally deputed to epileptic activity monitoring [1,2].
The principle of fMRI resides in the blood flow change associated with the hemodynamic response in brain areas involved in a given brain function; this generates a detectable signal, which is measured in terms of a blood-oxygen-level dependent (BOLD) contrast. On the other hand, EEG and MEG measure the brain activity directly, by recording, respectively, the electrical and magnetic fields generated by thousands of neurons that activate in synchrony.
EEG and MEG provide similar but not completely equivalent information about brain processes [3] and each technique has its advantages over the other. EEG recordings are more feasible, whereas recording by MEG is tricky and probably represents its major issue. The MEG apparatus requires distinguishing the brain’s magnetic activity (in the order of 10−15 T) from external interferences, especially the much stronger Earth’s magnetic field (10−5 T). Despite the need for a shielded room, super-conducting quantum interference detectors (SQUIDs) are also required and need a continuous cooling process by liquid helium (−270 °C). These factors make traditional MEG systems very demanding, being very bulky (with a weight of several tons) and very expensive, not only for the initial set up, but also for their maintenance.
Nevertheless, overall MEG is considered superior to EEG, because it normally has higher spatial accuracy related to the fact that magnetic versus electric fields are less distorted by the skull, scalp, and cerebrospinal fluid and because it better detects activities located in the gyri [4]. Further, it requires fewer neurons to be activated in synchrony to detect brain activity and source modeling, required to localize the detected field sources, is generally less complex and more accurate (for a comparison between EEG and MEG, see [5,6]).
In light of these technical considerations, in the current review paper we aimed to compare fMRI and MEG in terms of brain activity localization, from the perspective of surgical planning for brain tumor resection (Figure 1). Indeed, as well as fMRI, when MEG data are merged with high-resolution MR images, the resultant magnetic source imaging (MSI) can provide accurate localization of the detected signal sources and enable functional intra-operative navigation to guide safe tumor resection. However, this combined approach is poorly used for several reasons, first the poor availability of the MEG apparatus in clinical services.
Several studies have independently addressed the accuracy of fMRI and MEG in the localization of cognitive and sensorimotor functions in patient populations, for instance comparing the results from those obtained by intra-operative mapping. Localizing the functional activations associated with a given function is crucial, because tumor development can cause brain remodeling through neuroplasticity processes and activations can be displaced with respect to the healthy brain (see [7,8]). Identifying the potentially new location of these activations is fundamental to spare the removal of essential brain areas during surgery.
Evidence of reliability in the use of fMRI or MEG alone has been widely reported. A recent meta-analysis [9] reported the accuracy of fMRI in language mapping by taking intra-operative recordings via direct cortical stimulation (DCS) as a reference. Here, the reported sensitivity ranged from 51 to 80% and specificity 25 to 82%. Bizzi and colleagues [10] suggested that accuracy can by influenced by glioma grade, as they observed a remarkably lower sensitivity in glioblastomas (i.e., high-grade tumors) than in low-grade gliomas (65 vs. 93%), although specificity was higher (93 vs. 79%). Concerning MEG, Ellis et al. [11] commented that studies did not typically report a quantitative assessment of accuracy in terms of sensitivity and specificity; nevertheless, this technique has been widely observed to provide a reliable assessment of pre-operative localization of brain functional activity [12,13,14,15].
In the current review, we will first provide an overview of the methodological strengths and weaknesses of each technique, to then compare them more directly by commenting the results reported in studies that adopted both of them for pre-operative brain activity assessment. The final aim was to shed light on the advantages and disadvantages in applying one methodology versus the other and the related recommendation about their use in clinical practice.

2. Methodological Comparison between fMRI and MEG

Overall, agreement exists in relation to the pros and cons associated with each technique, regardless of the specific findings observed in the pertaining studies. Apart from the factors representing limitations for the acquisition procedure, for instance the presence of metallic implants in patients (which may prevent fMRI acquisition and cause severe artifacts in the case of MEG), the two techniques present characteristics that influence the quality of the measured brain activity that can either represent a drawback or a strong point for that technique.

2.1. Approach to Measuring the Brain Activity

Executing a given task activates the related brain network, with all the recruited areas simultaneously displayed by fMRI. On the other hand, MEG records the temporal development of the whole brain process, hence its spatiotemporal fingerprints; it detects the earliest component of the evoked response arising from primary eloquent areas, which can be separated from later components mainly arising from non-primary areas [16]. This is generally considered as an advantage over fMRI, given that it enables distinguishing between primary activated areas and areas that secondarily support that function. Nevertheless, this is positive, as fMRI can more easily capture the whole network associated with a specific brain function and investigate how the brain has been reshaped to compensate for glioma growth.

2.2. Spatial and Temporal Resolution

Concerning resolution, the two techniques can be considered as complementary. The advantage of fMRI relies in the high spatial resolution of a few millimeters (depending on the voxel size), although a drawback is represented by its poor temporal resolution; indeed, BOLD responses are detected after several seconds following the activation of a given brain region and also lasts several seconds after deactivation.
On the other hand, traditional MEG is characterized by a higher temporal resolution (in the order of milliseconds) but a poorer spatial resolution, especially when the source of the detected signal is deep (for reviews, see [17,18,19]). Even when the source is highly localized, resultant images are typically blurred. Spatial resolution is not homogeneous across the brain and can be affected by several factors, including distance from the sensors and signal-to-noise ratio [20]. Poor spatial resolution is primarily determined by the fact that the number of sensors covering the brain is finite, therefore localizing the source that generated the detected signal is generally problematic (see afterwards). This is particularly characteristic of earlier MEG devices, only possessing a few channels covering a limited portion of the brain.
Furthermore, classical MEG systems could not detect subcortical activity; in fact, MEG mainly detects relatively superficial currents in the fissural cortex, being sensitive to sources tangential to the skull; nevertheless, more recent whole-head systems with hundreds of sensor can cover the whole head and can detect activity even in subcortical regions [21]. However, these systems still have limited coverage of prefrontal areas and the cerebellum, limiting the detection of brain activity in these areas [20].
In the case of presurgical mapping, having high spatial resolution is fundamental, making fMRI more appropriate for this purpose, at least in general terms. In the following paragraphs, we will compare and discuss the in-depth suitability of both techniques from the perspective of neurosurgery.

2.3. Issues with Signal Detection and Processing

One of the major drawbacks associated with fMRI is represented by the inflow effect. In reality, BOLD signal may be altered by the signal coming from the blood flow in the draining veins, for instance in those supplying the central sulcus; the result is an alteration in neurovascular coupling. This was especially observed when using the conventional 1.5 T system with a gradient-echo sequence [22,23]. Further, the signal can be affected in the presence of brain lesions such as arteriovenous malformations or brain tumors and associated edema. In these cases, lesion removal is likely to restore signal detection [23]. In the case of tumors, the effect is attributed to angiogenesis and altered permeability of the tumor vasculature, especially for high-grade lesions [24]. In the case of low-grade gliomas, the process can be attributed to angiogenesis as well, but also to cellular dysfunction associated with tissue infiltration [25,26].
Nevertheless, these issues are scaled down by scans performed at high magnetic-field strength (3 T or greater) and the use of a spin-echo or gradient-echo EPI sequence instead of the traditional gradient-echo sequence. These solutions, indeed, decrease the effect of the inflow effect and then increase the BOLD contrast associated with the measured brain process [22,23,27,28]. This suggests that signal quality is highly influenced by both acquisition devices and acquisition and processing procedures.
Acquisition devices also influence signal detection by MEG. Sensors for magnetic signal detection are typically represented by gradiometers (axial or planar), although these are rather insensitive to deep brain sources. The latter can be, instead, collected by magnetometers, but these are also more sensitive to ambient noise. Modern devices include both gradiometers and magnetometers and their combination significantly increases the signal-to-noise ratio (for reviews, see [19,21,29,30]).
For both fMRI and MEG, signal detection may be challenging in correspondence to specific brain structures. For instance, partial-volume effects in fMRI can hamper signal detection when activation is confined to one bank of the sulcus [30]. Conversely, MEG can fail in the detection of an extended signal, when it originates from opposite banks of a sulcus [31]. In such cases, using MEG and fMRI could, instead, navigate around the problem.
Concerning signal processing, the main issue concerns MEG and, specifically, the solution to the inverse problem for signal localization; in other words, after recording the magnetic current on the scalp, it is necessary to identify the correct location of the brain source that generated it. The inverse problem is defined to be ill-posed, as there is not a unique solution to it. The most commonly used source estimation approach is represented by the equivalent current dipole (ECD), which captures the activity of tens of thousands of cortical neurons that activate in synchrony. A single ECD source can represent a valid solution in the case of strong dipolar patterns with a high signal-to-noise ratio, as is the case of the early components of somatosensory-evoked fields [21], but is limited otherwise.
Alternative approaches have been developed to provide a more reliable resolution to the inverse problem; these include spatial filtering and appear even more useful when analyzing widespread activation patterns such as those related to language functions [32]. Some authors, however, have underlined that defining the actual extent of a source can still remain problematic [18]. A widely used spatial-filtering approach is represented by beamforming [33,34], which excludes signals external to the source of interest (however, with difficulty in the case of highly synchronous sources). Source estimation can also be performed by distributed inverse methods, including minimum-norm estimation (MNE, [35]), dynamical statistical parametric mapping (dSPM, [36]), and standardized low-resolution brain electromagnetic tomography (sLORETA, [37]).
Nevertheless, even signal detection by fMRI has to be evaluated, and this can vary depending on the software used for processing and the selected options. For instance, different basis functions can be used to model the hemodynamic response (e.g., hemodynamic response function, Fourier set, finite impulse response), potentially giving different results.
Finally, another relevant issue pertains to the spatial co-registration procedure to the anatomical MR image. This is likely to be more challenging and less accurate for MEG than fMRI [38], even though appropriate procedures are recommended for fMRI, in order to avoid misalignment [39,40]. Indeed, whereas the problem with the latter may be represented by potential distortion due to susceptibility effects, co-registration with MEG is highly prone to human error. Reported co-registration errors for MEG are typically about 5 to 10 mm [41,42], although recently developed technological advances and algorithms can decrease errors to a few millimeters [43]. The most traditional approach was initially represented by the registration to reference points represented by anatomical landmarks (i.e., nasion, inion, and bilateral pre-auricular points). However, by using specific algorithms, such as contour fitting, accuracy can be increased [38,44].
All these concerns have to be weighted in light of clinical requirements and timing. Indeed, although very promising, these approaches are still time-consuming and it is sometimes difficult to integrate them into the clinical routine, considering also that patients can become tired more easily following intensive pre-operative assessments. For this reason, the best practical solution should be, generally, the one that provides the most reliable assessment within a reasonable time.

3. Direct Comparison between fMRI and MEG in Pre-Operative Mapping

Following this methodological comparison between the two approaches, we now comment on the findings from studies assessing pre-operative functional activation in patients with brain tumors by means of both fMRI and MEG. In order to identify peer-reviewed papers on this topic, we ran a paper search on the PubMed, Web of Science, and Scholar databases using the following keywords: (fMRI OR “functional MR”) AND (MEG OR magnetoencephalography OR MSI OR “magnetic source”) AND (glioma OR tumor OR lesion) AND (surgery OR preoperative).
We identified 16 studies, reporting findings from a highly variable number of patients (in the range of one to 90 patients). Sample sizes were generally small, with a median of 12 subjects (Figure 2A). The study results are reported in Table 1 and Table 2 in chronological order, from 1995 till present, with the purpose of providing a temporal evolution of the use of the two techniques. Indeed, improvements in both devices, processing algorithms and methods have enabled the increased reliability of the results from each approach. As commented on before, the improved reliability for fMRI can be attributed to the increased tomograph magnetic field strength (1.5 T vs. 3 T), the use of different EPI versus the classical gradient-echo sequence, and increased spatial resolution, with a consequent reduction in partial-volume effects. Concerning MEG, the reliability of solutions to the inverse problem have increased with the progressive use of whole-head devices, possessing an increased number of sensors, by combining both magnetometers and gradiometers, and by processing data by using alternative methods to single ECD.
The reliability of the findings from each technique was most frequently derived from the comparison with intra-operative mapping [11,23,28,44,45,46,48,49]. Alternatively, the confirmation of a good pre-operative assessment was given in the case of successful imaging-guided resection [47,50,51,53,54,55] and/or when the findings matched with those achieved by other techniques, such as the Wada test (in the case of laterality assessment; [47,48,50]) or stimulation methods including transcranial magnetic stimulation (TMS; [51]). Two studies had slightly different purposes than pre-operative planning. In one study [52], pre-operative mapping was performed to help plan the appropriate dose of chemotherapy and, in the other [56], to assess post-operative neuroplasticity changes associated with each language in bilinguals.
The second variable was the domain assessed. There was a dissociation between the frequency of motor and sensory tasks: a higher number of fMRI studies, compared to MEG studies, used motor localizers; vice versa, a higher number of MEG studies, compared to fMRI studies, used sensory localizers (Figure 2B). In contrast, the frequency of language assessments was comparable between fMRI and MEG studies. We observed that the earliest studies (up to 2003) focused exclusively on the central sulcus/sensorimotor cortex localization. The first studies investigating language were published in 2006 [47,48]; interestingly, all the selected studies focusing on language explored expressive language functions, although MEG was observed to be inaccurate at assessing these functions [57]. Overall, studies found general agreement between the results achieved by the two approaches independently.
Nevertheless, several cases of discrepancy were reported. Excluding partially different localizations attributable to the use of different tasks for the fMRI and MEG assessments (and, in case, for intra-operative mapping, too; see Figure 2C,D), the authors attributed the discrepancies to several factors. Most frequently, the ability to accurately detect functional activation was lower for fMRI, mainly represented by altered neurovascular coupling due to the close presence of a draining vein or edema associated with the pathological condition [23,28,55]. Pathology represented a notable limitation for fMRI. Inoue et al. [23] observed fMRI inaccuracy for the patients but not for the healthy controls and the possibility of restored accurate localization following surgery. In one case [47], a decreased BOLD contrast was observed in the vicinity of large and/or high-grade lesions.
In another case [49], fMRI failed in motor activation localization in the pre-central gyrus, but localized the source in the post-central gyrus. The authors commented that one potential limitation of this technique resides in its capacity to detect non-primary areas recruited to perform a given task [23], whereas MEG can discriminate between activation in primary and association areas. However, another study [53] observed the same phenomenon for MEG, too. Given that these authors carried out tumor resection based on pre-operative mapping and that the patient did not develop post-operative motor disturbances despite complete pre-central gyrus excision, it is crucial to understand whether apparently erroneous displacement should be interpreted, rather, in terms of functional reallocation of the investigated function in some case.
On the other hand, a few studies reported issues related to MEG acquisition. In detail, MEG could not be performed or provided poor results in cases of hemiparesis [44,45]—although fMRI could still show a residual activation [45]—or was difficult to execute with regularity for the proposed task [47]. In other cases, it produced severe artifacts due to the presence of dental implants [44,54].
Finally, we observed that, in three studies, pre-operative assessment by fMRI–MEG combination was further improved by the integration with relatively recent techniques, such as TMS [51] or diffusion-weighted/diffusion-tensor imaging [46,50,51]. TMS was used to confirm findings and to detect, within the fMRI functional network, areas that were essential to perform language tasks. Diffusion-weighted/diffusion-tensor imaging, on the other hand, was used to identify the integrity and potential dislocation of the subcortical connections in lesions involving the white matter; moreover, they helped in interpreting the functional imaging results. Several studies reported reliable and valuable pre-operative assessment by both TMS [58,59] and diffusion-weighted/diffusion-tensor imaging [60,61], claiming for their additional contribution and recommended use, at least when feasible.

Quantitative Comparison

A few studies provided a quantitative computation of potential discrepancies between fMRI–MEG localizations. Kober et al. [45] reported comparable sensorimotor area localizations by fMRI and MEG, although the Euclidean distance between them was not considered to be negligible, both for the sensory (15 ± 5 mm) and motor (10 ± 5 mm) activation. In Korvenoja et al. [49], who similarly aimed to localize the central sulcus, the Euclidean distance was inferior, with MEG dipoles located within 1–6 mm from the nearest fMRI activation voxel and 6–36 mm from the fMRI maximal z-score voxel. Zimmerman et al. [54] computed the Euclidean distance by adopting distributed inverse methods instead of single ECD and observed that, in 40% of the case of discrepancy, the mean distance was 3.8 mm for sensory fields with sLORETA and 7.4 mm for motor fields with MNE. In a subsequent study [55], the authors focused on language functions and used dSPM; they reported a mean Euclidean distance of 10.58 mm when taking into account the activation maxima and 9.06 mm for centroids. Although the first two studies underlined discrepancies in these findings, the latter concluded that, overall, there was a good correspondence between the pertaining results.
A thorough evaluation of accuracy based on the comparison with intra-operative mapping was carried out by Ellis and coworkers [11]. The authors computed the Youden’s J statistic [62], which combines sensitivity and specificity to estimate the overall accuracy of each approach. The statistic was computed at several distances between the fMRI/MEG activation and the DCS point. In detail, the authors found the highest J statistic for both motor and language fMRI at a 5 mm distance (J = 0.21 and 0.28, sensitivity = 0.34 and 0.31, specificity = 0.87 and 0.97, respectively) and for MEG at a 40 mm distance for motor mapping (J = 0.28, sensitivity = 0.86, specificity = 0.42) and a 15 mm distance for language mapping (J = 0.28, sensitivity = 0.56, specificity = 0.71). At the standard 10 mm distance used in the literature, there was a poor overlap between the localizations detected by the two approaches. However, their combination increased the accuracy for both the motor (J = 0.29, sensitivity = 0.62, specificity = 0.67) and language (J = 0.40, sensitivity = 0.69, specificity = 0.71) localizations. Overall, for motor mapping, the maximum J statistic (i.e., 0.35, sensitivity of 0.55, specificity of 0.79) was achieved using a 5 mm distance threshold for fMRI and a 15 mm distance threshold for MEG. For language mapping, this was achieved at the above-reported 10 mm threshold.
First, these findings suggest that when interpreting the results in reference to direct stimulation points, the distance threshold has to be taken into account. Depending on this distance and on the monitored function (motor or linguistic), either fMRI or MEG can be more accurate overall and that is necessary to find a compromise between sensitivity and specificity. Second, in light of these considerations, a combination of the two techniques appears as the best solution to assure reliable findings.

4. Conclusions

The gold standard in brain tumors surgery is represented by the awake procedure, during which an intra-operative mapping of the function of interest is performed. Nevertheless, for those patients for whom this procedure is not feasible or refused by the patients themselves, it is fundamental to achieve the most accurate function localization in order to prevent permanent post-operative sequelae. Adopting recent devices and processing procedures and algorithms is essential to increase the pre-operative mapping reliability. Additional cautions should be used. For instance, in order to reliably detect residual activity, especially in the case of fMRI, it is useful to adopt individual thresholds (i.e., by adjusting the p value when needed) when analyzing the data. However, this could be still insufficient to ensure a reliable pre-operative assessment [63]. In general, relying on a single technique might be risky in the case of surgery under general anesthesia, although, in combination with the assessment of the actual functional status of the patient, it could provide a general guideline for surgery planning.
In fact, in light of these limitations but also the mapping potential of both techniques, the majority of the included studies, both earlier [45,46,47,48] and more recent [11,51,52,53,54,55,56], underscored their complementarity: using both of them could enable a complete and reliable view of the brain activity. For instance, Kamada et al. [48] proposed the complementary use of both techniques to investigate language functions. In fact, MEG versus fMRI seemed to better detect temporoparietal activity (associated more with receptive language), whereas in several studies it seemed to less accurately detect signals from the frontal areas (associated more with expressive language, see [57]).
In the introductory part, we commented on how the availability of MEG devices is very limited worldwide, due to their cost and maintenance burden. However, more practical and economical MEG devices have been recently developed, which could increase MEG feasibility. For instance, optically pumped magnetometer (OPM) devices seem promising, although recording is still constrained to a limited signal bandwidth [20]. Importantly, they do not require expensive cryogenic components for maintenance. This system has reached a sensitivity comparable to that of SQUID sensors, but with increased signal power given that the sensors can be directly placed over the scalp. Alternatively, fMRI should be combined with other techniques; for instance, pre-operative assessment by fMRI could be confirmed by using non-invasive stimulation with more feasible TMS [51,58,59]. Further technological improvement will facilitate, in the near future, the distributed access to user-friendly devices which will enable even more accurate patient-tailored pre-operative surgical planning.

Author Contributions

E.C.: Conceptualization, investigation, data curation, writing—original draft preparation, writing—review and editing; B.T.: Conceptualization, writing—review and editing, supervision, validation. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the Italian Association for Cancer Research (AIRC)—Grant No.: I.G. 2017 Id.20387.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The authors accessed PubMed, Web of Science, and Scholar databases up to the end of 2022.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Bast, T.; Oezkan, O.; Rona, S.; Stippich, C.; Seitz, A.; Rupp, A.; Fauser, S.; Zentner, J.; Rating, D.; Scherg, M. EEG and MEG source analysis of single and averaged interictal spikes reveals intrinsic epileptogenicity in focal cortical dysplasia. Epilepsia 2004, 45, 621–631. [Google Scholar] [CrossRef] [PubMed]
  2. Chowdhury, R.A.; Zerouali, Y.; Hedrich, T.; Heers, M.; Kobayashi, E.; Lina, J.M.; Grova, C. MEG-EEG Information fusion and electromagnetic source imaging: From theory to clinical application in epilepsy. Brain Topogr. 2015, 28, 785–812. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  3. Malmivuo, J. Comparison of the properties of EEG and MEG in detecting the electric activity of the brain. Brain Topogr. 2012, 25, 1–19. [Google Scholar] [CrossRef] [PubMed]
  4. Hari, R. Magnetoencephalography: Methods and applications. In Niedermeyer’s Electroencephalography: Basic Principles, Clinical Applications, and Related Fields; Schomer, D.L., Lopes da Silva, F.H., Eds.; Lippincott Williams & Wilkins: Philadelphia, PA, USA, 2011; pp. 865–900. [Google Scholar]
  5. Cohen, D.; Cuffin, B.N. Demonstration of useful differences between magnetoencephalogram and electroencephalogram. Electroencephalogr. Clin. Neurophysiol. 1983, 56, 38–51. [Google Scholar] [CrossRef] [PubMed]
  6. Singh, S.P. Magnetoencephalography: Basic principles. Ann. Indian Acad. Neurol. 2014, 17, S107–S112. [Google Scholar] [CrossRef] [PubMed]
  7. Cargnelutti, E.; Ius, T.; Skrap, M.; Tomasino, B. What do we know about pre-and postoperative plasticity in patients with glioma? A review of neuroimaging and intraoperative mapping studies. Neuroimage Clin. 2020, 28, 102435. [Google Scholar] [CrossRef]
  8. Duffau, H. Brain plasticity and tumors. Adv. Tech. Stand. Neurosurg. 2008, 3, 3–33. [Google Scholar] [CrossRef]
  9. Weng, H.H.; Noll, K.R.; Johnson, J.M.; Prabhu, S.S.; Tsai, Y.H.; Chang, S.W.; Huang, Y.-C.; Lee, J.-D.; Yang, J.-T.; Yang, C.-T. Accuracy of presurgical functional MR imaging for language mapping of brain tumors: A systematic review and meta-analysis. Radiology 2018, 286, 512–523. [Google Scholar] [CrossRef] [Green Version]
  10. Bizzi, A.; Blasi, V.; Falini, A.; Ferroli, P.; Cadioli, M.; Danesi, U.; Aquino, D.; Marras, C.; Caldiroli, D.; Broggi, G. Presurgical functional MR imaging of language and motor functions: Validation with intraoperative electrocortical mapping. Radiology 2008, 248, 579–589. [Google Scholar] [CrossRef] [Green Version]
  11. Ellis, D.G.; White, M.L.; Hayasaka, S.; Warren, D.E.; Wilson, T.W.; Aizenberg, M.R. Accuracy analysis of fMRI and MEG activations determined by intraoperative mapping. Neurosurg. Focus 2020, 48, E13. [Google Scholar] [CrossRef] [Green Version]
  12. Kirsch, H.E.; Zhu, Z.; Honma, S.; Findlay, A.; Berger, M.S.; Nagarajan, S.S. Predicting the location of mouth motor cortex in patients with brain tumors by using somatosensory evoked field measurements. J. Neurosurg. 2007, 107, 481–487. [Google Scholar] [CrossRef] [PubMed]
  13. Nagarajan, S.; Kirsch, H.; Lin, P.; Findlay, A.; Honma, S.; Berger, M.S. Preoperative localization of hand motor cortex by adaptive spatial filtering of magnetoencephalography data. J. Neurosurg. 2008, 109, 228–237. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  14. Schiffbauer, H.; Berger, M.S.; Ferrari, P.; Freudenstein, D.; Rowley, H.A.; Roberts, T.P. Preoperative magnetic source imaging for brain tumor surgery: A quantitative comparison with intraoperative sensory and motor mapping. Neurosurg. Focus 2003, 15, E7. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  15. Tarapore, P.E.; Tate, M.C.; Findlay, A.M.; Honma, S.M.; Mizuiri, D.; Berger, M.S.; Nagarajan, S.S. Preoperative multimodal motor mapping: A comparison of magnetoencephalography imaging, navigated transcranial magnetic stimulation, and direct cortical stimulation. J. Neurosurg. 2012, 117, 354–362. [Google Scholar] [CrossRef] [Green Version]
  16. Hari, R.; Reinikainen, K.; Kaukoranta, E.; Hämäläinen, M.; Ilmoniemi, R.; Penttinen, A.; Salminen, J.; Teszner, D. Somatosensory evoked cerebral magnetic fields from SI and SII in man. Electroencephalogr. Clin. Neurophysiol. 1984, 57, 254–263. [Google Scholar] [CrossRef]
  17. Baillet, S. Magnetoencephalography for brain electrophysiology and imaging. Nat. Neurosci. 2017, 20, 327–339. [Google Scholar] [CrossRef]
  18. Hämäläinen, M.; Hari, R. Magnetoencephalographic characterization of dynamic brain activation: Basic principles and methods of data collection and source analysis. In Brain Mapping: The Methods; Mazziotta, J., Ed.; Elsevier: London, UK, 2002; pp. 227–253. [Google Scholar]
  19. Mäkelä, J.P.; Forss, N.; Jääskeläinen, J.; Kirveskari, E.; Korvenoja, A.; Paetau, R. Magnetoencephalography in neurosurgery. Neurosurgery 2006, 59, 493–511. [Google Scholar] [CrossRef]
  20. Gross, J. Magnetoencephalography in cognitive neuroscience: A primer. Neuron 2019, 104, 189–204. [Google Scholar] [CrossRef]
  21. Braeutigam, S. Magnetoencephalography: Fundamentals and established and emerging clinical applications in radiology. ISRN Radiol. 2013, 529463. [Google Scholar] [CrossRef] [Green Version]
  22. Constable, R.T.; Kennan, R.P.; Puce, A.; McCarthy, G.; Gore, J.C. Functional NMR imaging using fast spin echo at 1.5 T. Magn. Reson. Med. 1994, 31, 686–690. [Google Scholar] [CrossRef]
  23. Inoue, T.; Shimizu, H.; Nakasato, N.; Kumabe, T.; Yoshimoto, T. Accuracy and limitation of functional magnetic resonance imaging for identification of the central sulcus: Comparison with magnetoencephalography in patients with brain tumors. Neuroimage 1999, 10, 738–748. [Google Scholar] [CrossRef]
  24. Hou, B.L.; Bradbury, M.; Peck, K.K.; Petrovich, N.M.; Gutin, P.H.; Holodny, A.I. Effect of brain tumor neovasculature defined by rCBV on BOLD fMRI activation volume in the primary motor cortex. Neuroimage 2006, 32, 489–497. [Google Scholar] [CrossRef]
  25. Pillai, J.J.; Zacà, D. Clinical utility of cerebrovascular reactivity mapping in patients with low grade gliomas. World J. Clin. Oncol. 2011, 2, 397–403. [Google Scholar] [CrossRef]
  26. Zacà, D.; Jovicich, J.; Nadar, S.R.; Voyvodic, J.T.; Pillai, J.J. Cerebrovascular reactivity mapping in patients with low grade gliomas undergoing presurgical sensorimotor mapping with BOLD fMRI. J. Magn. Reson. Imaging 2014, 40, 383–390. [Google Scholar] [CrossRef] [Green Version]
  27. Gao, J.H.; Miller, I.; Lai, S.; Xiong, J.; Fox, P.T. Quantitative assessment of blood inflow effects in functional MRI signals. Magn. Reson. Med. 1996, 36, 314–319. [Google Scholar] [CrossRef]
  28. Morioka, T.; Yamamoto, T.; Mizushima, A.; Tombimatsu, S.; Shigeto, H.; Hasuo, K.; Nishio, S.; Fujii, K.; Fukui, M. Comparison of magnetoencephalography, functional MRI, and motor evoked potentials in the localization of the sensory-motor cortex. Neurol. Res. 1995, 17, 361–367. [Google Scholar] [CrossRef]
  29. Fred, A.L.; Kumar, S.N.; Kumar Haridhas, A.; Ghosh, S.; Purushothaman Bhuvana, H.; Sim, W.K.J.; Vimalan, V.; Givo, F.A.S.; Jousmäki, V.; Padmanabhan, P.; et al. A Brief introduction to magnetoencephalography (MEG) and its clinical applications. Brain Sci. 2022, 12, 788. [Google Scholar] [CrossRef]
  30. Puce, A.; Hämäläinen, M.S. A review of issues related to data acquisition and analysis in EEG/MEG studies. Brain Sci. 2017, 7, 58. [Google Scholar] [CrossRef]
  31. Ahlfors, S.P.; Han, J.; Lin, F.H.; Witzel, T.; Belliveau, J.W.; Hamalainen, M.S.; Halgren, E. Cancellation of EEG and MEG signals generated by extended and distributed sources. Hum. Brain Mapp. 2010, 31, 140–149. [Google Scholar] [CrossRef] [Green Version]
  32. Kober, H.; Möller, M.; Nimsky, C.; Vieth, J.; Fahlbusch, R.; Ganslandt, O. New approach to localize speech-relevant brain areas and hemispheric dominance using spatially filtered magnetoencephalography. Hum. Brain Mapp. 2001, 14, 236–250. [Google Scholar] [CrossRef]
  33. Sekihara, K.; Nagarajan, S.S.; Poeppel, D.; Marantz, A.; Miyashita, Y. Reconstructing spatiotemporal activities of neural sources using an MEG vector beamformer technique. IEEE Trans. Biomed. Eng. 2001, 48, 760–771. [Google Scholar] [CrossRef] [PubMed]
  34. Van Veen, B.D.; Van, D.W.; Yuchtman, M.; Suzuki, A. Localization of brain electrical activity via linearly constrained minimum variance spatial filtering. IEEE Trans. Biomed. Eng. 1997, 44, 867–880. [Google Scholar] [CrossRef] [PubMed]
  35. Baillet, S.; Mosher, J.C.; Leahy, R.M. Electromagnetic brain mapping. IEEE Signal Process. Mag. 2001, 18, 14–30. [Google Scholar] [CrossRef]
  36. Dale, A.M.; Liu, A.K.; Fischl, B.R.; Buckner, R.L.; Belliveau, J.W.; Lewine, J.D.; Halgren, E. Dynamic statistical parametric mapping: Combining fMRI and MEG for high-resolution imaging of cortical activity. Neuron 2000, 26, 55–67. [Google Scholar] [CrossRef] [Green Version]
  37. Pascual-Marqui, R.D. Standardized low-resolution brain electromagnetic tomography (sLORETA): Technical details. Methods Find Exp. Clin. Pharm. 2002, 24 (Suppl. D), 5–12. [Google Scholar]
  38. Kober, H.; Nimsky, C.; Vieth, J.; Fahlbusch, R.; Ganslandt, O. Co-registration of function and anatomy in frameless stereotaxy by contour fitting. Stereotact. Funct. Neurosurg. 2002, 79, 272–283. [Google Scholar] [CrossRef]
  39. Ganzetti, M.; Taberna, G.A.; Mantini, D. fMRI data processing in MRTOOL: To what extent does anatomical registration affect the reliability of functional results? Brain Imaging Behav. 2019, 13, 1538–1553. [Google Scholar] [CrossRef]
  40. Saad, Z.S.; Glen, D.R.; Chen, G.; Beauchamp, M.S.; Desai, R.; Cox, R.W. A new method for improving functional-to-structural MRI alignment using local Pearson correlation. Neuroimage 2009, 44, 839–848. [Google Scholar] [CrossRef] [Green Version]
  41. Hillebrand, A.; Barnes, G.R. Practical constraints on estimation of source extent with MEG beamformers. Neuroimage 2011, 54, 2732–2740. [Google Scholar] [CrossRef] [Green Version]
  42. Whalen, C.; Maclin, E.L.; Fabiani, M.; Gratton, G. Validation of a method for coregistering scalp recording locations with 3D structural MR images. Hum. Brain Mapp. 2008, 29, 1288–1301. [Google Scholar] [CrossRef]
  43. Sonntag, H.; Haueisen, J.; Maess, B. Quality assessment of MEG-to-MRI coregistrations. Phys. Med. Biol. 2018, 63, 075003. [Google Scholar] [CrossRef]
  44. Nimsky, C.; Ganslandt, O.; Kober, H.; Moller, M.; Ulmer, S.; Tomandl, B.; Fahlbusch, R. Integration of functional magnetic resonance imaging supported by magnetoencephalography in functional neuronavigation. Neurosurgery 1999, 44, 1249–1255. [Google Scholar] [CrossRef]
  45. Kober, H.; Nimsky, C.; Möller, M.; Hastreiter, P.; Fahlbusch, R.; Ganslandt, O. Correlation of sensorimotor activation with functional magnetic resonance imaging and magnetoencephalography in presurgical functional imaging: A spatial analysis. Neuroimage 2001, 14, 1214–1228. [Google Scholar] [CrossRef] [Green Version]
  46. Kamada, K.; Houkin, K.; Takeuchi, F.; Ishii, N.; Ikeda, J.; Sawamura, Y.; Kuriki, S.; Kawaguchi, H.; Iwasaki, Y. Visualization of the eloquent motor system by integration of MEG, functional, and anisotropic diffusion-weighted MRI in functional neuronavigation. Surg. Neurol. 2003, 59, 352–360. [Google Scholar] [CrossRef]
  47. Grummich, P.; Nimsky, C.; Pauli, E.; Buchfelder, M.; Ganslandt, O. Combining fMRI and MEG increases the reliability of presurgical language localization: A clinical study on the difference between and congruence of both modalities. Neuroimage 2006, 32, 1793–1803. [Google Scholar] [CrossRef]
  48. Kamada, K.; Takeuchi, F.; Kuriki, S.; Todo, T.; Morita, A.; Sawamura, Y. Dissociated expressive and receptive language functions on magnetoencephalography, functional magnetic resonance imaging, and amobarbital studies: Case report and review of the literature. J. Neurosurg. 2006, 104, 598–607. [Google Scholar] [CrossRef]
  49. Korvenoja, A.; Kirveskari, E.; Aronen, H.J.; Avikainen, S.; Brander, A.; Huttunen, J.; Ilmoniemi, R.J.; Jääskeläinen, J.E.; Kovala, T.; Mäkelä, J.P.; et al. Sensorimotor cortex localization: Comparison of magnetoencephalography, functional MR imaging, and intraoperative cortical mapping. Radiology 2006, 241, 213–222. [Google Scholar] [CrossRef]
  50. Kamada, K.; Todo, T.; Masutani, Y.; Aoki, S.; Ino, K.; Morita, A.; Saito, N. Visualization of the frontotemporal language fibers by tractography combined with functional magnetic resonance imaging and magnetoencephalography. J. Neurosurg. 2007, 106, 90–98. [Google Scholar] [CrossRef]
  51. Choudhri, A.F.; Narayana, S.; Rezaie, R.; Whitehead, M.T.; McAfee, S.S.; Wheless, J.W.; Boop, F.A.; Papanicolaou, A.C. Same day tri-modality functional brain mapping prior to resection of a lesion involving eloquent cortex: Technical feasibility. Neuroradiol. J. 2013, 26, 548–554. [Google Scholar] [CrossRef] [Green Version]
  52. De Martin, E.; Duran, D.; Ghielmetti, F.; Visani, E.; Aquino, D.; Marchetti, M.; Sebastiano, D.R.; Cusumano, D.; Bruzzone, M.G.; Panzica, F.; et al. Integration of functional magnetic resonance imaging and magnetoencephalography functional maps into a CyberKnife planning system: Feasibility study for motor activity localization and dose planning. World Neurosurg. 2017, 108, 756–762. [Google Scholar] [CrossRef]
  53. Izutsu, N.; Kinoshita, M.; Yanagisawa, T.; Nakanishi, K.; Sakai, M.; Kishima, H. Preservation of motor function after resection of lower-grade glioma at the precentral gyrus and prediction by presurgical functional magnetic resonance imaging and magnetoencephalography. World Neurosurg. 2017, 107, 1045-e5. [Google Scholar] [CrossRef]
  54. Zimmermann, M.; Rössler, K.; Kaltenhäuser, M.; Grummich, P.; Brandner, N.; Buchfelder, M.; Dörfler, A.; Kölble, K.; Stadlbauer, A. Comparative fMRI and MEG localization of cortical sensorimotor function: Bimodal mapping supports motor area reorganization in glioma patients. PLoS ONE 2019, 14, e0213371. [Google Scholar] [CrossRef] [Green Version]
  55. Zimmermann, M.; Rössler, K.; Kaltenhäuser, M.; Grummich, P.; Yang, B.; Buchfelder, M.; Dörfler, A.; Kölble, K.; Stadlbauer, A. Refined functional magnetic resonance imaging and magnetoencephalography mapping reveals reorganization in language-relevant areas of lesioned brains. World Neurosurg. 2020, 136, e41–e59. [Google Scholar] [CrossRef]
  56. Quiñones, I.; Amoruso, L.; Pomposo Gastelu, I.C.; Gil-Robles, S.; Carreiras, M. What can glioma patients teach us about language (re) organization in the bilingual brain: Evidence from fMRI and MEG. Cancers 2021, 13, 2593. [Google Scholar] [CrossRef]
  57. Huang, C.W.; Huang, M.X.; Ji, Z.; Swan, A.R.; Angeles, A.M.; Song, T.; Huang, J.W.; Lee, R.R. High-resolution MEG source imaging approach to accurately localize Broca’s area in patients with brain tumor or epilepsy. Clin. Neurophysiol. 2016, 127, 2308–2316. [Google Scholar] [CrossRef]
  58. Krieg, S.M.; Shiban, E.; Buchmann, N.; Gempt, J.; Foerschler, A.; Meyer, B.; Ringel, F. Utility of presurgical navigated transcranial magnetic brain stimulation for the resection of tumors in eloquent motor areas. J. Neurosurg. 2012, 116, 994–1001. [Google Scholar] [CrossRef]
  59. Krieg, S.M.; Shiban, E.; Buchmann, N.; Meyer, B.; Ringel, F. Presurgical navigated transcranial magnetic brain stimulation for recurrent gliomas in motor eloquent areas. Clin. Neurophysiol. 2013, 124, 522–527. [Google Scholar] [CrossRef] [PubMed]
  60. Cargnelutti, E.; Maieron, M.; Ius, T.; Skrap, M.; Tomasino, B. Relation between reading performance and white-matter alteration and reorganization in neurosurgical patients. Front. Neurol. 2020, 11, 563259. [Google Scholar] [CrossRef]
  61. Salvati, L.F.; De Marco, R.; Palmieri, G.; Minardi, M.; Massara, A.; Pesaresi, A.; Cagetti, B.; Melcarne, A.; Garbossa, D. The relevant role of navigated tractography in speech eloquent area glioma surgery: Single center experience. Brain Sci. 2021, 11, 1436. [Google Scholar] [CrossRef]
  62. Youden, W.J. Index for rating diagnostic tests. Cancer 1950, 3, 32–35. [Google Scholar] [CrossRef]
  63. Meier, M.P.; Ilmberger, J.; Fesl, G.; Ruge, M.I. Validation of functional motor and language MRI with direct cortical stimulation. Acta Neurochir. 2013, 155, 675–683. [Google Scholar] [CrossRef] [PubMed]
Life 13 00609 g001 550
Figure 1. Example of spatial and temporal information provided by fMRI and MEG during a representative task (i.e., object naming) in pre-operative mapping. (A) fMRI functional overlay on a glass brain, providing the entire naming-related network (data from our lab, acquired by a 3 T tomograph). (B) Same fMRI activations overlaid on sagittal and axial structural images (note the relevant nodes represented by the calcarine cortex, fusiform gyrus, Wernicke’s, and Broca’s areas); these images can provide a hint on the distance between the activation center of mass and potential lesions. (C,D) Detail of one activation cluster (high-field scanners allow the acquisition of images at high spatial resolution) and related averaged fMRI signal, resulting from the overall activation across several task repetitions (i.e., blocks) throughout the whole task presentation. (E) Temporal fingerprints of the object-naming task, provided by MEG high temporal resolution (data from Grummich et al., 2006); temporal information can be relevant for planning the proper delay after stimulus presentation to be used during intra-operative stimulation mapping.
Figure 1. Example of spatial and temporal information provided by fMRI and MEG during a representative task (i.e., object naming) in pre-operative mapping. (A) fMRI functional overlay on a glass brain, providing the entire naming-related network (data from our lab, acquired by a 3 T tomograph). (B) Same fMRI activations overlaid on sagittal and axial structural images (note the relevant nodes represented by the calcarine cortex, fusiform gyrus, Wernicke’s, and Broca’s areas); these images can provide a hint on the distance between the activation center of mass and potential lesions. (C,D) Detail of one activation cluster (high-field scanners allow the acquisition of images at high spatial resolution) and related averaged fMRI signal, resulting from the overall activation across several task repetitions (i.e., blocks) throughout the whole task presentation. (E) Temporal fingerprints of the object-naming task, provided by MEG high temporal resolution (data from Grummich et al., 2006); temporal information can be relevant for planning the proper delay after stimulus presentation to be used during intra-operative stimulation mapping.
Life 13 00609 g001
Life 13 00609 g002 550
Figure 2. Patients and tasks of the selected studies. (A) The number of patients involved in each selected study. (B) The percentage of studies that investigated the motor, sensory, and language functions by fMRI and MEG. (C,D) The correspondence between fMRI and MEG concerning the adopted task/stimulation, respectively for the motor and sensory function and for the language function (same number is attributed to same task or group of tasks across the studies).
Figure 2. Patients and tasks of the selected studies. (A) The number of patients involved in each selected study. (B) The percentage of studies that investigated the motor, sensory, and language functions by fMRI and MEG. (C,D) The correspondence between fMRI and MEG concerning the adopted task/stimulation, respectively for the motor and sensory function and for the language function (same number is attributed to same task or group of tasks across the studies).
Life 13 00609 g002
Table
Table 1. Selected studies.
Table 1. Selected studies.
AuthorsSample SizeLesionAimReferenceAdditional Relevant ExaminationsResultsfMRI–MEG Agreement
Morioka et al. (1995) [28]5 (4 F)Various (glioma, AVM, cyst, metastasis, MG) in peri-rolandic areasPre-operative sensorimotor cortex localizationIntra-operative (median-nerve) somatosensory stimulation (phase-reversal algorithm)MR angiography, motor-evoked potentialsfMRI failed correct localization in 4/5 patients due to edema or large cortical vein activation; MEG localized the sensory cortex correctly in all patients (agreement with intra-operative mapping in 4/5 patients)N/A
Inoue et al. (1999) [23]11 (7 F)Various (MG, lymphoma, cavernoma) in frontal/parietal areasPre-operative CS localizationIntra-operative (median-nerve) somatosensory stimulation (phase reversal and maximum amplitude algorithms) in 10 ptsMR angiographyBy fMRI, the CS could not be defined in 7/11 affected hemispheres (wider activation areas including large veins). MEG defined it correctly in all, with agreement with intra-operative mappingCoincidence in 9/11 patients for the affected hemisphere and in all for the healthy patients (and in all controls). In case of discrepancy (n = 2), intra-operative mapping confirmed MEG- and not fMRI-defined sulcus to be correct (in one case, fMRI defined the sulcus correctly following tumor resection)
Nimsky et al. (1999) [44]7 (6 F)Various (glioma, cyst, metastasis, MG) adjacent to the motor cortexPre-operative central region localizationIntra-operative (median-nerve) somatosensory stimulation (phase-reversal algorithm)NoneCorrect CS identification by both in all patients confirmed by intra-operative mapping (but MEG could not be performed in 2 patients because of severe artifacts or hemiparesis)Correspondent results in all patients for both motor and sensory activations
Kober et al. (2001) [45]34 (24 F)Various (glioma, angioma, cavernoma, cyst, epilepsy, MG, metastasis) adjacent to the motor cortexPre-operative central region localizationIntra-operative (median-nerve) somatosensory stimulation (phase-reversal algorithm)NoneCS localization successful in all cases using either fMRI or MEG (94% with sensory fMRI, 97% with motor fMRI, 97% with sensory MEG, and 79% with motor MEG). The latter could not be used in patients with hemiparesis, in which a small fMRI activity was detected. No clear superiority of one modality over the otherOverall, correspondent results, although localization significantly differed (Euclidean distance) for both sensory (15 ± 5 mm) and motor (10 ± 5 mm) activations
Kamada et al. (2003) [46]4 (1 F)Various (glioma, cavernoma) involving the motor areaPre-operative central region localizationIntra-operative (median-nerve) somatosensory stimulation (phase-reversal algorithm)Anisotropic diffusion-weighted MRIfMRI-MEG integration correctly identified the CSN/A
Grummich et al. (2006) [47]90Various (glioma, AVM, metastasis, MG, other) in frontal or temporo-parietal areas of the dominant hemispherePre-operative language mappingImaging-guided resection without post-operative sequelae; comparison with Wada test for 10 ptsWada test for 10 ptsBroca and Wernicke’s areas correctly localized in all patients; however, fMRI data were too weak in the case of high-grade/large lesions and MEG data could not be used for tasks not performed regularly by the patientsOverall congruence in 77% of cases with greater agreement for Wernicke’s than Broca’s area
Kamada et al. (2006) [48]1 (M) + 20 as controlsGlioma in the right insula (in frontal/temporal areas for controls)Pre-operative language mappingImaging-guided resection with mild post-operative language deficits + Wada test + control ptsWada testBoth fMRI and MEG showed left-hemisphere dominance for expressive tasks and right-hemisphere dominance for receptive tasks (in agreement with findings from Wada test)Concordant findings (not directly compared) for the word categorization task
Korvenoja et al. (2006) [49]15 (9 F)Various (glioma, cavernous hemangioma, MG) close to the primary sensorimotor cortexPre-operative CS localizationIntra-operative (median-nerve) somatosensory stimulation (phase-reversal algorithm) and/or cortical stimulation (phase reversal) NoneMEG correctly localized the CS in all patients (N20m goodness-of-fit = 88.8–98.2%) and fMRI in 73% (in 27% it localized it in the post-central gyrus)MEG dipoles within 1–6 mm Euclidean distance from nearest fMRI activation voxel and 6–36 mm from the fMRI maximal z-score voxel
Kamada et al. (2007) [50]22 (13 F)Glioma in frontal/temporal areas of the language-dominant hemispherePre-operative language mappingIntra-operative cortical and subcortical mapping during execution of language tasks for 2 ptsWada test, diffusion-tensor imagingBoth identified language areas and hemispheric language dominance correctlyN/A
Choudhri et al. (2013) [51]1 (F)Glioma in the left inferior peri-rolandic cortexPre-operative facial motor area and language mappingImaging-guided resection; TMSTMS, diffusion-tensor imaging and trackingBoth identified bilateral (but predominantly left-sided) language activation; agreement between MEG and TMS in localizing early somatosensory response; overall agreement between fMRI and TMS (slight discrepancy attributed to the use of different tasks)N/A
De Martin et al. (2017) [52]4 (2 F)Metastasis in the right-hand motor cortexPre-operative motor activity localization and dose planningN/ANoneBoth localized the functional areas, enabling reduced irradiation during CyberKnife treatment; in one pt, MEG could map functional area within the tumorN/A
Izutsu et al. (2017) [53]1 (M)Glioma in the right precentral gyrusPre-operative primary motor area localizationImaging-guided resection without post-operative sequelaeNoneBoth showed motor activation shift to the post-central gyrus (and contralesional pre-central gyrus)N/A
Zimmermann et al. (2019) [54]13 (4 F)Glioma or AVM/hemangioma close to the sensorimotor cortexPre-operative sensorimotor cortex localizationImaging-guided resection with general deficit improvement + 3 (1 F) healthy controlsNonefMRI–MEG agreement in the detection of functional reorganization in five pts. In one pt, MEG data were affected by severe artifacts 60% of MEG localizations mapped within fMRI activations (the remaining localized at a mean Euclidean distance of 3.8 mm for sensory fields with sLORETA and 7.4 mm for motor fields with MNE)
Ellis et al. (2020) [11]19Various (glioma, cavernoma, metastasis) close to eloquent areasPre-operative motor and language mapping and quantitative assessment of its accuracyIntra-operative somatosensory stimulation (phase-reversal algorithm) and direct cortical stimulation. Computation of the Youden’s J statistics to assess imaging mapping sensitivity and specificity (at a given distance between intra-operative and imaging localizations)NoneAccuracy highly dependent on distance to direct cortical stimulation site: highest accuracy (J statistic) for both motor and language fMRI at a 5-mm distance and for MEG at 40-mm distance for motor mapping and 15 mm distance for language mapping. fMRI–MEG combination increased accuracy. Poor overlap between locations identified by each modality at the most commonly used 10-mm distance
Zimmermann et al. (2020) [55]18 (8 F)Various (glioma, cavernoma, MG) in the left, language-dominant hemispherePre-operative language mappingImaging-guided resection with general deficit improvement + 3 (1 F) healthy controlsDiffusion-tensor imagingfMRI could not identify the activity in 5/13 language areas in 3/18 patients (with vascular/hemorrhagic alterations), whereas MEG provided strong activationsExcluding areas in which fMRI could not identify activity, congruent activations (mean Euclidean distance = 10.58 mm for activation maxima and 9.06 mm for centroids)
Quiñones et al. (2021) [56]5 (all M)Glioma (all low-grade) in brain areas involved in bilingual languageLanguage mapping changes in bilinguals between pre- and post-surgeryN/ANoneBoth revealed post-operative reorganization taking place for the two languages differentlyComplementary and even convergent findings between the alpha/theta longitudinal indexes and fMRI longitudinal lateralization indexes
Note. AVM = arterial-vascular malformation; CS = central sulcus; MG = meningioma; MNE = minimum-norm estimate; N/A = non-available information; sLORETA = standardized low-resolution brain electromagnetic tomography; TMS = transcranial magnetic stimulation.
Table
Table 2. Methodological details of fMRI and MEG acquisition and processing information in the selected studies.
Table 2. Methodological details of fMRI and MEG acquisition and processing information in the selected studies.
fMRIMEG
AuthorsTomographImaging SequenceSpatial ResolutionTaskDesignActivation Map DefinitionEquipmentTaskSource LocalizationCo-Registration to MRI
Morioka et al. (1995) [28]1.5 TEPIFOV = 240 × 180 mm2, matrix = 256 × 128, slice thickness = 7 mmSelf-paced hand clenchingN/AIdentification of hyperintense activated region37-channel neuro-magnetometerFinger (no. 1, 2, and 5) stimulation (ISI = 400–500 ms, sampling rate = 520 Hz, filtering: 1–55 Hz)ECD (N20 current peak)ECD superimposition on MR images
Inoue et al. (1999) [23]1.5 TGradient-echo EPIFOV = 240 × 240 mm2, matrix = 64 × 64, slice thickness = 10 mmHand clenching once/sec (30 sec/block)Three activation and three resting blocks (10 EPI volumes each)Cross-correlation (coefficient > 0.06 or >0.08) MR-linked 122-channel systemMedian nerve stimulation (200 stimuli, sampling rate = 1280 Hz, filtering = 0.03–400 Hz)ECD (N20m current peak)ECD superimposition on MR images
Nimsky et al. (1999) [44]1.5 TEPIFOV = 200 × 200 mm2, matrix = 64 × 64 (interpolated 128 × 128), slice thickness = 3 mmMotor activation: hand clenching (once/sec); sensory activation: finger tactile stimulation (ISI = 800 ms)Three activation and three resting blocks (10 EPI volumes each)Cross-correlation (coefficient > 0.65 for motor and >0.55 for sensory activation)2 × 37-channel bio-magnetometerMotor activation: brisk index finger flection (100 stimuli, ISI = 3000–5000 ms, sampling rate = 520.8 Hz, filtering = 1–100 Hz); sensory activation: see fMRI (200 stimuli, sampling rate = 1041.7 Hz, filtering = 1–200 Hz)Single ECD (first peak for motor activation and M30 for sensory activation; correlation > 0.95) ECD superimposition on MR images (by contour-fit algorithm)
Kober et al. (2001) [45]1.5 TEPIFOV = 200 × 200 mm2, matrix = 64 × 64 (interpolated 128 × 128), slice thickness = 3 mmMotor activation: hand clenching (once/sec); sensory activation: finger tactile stimulation (ISI = 800 ms)Three activation and three resting blocks (10 EPI volumes each)Cross-correlation (coefficient > 0.65 for motor and >0.55 for sensory activation)2 × 37-channel bio-magnetometerMotor activation: brisk finger flection (100 stimuli, ISI = 3000–5000 ms, sampling rate = 520.8 Hz, filtering = 0.1–100 Hz); sensory activation: see fMRI (200 stimuli, ISI = 800 ms, sampling rate = 1041.7 Hz, filtering = 1–200 Hz)Single ECD (P35m current peak; correlation > 0.95 or + current localization by spatial filtering if <0.95)ECD superimposition on MR images (by contour-fit algorithm)
Kamada et al. (2003) [46]1.5 TEPIFOV = 300 × 300 mm2, matrix = 128 × 128, slice thickness = 5 mm (2.5 mm gap)Self-paced finger tapping (approximately once/sec)Three activation and four resting blocks (five EPI volumes each)Cross-correlation (Z > 3.5)204-channel bio-magnetometerMedian nerve stimulation (200 stimuli, ISI = 211 ms, filtering (averaged signals) = 1–70 Hz)Single ECD (N20m current peak; correlation > 0.95 and confidence volume < 200 mm3)ECD superimposition on MR images
Grummich et al. (2006) [47]1.5 TEPIVoxel size = 3 mm3Visually presented tasks selected according to the patient’s clinical features: word-reading task (300 stimuli), sentence-reading task (61 stimuli), picture-naming task (75 stimuli), verb-generation task, arithmetic task. ISI = 900–2000 msSix activation blocks interleaved with resting blocksCross-correlation (coefficient > 0.03, p < 0.000045, k ≥ 6)2 × 37-channel bio-magnetometerSee fMRI (but with ISI = 1200–2300 ms, sampling rate = 520.8 Hz, filtering = 0.1–200 Hz + 0.03–1 Hz)Single ECD (correlation > 0.94 or current-density reconstruction for lower correlations)ECD superimposition on MR images (by contour-fit algorithm)
Kamada et al. (2006) [48]1.5 TEPIFOV = 260 × 260 mm2, matrix = 64 × 128, slice thickness = 4 mm (2 mm gap)Two tasks: expressive task (verb generation from acoustically presented nouns (ISI = 1600–2400 ms)) and receptive task (word categorization of visually displayed words (ISI = 1800–2200 ms)). Three activation and four resting blocks (five EPI volumes each)Z-score estimation (clusters with Z > 2.2 and k > 10)204-channel bio-magnetometerWord categorization of visually displayed words (150 stimuli, ISI = 2800–3200 ms, filtering (averaged signals) = 0.01–30 Hz)Single ECD (+multiple-current estimates to confirm results; correlation > 0.90)ECD superimposition on MR images
Korvenoja et al. (2006) [49]1.5 TGradient-echo EPIFOV = 256 × 256 mm2, matrix = 128 × 128, slice thickness = 3 mmSelf-paced hand clenchingAlternated activation and resting blocks (91–128 EPI volumes)Z-score estimation (clusters with Z > 1.6 and neighborhood weighting = 0.427)122- or 306-channel magnetometerMedian-nerve stimulation (about 200 stimuli, sampling rate = 987 Hz, filtering = 0.03–320 Hz)ECD (N20m current peak)ECD superimposition on MR images
Kamada et al. (2007) [50]1.5 TEPIFOV = 260 × 260 mm2, matrix = 64 × 128, slice thickness = 4 mm (2 mm gap)Verb generation from acoustically presented nouns (ISI = 1600–2400 ms)Three activation and four resting blocks (five EPI volumes each)Z-score estimation (clusters with Z > 2.2)204-channel bio-magnetometerWord categorization of visually displayed words (150 stimuli, ISI = 2800–3200 ms, filtering (averaged signals) = 0.01–30 Hz)Single ECD (N400m current peak; correlation = > 0.85)ECD superimposition on MR images
Choudhri et al. (2013) [51]3 TN/AN/AMotor tasks: tongue movement and lip puckering; language tasks: silent word generation, object naming, and sentence completionAlternated activation and resting blocks (five volumes each)N/A248-magnetomer systemMotor activation: index finger stimulation; language activation: auditory word recognition (sampling rate = 508 Hz, frequency range = 0.1–20 Hz)N/ADipole superimposition on MR images
De Martin et al. (2017) [52]3 TGradient-echo EPIVoxel size = 2.5 mm3 Two tasks: hand lifting and hand lowering (21 sec/block)20 activation blocks alternated with resting blocksGLM (voxels with p < 0.01 and Bonferroni correction)306-channel neuro-magnetometerBrisk hand extension following a visual stimulation (at least 100 stimuli, sampling rate = 1000 Hz, filtering = 0.01–100 Hz)dSPM (movement-related field peak; threshold = 80% of maximum value)ECD superimposition on MR images (by iterative closest-point algorithm)
Izutsu et al. (2017) [53]3 TSingle-shot EPIVoxel size = 3 mm3 (3.75 mm gap)Hand grasping (15 sec/block)Three alternated activation and resting blocksGLM (clusters with t > 4.0)160-channel systemHand grasping following either visual or auditory stimulation (ISI = 5500 ms, sampling rate = 1000 Hz, filtering > 200 Hz)Event-related desynchronization (synthetic aperture magnetometry—a beamforming approach—to compute power changes in theta, alpha, beta, and low- and high-gamma bands)N/A
Zimmermann et al. (2019) [54]1.5 TEPIFOV = 192 × 192 mm2, matrix = 64 × 64, slice thickness = 3 mmTwo tasks: flection-extension of all digits or toesThree activation and three resting blocks (30 EPI volumes each)Linear correlation (coefficient > 0.03, p < 0.000045, and clusters with k ≥ 4)248-magnetomer systemFlection-extension of digits two to five or toes (two separate tasks) following a tactile stimulation (300 stimuli, ISI = 3600 ms, sampling rate = 678 Hz, filtering = 0.01–200 Hz)MNE, dSPM, sLORETAECD superimposition on MR images
Ellis et al. (2020) [11]3 TSingle-shot fast-field EPIVoxel size = 3 mm3Motor tasks: finger tapping, foot movement, and lip pursing (following an acoustic signal (ISI = 1000 ms)); language task: word reading (15 sec/block)10 activation and 10 resting blocks (five EPI volumes each)Thresholded by eye306-channel system (102 magnetometers and 204 planar radiometers)Motor tasks (11 patients): finger tapping, foot movement, and face movement, following an auditory stimulation (about 120 stimuli, ISI = 3500–4000 ms); language task (11 patients): word reading (ISI = 3000 ms) (for both: sampling rate = 1000 Hz, filtering = 0.01–330 Hz)Single ECD (Nelder-Mead nonlinear search algorithm; ≥90% variance and 95% confidence volumes < 3 cm)ECD superimposition on MR images (by iterative closest-point algorithm and using a spherically symmetric single-conductor head model)
Zimmermann et al. (2020) [55]1.5 TEPIFOV = 192 × 192 mm2, matrix = 64 × 64, slice thickness = 3 mmTwo tasks: verb conjugation and sentence building (up to 150 stimuli each)Three activation and three resting blocks (30 EPI volumes each)Linear correlation (coefficient > 0.03, p < 0.000045, and clusters with k ≥ 4—or until signal appearance)248-magnetomer systemTwo tasks: verb-conjugation task (300 stimuli, ISI = 3000 ms) and sentence-building task (300 stimuli, ISI = 2000 ms) from visually presented words (for both: sampling rate = 678 Hz, filtering = 0.1–200 Hz (0.03–95 Hz for averaged signals) + 50- and 60-Hz Notch) dSPMECD superimposition on MR images
Quiñones et al. (2021) [56]3 TEPIvoxel size = 2 mm3Naming task (object- and action-naming) in the two known languages (44 stimuli for each category, ISI = 2000–8000 ms)event-related design (368 EPI)Robust weighted least-squares regression with FDR correction (p < 0.05, height Threshold: p < 0.001, and k > 50)360-channel systemSee fMRI (but ISI = 2000–3000 ms, sampling rate = 1000 Hz, filtering = 0.01–330 Hz)Time-frequency representations (cluster-based permutation approach to compute power changes in theta, alpha, and beta bands)ECD superimposition on MR images
Note. dSPM = dynamical statistical parametric mapping; ECD = equivalent current dipole; EPI = echo-planar imaging; FDR = false discovery rate; GLM = general linear model; k = number of voxels; ISI = inter-stimulus interval; MNE = minimum-norm estimate; N/A = non-available information; sLORETA = standardized low-resolution brain electromagnetic tomography.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Cargnelutti, E.; Tomasino, B. Pre-Operative Functional Mapping in Patients with Brain Tumors by fMRI and MEG: Advantages and Disadvantages in the Use of One Technique over the Other. Life 2023, 13, 609. https://doi.org/10.3390/life13030609

AMA Style

Cargnelutti E, Tomasino B. Pre-Operative Functional Mapping in Patients with Brain Tumors by fMRI and MEG: Advantages and Disadvantages in the Use of One Technique over the Other. Life. 2023; 13(3):609. https://doi.org/10.3390/life13030609

Chicago/Turabian Style

Cargnelutti, Elisa, and Barbara Tomasino. 2023. "Pre-Operative Functional Mapping in Patients with Brain Tumors by fMRI and MEG: Advantages and Disadvantages in the Use of One Technique over the Other" Life 13, no. 3: 609. https://doi.org/10.3390/life13030609

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop