Introduction
Gliomas represent the most common malignant entity of neoplasms of the central nervous system (CNS), accounting for approximately 50 % of all malignant brain tumors [1 ]
[2 ]. According to the 2021 World Health Organization (WHO) classification of tumors of the CNS, gliomas can be categorized into different entities according to combined histological and molecular grading [3 ]. High-grade astrocytoma and glioblastoma are particularly common high-grade tumors (WHO grades 3 and 4) and have extraordinarily poor prognoses (5-year survival rates below 30 %) [1 ]
[2 ]. Therapy in most cases includes neurosurgical tumor resection and extended focal irradiation, as well as adjuvant chemotherapy [4 ]
[5 ]
[6 ].
During the course of disease, cranial magnetic resonance imaging (MRI) is paramount for the diagnosis, prognosis estimation, and treatment response assessment and monitoring. Specifically, initial imaging prior to tumor resection allows not only assessment of the distinct location of tumor growth and involved structures but also image-based tumor grading and phenotyping [7 ]
[8 ]. Furthermore, preoperative MRI provides images crucial for neurosurgical tumor resection planning and guidance, which can include the assessment of the involvement of functionally eloquent brain structures using additional techniques such as functional MRI (fMRI) and tractography of subcortical white matter (WM) pathways [9 ]
[10 ]. Lately, navigated transcranial magnetic stimulation (nTMS) has found its way into the armamentarium of the preoperative workup of patients with glioma, providing image-based functional mapping data with the major goal of sparing functionally eloquent brain tissue from harm during resection [11 ]
[12 ]. Functional data derived from fMRI or nTMS mapping can also be effectively combined with diffusion-weighted MRI to establish function-based tractography of major WM bundles, such as the corticospinal tract (CST) or arcuate fascicle (AF) [12 ]
[13 ].
Against this background, the purpose of this narrative review article is to provide an overview of advanced preoperative MRI and functional mapping. Specifically, we review applications such as diffusion-weighted imaging including fiber tractography, magnetic resonance spectroscopy (MRS), perfusion imaging, contrast-enhanced T1-weighted imaging, fMRI, and nTMS. Relevant studies were identified by PubMed search (http://www.ncbi.nlm.nih.gov/pubmed ; Supplementary Table ).
Advanced Preoperative Imaging
Overview of methods
Conventional structural MRI defines the standard approach in neuro-oncological imaging, including axial fluid-attenuated inversion recovery (FLAIR), axial diffusion-weighted imaging, axial T2-weighted, and three-dimensional (3D) T1-weighted sequences before and after the administration of contrast agents using a 1.5-Tesla MRI system at minimum [14 ]
[15 ]. This approach is commonly supplemented by further advanced sequences, depending on technical feasibility, time constraints, and individual needs with respect to interdisciplinary clinical requirements: diffusion tensor imaging (DTI) or specific high-resolution iso-volumetric 3D imaging may be added for dedicated neurosurgical needs or radiosurgical planning [14 ]
[16 ]
[17 ].
Diffusion-weighted imaging
Diffusion-weighted imaging in neuro-oncology covers a broad spectrum of different sequences and approaches. Most commonly, the DTI technique is used during the clinical routine, which investigates the shape of diffusion considering direction (eigenvectors) as well as diffusivity (eigenvalues) and allows extraction of scalar measures such as the fractional anisotropy (FA), either for specific regions of interest (ROIs) or the whole brain [18 ]
[19 ]. Commonly, maximal and/or mean FA values are significantly higher in high-grade glioma compared to low-grade glioma, with a pattern of infiltration and disruption of fibers being more characteristic for high-grade tumors [20 ]
[21 ]
[22 ]. Specifically, cutoff values of 0.129 (mean FA) and 0.219 (maximal FA) have been proposed to distinguish between low- and high-grade glioma, with a resulting specificity of 69.2 %/76.9 % and sensitivity of 93.3 %/100 % [21 ]. Additionally, studies used the FA to discriminate between tumors according to the isocitrate dehydrogenase (IDH) mutation status, which has become a relevant diagnostic marker since mutation correlates to less aggressive biologic behavior and better clinical outcome compared to the wild-type status [23 ]. Maximal FA and the ratio of maximal FA (maximal FA divided by the contralateral normal FA) were significantly different between oligodendroglial tumors with IDH mutations and those without mutations (area under the curve [AUC]: 0.79 and 0.82) [24 ]
[25 ]. Furthermore, DTI-derived parameters, in particular mean diffusivity and FA, can visualize tumor cell densities and infiltration [26 ]
[27 ]. This relevant information is, however, overlaid by free-water contamination, which is particularly relevant for the peritumoral edematous region. Thus, several strategies have been developed to disentangle and bias-correct the “true” diffusion signal, which could increase the diagnostic value of DTI-derived metrics [28 ]
[29 ]
[30 ].
Besides its role for tumor grading, the DTI technique can be used to visualize the spatial course of WM pathways, which can appear unaffected, deviated, infiltrated, or destroyed (entire or partial disintegrity) due to the tumor mass as depicted in color-coded FA maps ([Fig. 1 ]) [31 ]. Yet, most notably, the DTI technique has been used to conduct fiber tractography to delineate specific subcortical WM pathways prior to tumor resection.
Fig. 1 Diffusion tensor imaging (DTI). Axial fluid-attenuated inversion recovery (FLAIR) A , DTI-derived fractional anisotropy (FA) color map B , and T1-weighted images before C and after D administration of a gadolinium-based contrast agent. Conventional structural sequences are indicative of a left-hemispheric high-grade glioma affecting the precentral, postcentral, and superior and middle frontal gyrus, which affects the spatial architecture of subcortical white matter (WM) pathways according to the color-coded FA map. Specifically, tracts are deviated and partially destroyed due to tumor growth when compared to the contralateral unaffected hemisphere.
It needs to be emphasized that although widely used in the clinical routine, the DTI method has relevant drawbacks because a single tensor can only resolve a single fiber direction within an imaging voxel, while the vast majority of WM voxels may be constituted of more than a single fiber [32 ]
[33 ]
[34 ]. Hence, novel methods have been developed lately, which may partially compensate for the drawbacks of DTI and could provide information beyond a simple diffusion scalar by emphasizing the importance of more complex 3D patterns of diffusion within the brain. Diffusion kurtosis imaging (DKI) is an approach to provide a more accurate model of diffusion and to capture non-Gaussian diffusion patterns as representative markers for tissue heterogeneity [35 ]. For glioma grading, it has been shown that DKI-derived mean, radial, and axial kurtosis were significantly higher in high-grade than in low-grade gliomas, probably as a result of a higher degree of tissue complexity in high-grade glioma, while conventional diffusion parameters (e.g., FA and MD) were not significantly different between grades [36 ]. Moreover, neurite orientation dispersion and density imaging (NODDI) is a technique for estimating the microstructural complexity of dendrites and axons [37 ]. In glioma, NODDI for evaluation of the T2-hyperintense region around contrast-enhancing tumor parts might facilitate differentiation between the region infiltrated by the tumor and edematous or normal tissue [38 ]. For the peritumoral region, it has also been proposed that metrics derived from NODDI could be helpful for differentiating between metastatic lesions and glioma [39 ].
A promising approach particularly for the purpose of fiber tracking is high angular resolution diffusion imaging (HARDI), which excels in detecting the orientational distribution of water diffusion and, thus, could also resolve complex fiber configurations [40 ]
[41 ]. Exemplarily, one study using both DTI- and HARDI-based tractography has demonstrated that the HARDI-based approach displayed more compact fiber bundles and more neuroanatomically plausible fibers in the vicinity of the tumor and within the peritumoral region, which were not tracked using DTI [42 ]. Furthermore, HARDI q-ball tractography (using residual bootstrap) enables prediction of long-term language deficits following tumor resection [43 ]. Another novel approach is multi-level fiber tracking (MLFT) as an attempt to add branches to reconstructed WM pathways that do not reach a predefined target region [44 ]. Specifically, based on a conventional diffusion-weighted MRI sequence, MLFT has been shown to provide CST reconstructions with higher radial extent, thus enabling delineation of CST fanning with a wider angular range [44 ]. While such advanced methods have not yet been broadly implemented in the clinical routine, they may have the potential to considerably improve diffusion-weighted MRI including tractography.
Magnetic resonance spectroscopy
The application of MRS allows for noninvasive metabolic quantification by means of a spectrum of peaks that represent metabolite intensities resonating at different frequencies, which is often referred to as “virtual biopsy” [45 ]
[46 ]. Proton MRS is commonly used in the clinical setting and is derived from one or more voxels of interest placed within the tumor volume or surrounding tissue ([Fig. 2 ]) [45 ]
[46 ].
Fig. 2 Proton magnetic resonance spectroscopy (MRS). Placement of the voxel of interest for MRS in sagittal, coronal, and axial view of the fluid-attenuated inversion recovery (FLAIR) sequence A , together with the obtained spectrum of metabolites (Cr2/Cr: creatine, Glx: glutamate and glutamine, mI: myo-inositol, Cho: choline; NAA: N-acetylaspartate; Lac: lactate; Lip: lipids). The spectrum is indicative of a brain tumor with slightly increased Cho (at ~3.22 ppm) and decreased NAA (at ~2.02 ppm) compared to reference values known for healthy brain tissue.
An early study proposed that MRS has potential in the diagnosis of low- vs. high-grade tumors and high-grade tumors vs. metastases when used as part of a multi-sequence MRI protocol [47 ]. Another study investigated the added value of MRS, showing that MRS data improved low- and high-grade tumor prediction when compared to conventional MRI alone (AUC low-grade tumors: 0.93 vs. 0.81; AUC high-grade tumors: 0.93 vs. 0.85 for MRI with MRS vs. conventional MRI alone) [48 ]. In this context, relatively increased total choline (Cho) and decreased total N-acetylaspartate (NAA) are diagnostic characteristics indicative of brain tumors [45 ]
[49 ]. Beyond tumor grading, MRS has also been shown to be able to identify subtypes of glioma with IDH mutations, and the prominent signal at 1.3 ppm that stems from lipids of cytoplasmic droplets associated with necrosis or hypoxia has been shown to correlate with higher tumor aggressiveness or poor survival [50 ]
[51 ].
Perfusion imaging
Several techniques are available for measuring perfusion, including dynamic susceptibility contrast (DSC) imaging and arterial spin labeling (ASL) [52 ]
[53 ]
[54 ]. In the clinical setting, DSC imaging is the most common option. It requires a bolus of contrast agent passing through the capillary bed of the brain, causing measurable susceptibility-induced signal loss on T2*-weighted imaging ([Fig. 3 ]) [55 ]. A fundamentally different technique is ASL, which does not need the application of any contrast media, but instead makes use of the labeling of arterial blood that flows to the brain [55 ]. Common parameters that can be extracted from DSC perfusion are relative cerebral blood flow (CBF), relative cerebral blood volume (CBV), and mean transit time, while ASL measurements may be mostly restricted to CBF [54 ]
[55 ].
Fig. 3 Dynamic susceptibility contrast (DSC) perfusion. Axial fluid-attenuated inversion recovery (FLAIR) A , color-coded map for relative cerebral blood volume (CBV) derived from DSC imaging B , and T1-weighted images before C and after D administration of a gadolinium-based contrast agent. Conventional structural sequences are indicative of a right-hemispheric high-grade glioma of the temporal lobe, with increased relative CBV at the contrast-enhancing tumor borders and decreased relative CBV in the necrotic tumor core according to DSC perfusion.
Notably, there is a strong correlation between the glioma grade and DSC-derived relative CBV, with high-grade tumors typically presenting with markedly higher relative CBV than low-grade tumors or normal-appearing WM [56 ]
[57 ]
[58 ]. In view of earlier work showing that increased relative CBV indeed correlates with neoangiogenesis, these results corroborate the potential of perfusion imaging to visualize this central oncogenic process in high-grade gliomas [59 ]
[60 ]
[61 ]. Furthermore, relative CBV was shown to be increased up to about one year before contrast enhancement is visualized on T1-weighted sequences for low-grade gliomas that undergo a malignant transformation [62 ]. Yet, a very common challenge to relative CBV quantification from DSC perfusion is that the presence of a leaky blood-brain barrier can confound measurements, which needs to be corrected for [63 ]. A multitude of methods are available to address leakage correction, yet no universally accepted approach has been revealed [63 ]
[64 ]. Nevertheless, in the clinical routine, most tools for the analysis of DSC perfusion data nowadays incorporate correction steps to mitigate bias due to leakage.
Regarding ASL-derived CBF, both maximum CBF and maximum relative CBF have shown to be significantly higher in high-grade than low-grade gliomas (AUC maximum CBF: 0.83; AUC maximum relative CBF: 0.86) [65 ]. Furthermore, ASL-derived CBF maps allowed stratification of survival in the case of glioblastoma and could be used to differentiate gliomas with respect to IDH mutation status [66 ]
[67 ]. It has recently been suggested that ASL perfusion may predict malignant progression within one year among patients with glioma WHO grade II [68 ]. In essence, the advantages of ASL are that CBF quantification is not affected by leakage effects, and it does not require administration of a contrast agent. In light of ongoing debates regarding gadolinium depositions from contrast media within the brain, this characteristic could be regarded as being of special interest [69 ]. Yet, ASL imaging typically has a lower signal-to-noise ratio than DSC perfusion, and the relevance of contrast media-free imaging is relativized in most cases since contrast agents are applied anyway for later T1-weighted imaging to evaluate contrast enhancement of brain tumors.
Contrast-enhanced T1-weighted imaging
Imaging with T1-weighted sequences before and after the administration of a contrast agent is an integral part of an imaging protocol in neuro-oncology. The T1 relaxation time is shortened by gadolinium-based contrast agents, which increase tissue contrast by accentuating areas where leakage into interstitial tissue is present due to blood-brain barrier disruption, with resulting parenchymal enhancement being positively correlated to the tumor grade with few exceptions [70 ]
[71 ]. Most commonly, turbo field echo (TFE) imaging before and after contrast administration is used to assess tumor-related contrast enhancement and spread, but recent studies have suggested improved depiction of intracranial contrast-enhancing pathology with advanced sequences [72 ]
[73 ]. Specifically, T1-weighted black-blood sequences may better delineate therapy-naïve high-grade gliomas with higher contrast-to-noise ratios when compared to established TFE sequences, which was also confirmed for intraoperative MRI during tumor removal where assessment of the extent of tumor resection could be accelerated [73 ]
[74 ].
Advanced image analysis
With advancements in scanner technology, a multi-sequence protocol including imaging for diffusion, perfusion, metabolism, and function in addition to conventional structural sequences (i. e., T1- and T2-weighted and FLAIR sequences) can become feasible in most patients within a reasonable scan time, which is partly due to the introduction of different image acquisition acceleration techniques for clinical routine MRI [75 ]
[76 ]
[77 ]
[78 ]. The rich information on tumor biology contained herein reflects many key cellular and oncogenic aspects, including cellularity, proliferation, neoangiogenesis, and invasion, with the opportunity to extract and define quantitative MRI-based biomarkers for neuro-oncological imaging [7 ]. While glioma genotyping based on tissue probes as gathered from biopsy or tumor resection remains the reference standard, genotype predictions by advanced MRI could support clinical decision-making and individual patient management that is tailored to the distinct tumor characteristics [7 ]. Leveraging the rich information from multi-sequence MRI for training multi-parametric models to infer tumor biology is therefore an active field of research, both at initial diagnosis and along the disease course [79 ]
[80 ]
[81 ].
Mapping of Brain Function
Overview of methods
For the preoperative workup of patients, functional mapping is of high importance in addition to structural MRI when the tumor is supposed to affect functionally eloquent brain structures (e. g., the hand knob as the center of primary motor function or the left-hemispheric opercular and triangular parts of the inferior frontal gyrus harboring the Brocaʼs area). Major techniques used for this purpose are fMRI, magnetoencephalography (MEG), and nTMS. While MEG is rather expensive and not widely available in most countries, fMRI is the standard approach in many centers. More recently, nTMS has been made available for preoperative functional mapping [11 ]
[12 ].
Functional magnetic resonance imaging
Methodologically, fMRI indirectly measures neuronal activation by making use of the deoxyhemoglobin-to-oxyhemoglobin ratio as a contrast mechanism, which is referred to as the blood oxygenation level-dependent (BOLD) signal that can be used to map function within the brain when combined with a task (e. g., finger tapping task to detect motor function within the brain) ([Fig. 4 ]) [82 ]
[83 ]
[84 ]
[85 ]. While task-based fMRI is the most common technique for presurgical functional mapping among patients with brain tumors, resting-state fMRI, which measures spontaneous low-frequency fluctuations in the BOLD signal between regions to detect functional networks, has also been applied recently [85 ]
[86 ]
[87 ]. Regarding preoperative motor mapping by task-based fMRI, most studies demonstrated that task-based fMRI is an adequate method to localize motor function, and it could facilitate surgical planning and decrease the time needed for intraoperative mapping using direct electrical stimulation (DES) [88 ]
[89 ]
[90 ]. Specifically, the sensitivity and specificity of task-based fMRI for the delineation of motor function have been reported to range from 71 % to 100 % and 68 % to 100 %, respectively [88 ]
[89 ]
[90 ]. Yet, the specificity and sensitivity for the preoperative localization of language function using task-based fMRI showed higher variability, with sensitivity ranging from 59 % to 100 % and specificity ranging from 0 % to 97 % compared to DES [88 ]
[91 ]
[92 ]. The variability regarding sensitivity and specificity across studies may be related to a variety of factors, including differences in the language tasks that are used, the MRI hardware, and the software including analysis paradigms [85 ]
[93 ]. For instance, an appealing option to tackle issues related to fMRI data alignment, which is a prerequisite for comparing features such as brain activity at corresponding locations across patients, can be based on global functional connectivity patterns, which facilitates matching of functionally corresponding areas in a more accurate fashion than conventionally used anatomical alignment [94 ]. Furthermore, non-rigid image registration algorithms may overcome limitations regarding alignment for longitudinal studies and particularly for registering presurgical to intraoperative datasets including the registration of fMRI to anatomical sequences [95 ]. A longitudinal design may be chosen in particular to track down plastic reorganization of the brain in response to the presence and growth patterns of glioma by means of changes in the fMRI signal and connectivity profiles over time, which could relate to measurable reallocation of motor or language areas [96 ]
[97 ]
[98 ].
Fig. 4 Functional magnetic resonance imaging (fMRI). Task-based fMRI with derived activation maps in axial view to localize motor function A and language function B . A finger-tapping task and toe-movement task were used to localize motor function, which was located lateral to the tumor for upper extremity motor representation (middle image, A ) and medial to the tumor for lower extremity motor representation (right image, A ). Specifically, motor activation maps primarily overlapped with the precentral gyrus bilaterally as well as with parts of the superior frontal gyrus of the right hemisphere (middle and right image, A ). A picture-naming task was used to localize language function, which was located anterior to the tumor (right image, B ). Specifically, left-hemispheric fronto-temporal parts of the language network overlapped with the language activation map (right image, B ).
A main criticism regarding fMRI is that tumor vasculature can lose the ability to autoregulate, which – together with tumor-related compressive effects on venules and larger veins and arteriovenous shunting – can render BOLD signal evaluations imprecise and, thus, impacts the accuracy of findings, particularly for patients with high-grade glioma [99 ]
[100 ]
[101 ]. Due to this neurovascular uncoupling, task-based fMRI can be considered more accurate and useful in low-grade compared to high-grade gliomas [100 ]
[102 ]
[103 ]. Another issue is that fMRI activation maps could show false-positive results by outlining a region larger than the actual functionally eloquent area when correlated to DES, which could negatively influence the extent of tumor resection [104 ].
Navigated transcranial magnetic stimulation
Functional mapping using magnetic stimulator devices is based on the principle of electro-magnetic induction [105 ]
[106 ]
[107 ]. Brief high-current pulses are produced by a magnetic coil, which is placed above the scalp [105 ]
[106 ]
[107 ]. A transient electric field is then induced perpendicular to the magnetic field, which is capable of causing neuronal activation with different extents and effects, depending on factors such as stimulation intensity, pulse shape, and frequency [105 ]
[106 ]
[107 ]. The fundamental difference between nTMS and other techniques is that when a physiological response is evoked by stimulation of a cortical area, that specific cortical area is causally related to the response since a so-called “virtual lesion” is induced by nTMS [12 ]
[106 ]. Furthermore, it is believed that responses to nTMS are not biased due to tumor characteristics (e. g., related to increased perfusion), making the technique potentially more robust and reliable than the presurgical alternatives.
The transformation into an advanced functional mapping device with very close links to imaging is inherently linked to the recent combination of magnetic stimulation with precise neuronavigation based on structural MRI data, defining the technique as nTMS ([Fig. 5 ]) [12 ]
[106 ]. Systems with the highest accuracy to identify and spatially enclose functional brain tissue use electric-field-based neuronavigation, which can be achieved through individual modelling that takes into account parameters such as skull thickness, affecting the coil-cortex distance, and coil tilting [12 ]
[106 ]. Importantly, a simple method to guide magnetic stimulation (e. g., using standard coil location with respect to external landmarks of the skull) would not be acceptable for preoperative mapping in neuro-oncology as there is a high risk of imprecision [12 ]
[106 ]. The starting point for mapping by nTMS is given by co-registration of the respective structural MRI (i. e., high-resolution 3D contrast-enhanced T1-weighted sequences) to the actual head of the patient. Once registration is completed, the stimulation coil can be freely navigated during mapping and tracked on the MRI-based head model within the nTMS system, thus allowing stimulation across hemispheres to pinpoint sites responsible for brain functions such as active movement or speech and language [11 ]
[12 ].
Fig. 5 Navigated transcranial magnetic stimulation (nTMS). Neuronavigational view with a three-dimensional (3D) head model based on a contrast-enhanced T1-weighted sequence for motor mapping A and language mapping B by nTMS in a patient with a left-hemispheric contrast-enhancing tumor affecting the ventral precentral and opercular region of the inferior frontal gyrus. The white spots indicate motor-positive stimulation points A or language-positive stimulation points B , i. e. points that are considered part of the cortical primary motor or language representation. Judgement is based on motor-evoked potentials (MEPs) for nTMS motor mapping, which are derived from continuously recorded electromyography (EMG) of upper and lower extremity muscles contralateral to the tumor-affected hemisphere during stimulation C . Regarding language mapping, transient impairments during performance of a task such as object naming can be elicited by nTMS, which can be used to judge on the spatial location and characteristics of language-positive stimulation points (e. g., typically speech arrests due to targeted stimulation of the Broca's area or semantic paraphasia occurs due to stimulation of parietal or posterior temporal cortex). The use of precise neuronavigation qualifies nTMS as a preoperative tool to map cortical function, which is established through infrared tracking of the coil during stimulation and registration of the patient’s head to the respective image data D . The stimulating coil can then be tracked during pulse application in relation to individual brain anatomy D .
The primary use case for nTMS in neuro-oncology is the mapping of motor function to identify the motor hotspot and boundaries of the primary motor cortex ([Fig. 5 ]). Using electromyography (EMG) of upper and lower extremity muscles during cortical stimulation by the coil, motor-evoked potentials (MEPs) can be elicited and related to a specific site of stimulation. When such MEPs reach a certain amplitude threshold and fall within a muscle-characteristic latency, motor-positive points are defined that are considered essential for primary motor function [108 ]
[109 ]. Compared to DES, presurgical nTMS has repeatedly demonstrated high accuracy [110 ]
[111 ]. Notably, significantly better agreement between nTMS and DES has been achieved for determining the primary motor cortex when compared to fMRI against DES [110 ]
[111 ]. Furthermore, motor mapping by nTMS may make it possible to reveal plastic reallocation of the motor cortex related to tumor growth, demonstrating location changes of the primary motor area by repeated mapping over time [112 ]
[113 ]. Regarding clinical outcome, use of preoperative nTMS motor mapping could improve the extent of tumor resection and survival [114 ]. Yet, data from randomized controlled trials are currently lacking to confirm positive impact on the clinical course besides the distinct value of the technique for tumor resection planning and intraoperative guidance.
Furthermore, language mapping by nTMS is increasingly used in patients with language-eloquent brain tumors ([Fig. 5 ]). The principle is that stimulation by nTMS can cause several instances of transient impairment (e. g., during performance of an object-naming task), which can be recorded and spatially correlated to the site of stimulation [109 ]
[115 ]
[116 ]. Correlations of results from nTMS language mapping to DES are not as satisfactory as for motor mapping, which currently suggests primary application for so-called “negative mapping” (i. e., a language-negative stimulation spot of nTMS is almost always also negative during DES) [117 ]
[118 ]. Thus, several methodological studies have been performed to increase the specificity of nTMS language mapping, testing a variety of stimulation protocol optimizations (e. g., coil orientation or frequency of stimulation) [119 ]
[120 ]
[121 ].
Function-based tractography
Fiber tractography may become most powerful when combined with functional data. Activation maps derived from fMRI-based motor or language assessment or derived from nTMS mapping can be used for ROI seeding, with the aim of establishing tractography based on individual functional data [12 ]
[13 ]. In this context, previous studies have proposed that fMRI-guided fiber tracking enables reconstruction of relevant WM bundles belonging to a specific functional system, and that the evaluation of the lesion-to-activation distance (i. e., distance measurement between the tumor and a specific WM bundle as derived from fiber tractography) may be relevant to assess postoperative functional outcome [122 ]
[123 ]
[124 ]. Specifically, it has been proposed that the risk of postoperative functional decline is considerably lower in patients in whom the lesion-to-activation distance was at least 10 mm [123 ]
[124 ]. Similar to the approach using fMRI-derived activation maps as functional seeding data, motor- or language-positive points derived from nTMS can also be used to generate ROIs for tractography of the CST or language-related subcortical pathways such as the AF ([Fig. 6 ]) [125 ]
[126 ]
[127 ]
[128 ]. Furthermore, nTMS-based tractography may enable preoperative risk stratification for surgery-related motor or language impairment, making it possible to define a cutoff value of a minimum tract-to-tumor distance to avoid perioperative functional decline [129 ]
[130 ]
[131 ]
[132 ]. In essence, the combination of multi-sequence MRI with functional data from fMRI or nTMS and derived tractography represents a seamless multi-modal approach that combines structural and functional information for imaging in neuro-oncological patients ([Fig. 7 ]).
Fig. 6 Fiber tractography based on functional mapping. Fiber tracking using motor maps A and language maps B derived from navigated transcranial magnetic stimulation (nTMS) allows delineation of subcortical white matter (WM) pathways. Using the nTMS-derived motor map (motor-positive nTMS points, green) as the region of interest (ROI) for tractography allows for delineation of the corticospinal tract (CST) in somatotopic organization (separate parts for upper and lower extremity muscle representations, separated by the tumor volume in red; A ). Similarly, using the nTMS-derived language map (language-positive nTMS points, purple) as the ROI enables tracking of language-related WM pathways within the brain purely based on functional data B .
Fig. 7 Multi-modal fiber tractography. Fiber tracts belonging to the corticospinal tract (CST, orange) and the language network (pink) as derived from tractography using cortical maps of navigated transcranial magnetic stimulation (nTMS) for generation of regions of interest (ROIs). Motor and language mapping, nTMS-based tractography, and magnetic resonance imaging (MRI) can be effectively combined within a multi-modal approach to outline individual structural and functional anatomy. Fibers are fused with a fluid-attenuated inversion recovery (FLAIR) sequence in axial view A and displayed within a three-dimensional (3D) head model in sagittal view B and parasagittal view C .
Conclusion
Advanced imaging and mapping during the preoperative workup of neuro-oncological patients enables the noninvasive assessment of a multitude of characteristics relevant to tumor grading and prediction. With advancements in scanner technology including parallel imaging for the acceleration of acquisitions, a multi-sequence protocol including imaging for diffusion, perfusion, metabolism, and function in addition to conventional structural sequences becomes feasible in most patients within a reasonable scan time. The use of preoperatively acquired MRI data in combination with nTMS mapping harbors great potential for comprehensive multi-modal approaches that integrate structural with functional data.
