Magnetic resonance relaxometry in quantitative imaging of brain gliomas: A literature review

Abstract

Magnetic resonance (MR) relaxometry is a quantitative imaging method that measures tissue relaxation properties. This review discusses the state of the art of clinical proton MR relaxometry for glial brain tumors. Current MR relaxometry technology also includes MR fingerprinting and synthetic MRI, which solve the inefficiencies and challenges of earlier techniques. Despite mixed results regarding its capability for brain tumor differential diagnosis, there is growing evidence that MR relaxometry can differentiate between gliomas and metastases and between glioma grades. Studies of the peritumoral zones have demonstrated their heterogeneity and possible directions of tumor infiltration. In addition, relaxometry offers T2* mapping that can define areas of tissue hypoxia not discriminated by perfusion assessment. Studies of tumor therapy response have demonstrated an association between survival and progression terms and dynamics of native and contrast-enhanced tumor relaxometric profiles. In conclusion, MR relaxometry is a promising technique for glial tumor diagnosis, particularly in association with neuropathological studies and other imaging techniques.

Introduction

Gliomas account for 80% of primary malignant brain tumors, among which glioblastoma is most common. ¹ Informed tumor treatment decisions require the development of imaging techniques that identify tumor infiltration borders, and quickly and accurately detect tumor relapses. Proton MR relaxometry, or the quantification of tissue proton magnetic relaxation times and rates (which are inverse of each other), is currently receiving significant attention. Relaxometry is intended to overcome issues of conventional MR imaging, such as qualitative visual interpretation and the absence of standardization for signal intensities. This review presents a basic introduction to MR relaxometry in clinical research (Figure 1^2–4) and explores its application to brain tumors.

Figure 1.

Pulse sequence and signal intensity models for (a) conventional T1 mapping; (b) conventional T2 mapping; (c) MR fingerprinting (MRF); and (d) QRAPMASTER. Each method acquires a time series of images and the temporal signal from each pixel is fitted to a model (analytical for conventional T1 and T2 mapping, simulated using Bloch equations for MRF and QRAPMASTER). Assuming three inversion times with a long repetition time (TR) (6 s) for T1 and multi-echo acquisition for T2, the scan time of each sequence was approximately 4,608, 1,536, 5 ^2,3, and 16 ⁴ seconds per slice for T1, T2, MRF, and QRAPMASTER, respectively. An inversion-recovery sequence is conventionally used for T1 mapping, which varies the inversion time (TI) between images in the time series (a). A spin-echo sequence with variable echo time (TE) is used in conventional T2 mapping (b). A long and constant TR is generally used in both sequences to ensure full recovery between measurements, as well as eliminate confounding effects that result in significantly longer acquisition times. In MRF, TR and excitation flip angle θ are instead varied in each measurement such that the resulting signal contains contributions from both T1 and T2 in each image, thereby leading to higher measurement efficiency. The resulting signal is matched to the closest entry in the precomputed dictionary of fingerprints. In addition to higher measurement efficiency, MRF is more robust to system imperfections that deviate from the analytical models used in conventional T1 and T2 mapping.

We reviewed published articles in PubMed up to (and including) September 2022. Main keywords (in various combinations) included “glioma”, “glioblastoma”, “relaxometry”, “magnetic resonance relaxometry”, “magnetic resonance fingerprinting.”

Conventional MR relaxometry and synthetic MRI

Conventional T1 and T2 mapping techniques (Figure 1 (a) and (b)) acquire multiple images with different T1 or T2 weights, respectively, and fit the images’ signal intensities to the theoretical models of T1 or T2 relaxation. ⁵ However, these techniques are inefficient, requiring multiple repetitions and mapping of just one parameter at a time. Further, these techniques use very simple data fitting models that do not consider the effect of hardware variability, thus reducing both accuracy and reproducibility. The development of newer and more rapid pulse sequences solved the clinical feasibility problem of MR relaxometry. ⁶

T1 relaxometry was initially implemented using time-consuming inversion-recovery and saturation-recovery sequences.^7,8 A more rapid Look-Locker inversion recovery pulse sequence in modified form still constitutes one of the basic methods to map T1 relaxation.^9–11 Another and faster T1 relaxometry method employs spoiled gradient echo sequences with variable flip angles. There are also techniques of T2 confound reduction during acquisition or postprocessing. ¹² T2 measurements were originally developed using the Carr-Purcell-Meiboom-Gill multiple spin-echo pulse sequence, in which T2 values are calculated by fitting signal decay to a mono-exponential model. ¹³ Acquisition time may be optimized using mapping sequences based on steady-state free precession (SSFP) ¹¹ In addition, dependence of SSFP images on T1 and T2 constitutes a basis for simultaneous measurements of these times, thus possessing ultimate perspective for clinical studies and practice. Modern SSFP sequences are also intended for high-field MR relaxometry and allow for 3D mapping. ¹⁴

Several techniques allow for the simultaneous acquiring of multiple parameters, including relaxometric (e.g., proton density). ¹⁵ Quantitative data obtained from relaxometry may be processed to yield reconstructed (“synthetic”) MR images. ¹⁶ To date, some technologies are available as commercial products. MRF promises to provide a breakthrough in both MR relaxometry speed and accuracy by enabling accelerated and multiparametric acquisition using a single-pulse sequence (originally, inversion-recovery balanced SSFP) and accurate Bloch equation-based pattern matching (Figure 1(c)). ¹⁷ In MRF, acquisition is deliberately altered such that each tissue generates a different temporal signal evolution that exclusively corresponds to a set of MR parameter values. Signal evolutions can be simulated using a physical model (e.g., the Bloch equation) and prestored in a dictionary or database. After acquisition, a pattern recognition algorithm locates the database entry that best matches the acquired signal, as in a standard fingerprinting approach. MRF enables rapid mapping of multiple MR tissue properties using a single acquisition. Mapping is not limited to the conventional MR relaxation parameters (T1 and T2) and can also include system imperfections, such as B0 and B1, as well as perfusion ¹⁸ and chemical exchange ¹⁹ maps. Inclusion of these parameters enables robust and reproducible quantitative MRI across different scanners and time points in longitudinal studies. MRF uses multiple images per slice to reconstruct tissue parameters, which results in quantitative accuracy of the reconstructed tissue maps and, consequently, of the synthesized images. In addition, MRF relies on a simulated dictionary that takes into account the effect of all potential contributors to the measured signal. ²⁰ Although MRF mostly acquires 2D multi-slice data, ² a 3D technology is also available. ²¹

QRAPMASTER (quantification of relaxation times and proton density by multi-echo acquisition of a saturation recovery using turbo spin-echo readout)^4,22 is a competing approach for quantitative mapping of tissue parameters (Figure 1(d)). The pulse sequence consists of a saturation module that is applied to a slice, n, followed by a two-echo, fast-spin-echo (FSE) imaging module applied to a slice, m. Varying the values of n and m effectively varies the saturation delay experienced by each slice. Using four saturation delays and two echoes, the acquisition yields eight images per slice. A least-squares fit to the signal equation provides the quantitative proton density and T1, T2, R1, and R2 maps, which are then used to synthesize the desired image contrast (SyMRI). ²³ A recent multicenter trial ²⁴ deemed QRAPMASTER “non-inferior” compared to conventional images.

MR relaxometry in glioma differential diagnosis and assessment of peritumoral edema infiltration zone

Original work on clinical relaxometry in glioma diagnosis began in the 1980s and is ongoing. Earlier studies demonstrated shorter T1 values in grade I gliomas compared to higher grade gliomas. ²⁵ Grade II–IV gliomas showed significant overlap, especially in T2 and proton density. ²⁶ A recent study implementing MRF did not find significant differences between solid parts of low-grade and high-grade gliomas. ²⁷ In contrast, de Blank et al. demonstrated the ability of T1 and T2 to discriminate tumor cores of low-grade gliomas and high-grade tumors in pediatric and young adolescent patients (in the strictly pediatric cohort, only T1 afforded differentiation). The four high-grade tumors in this study also included two non-glial tumors: one medulloblastoma and one atypical teratoid/rhabdoid tumor. ²⁸ Ge et al. ²⁹ developed a combination of SyMRI T1, proton density, cerebral blood flow (pseudo-continuous arterial spin labeling perfusion, pCASL) and apparent diffusion coefficient (ADC) for powerful (95.5% sensitivity, 100% specificity) discrimination between low-grade and high-grade gliomas.

Brain tumor differential diagnosis may be added by studying the peritumoral zone, which may consist predominantly of vasogenic edema (in cases of metastasis or meningioma) or of cellular infiltration (in cases of glial tumor). ³⁰ De Blank et al. ²⁸ found that MRF technology could separate solid tumor from peritumoral white matter based on T1 and T2 values. Piper et al. showed differences of T1 and fractional anisotropy in both tumor tissue and edema of meningiomas and low-grade gliomas. Tumor core and edema in low-grade glioma had higher T1 and lower fractional anisotropy than did respective regions of meningioma. Although this may counteract the concept of “aqueous” vasogenic edema in meningioma, the authors suggested that meningiomas had a more organized structure of both core region and peritumoral edema. ³¹ These findings were also supported by De Belder et al. ³² who showed higher fractional anisotropy in both tumor core and perifocal edema of meningiomas compared to high-grade gliomas. Some expectations relate to T1rho (T1ρ) relaxometry, in which a time constant of spin-lattice tissue relaxation in the presence of a rotating frame is measured (rotating frame implies additional radiofrequency [“spin-lock”] pulse in the transverse plane) ³³ Thus, T1ρ is dependent on both T1 and T2 and is sensitive to macromolecular environment and tissue acidity. Villanueva-Meyer et al. showed that T1ρ of metastases was higher than that of low-grade and high-grade gliomas, while T1ρ values did not differ between glioma grades with higher variability. This may be explained by the nature of edema, as the presence of macromolecules and higher tissue acidity induced by infiltrative cells may lower T1ρ. ³⁰ An earlier study by Oh et al. ³⁴ showed that T2 in the peritumoral edema closest to the lesion (within 1 cm from enhancing tumor margin) in metastases and meningiomas was higher than in high-grade gliomas, which aligns with the edema origin in these tumor types (such differentiation was impossible on the basis of ADC).

In contrast, Badve et al. did not demonstrate differences in the peritumoral zone of glioblastomas and metastases according to MRF T1 or T2 but found significant T1 distinction in this region between low-grade gliomas and glioblastomas (T1 in glioblastomas was higher). In addition, tumor cores of low-grade gliomas had higher T2 than those of metastases. Differences in relaxometric values between tumor cores of glioblastoma and metastases neared significance. ²⁷ In a further study, application of MRF data texture analysis revealed significant differences between solid parts of both low-grade gliomas or high-grade gliomas and metastases. For peritumoral white matter, there were differences between all three tumor types. Lower T2 entropy and higher T1 entropy were associated with higher survival in glioblastoma patients. ³⁵ In a larger group, 3D MRF with texture analysis also showed differences between glioblastomas and metastases, between grade 2 and grade 4 gliomas as well as between grade 3 and grade 4 gliomas. T1 features also correlated with survival in glioblastoma patients. ³⁶ Oh et al. ³⁴ found high-grade gliomas to have higher T2 than did metastases and meningiomas. De Blank et al. ²⁸ demonstrated that T1 and T2 values in low-grade gliomas more closely resembled intact white matter than did those in high-grade tumors.

The capability of MR relaxometry to diagnose glioma infiltration has also been intensively studied. Earlier experimental works on murine models demonstrated gradual T1 and T2 reduction from the tumor core to the periphery. ³⁷ Although earlier research showed no difference in T2 between immediate and peripheral peritumoral zones in patients with high-grade gliomas, ³⁴ a recent study by Blystad et al. ³⁸ revealed heterogeneity in relaxation rates (R1 and R2, defined with synthetic MRI) in immediate high-grade glioma peritumoral zones, which may be caused by tumor infiltration. In addition, post-contrast slope of R1 decrease through the tumor section from core to periphery was higher than before contrast injection. Later, R1 was shown to detect significant enhancement in peritumoral zone of malignant gliomas, compared to normal-appearing white matter. This possibly reflects tumor infiltration. ³⁹ Relaxometry may be combined with other MRI modalities to determine tissue-specific correlates. Combination of T2 mapping with ¹H spectroscopy may show potential directions of more active glioma infiltrative growth with positive correlations between T2 and Cho spectrum. ⁴⁰

Glioma-related tissue hypoxia may be assessed by detection of different hemoglobin forms based on T2* changes. ⁴¹ Besides magnetic susceptibility, T2* depends on spin-spin relaxation (described by T2), and thus may be obscured by other pathologic changes. ⁴² Minimizing such influences and revealing signal changes related to susceptibility implies the usage of T2′ and represents T2* corrected for spin-spin effects. This may be calculated with the formula, 1/T2′ = 1/T2*−1/T2. T2′ values depend on hemoglobin oxygen saturation, and tissue oxygen consumption leads to T2′ decrease. High-grade gliomas show lower T2′ values compared with grade II gliomas and are characterized by inverse correlation between regional CBV and T2′, which suggests higher oxygen consumption in more malignant and highly vascularized tumors. The presence of perfusion-negative regions with low T2′ may reflect a higher degree of malignancy preceding tumor-related angiogenesis, ⁴¹ and therefore favor additional sites for biopsy. ⁴³ Possible other factors (myelin destruction, iron deposition, etc.) have similar effects on T2′. Nevertheless, hemoglobin saturation can be varied by several interventions (e.g., induced hypercapnia or oxygenation), whereas the other factors remain more static. ⁴⁴

MR relaxometry and neuropathology of gliomas

Quantitative parameters of different imaging methods need to be matched to respective neuropathological data, e.g., cellularity or proliferative index. MRI-guided localized biopsies yielded inverse correlations between tumor cellularity and fluid attenuated inversion recovery (FLAIR) signal or ADC, as well as positive correlations with signal on contrast-enhanced T1-weighted images, providing a system for MRI-based prediction of regions with varying cellularity. ⁴⁵ As for relaxometry, the number of such studies is quite modest. Kinoshita et al. found T1 of 1850–3200 ms to predict a higher tumor cell density (although T2 and ADC did not correlate with tumor cell density in the study). In addition, high ¹¹C-methiotnine uptake in PET was predicted by T1 longer than 1850 ms but shorter than 3200 ms, or by T2 higher than 115 ms but shorter than 225 ms. ⁴⁶

Our previously published results showed a tendency toward inverse correlation between Ki-67 index and T2 in enhancing malignant glioma zone (R = −0.46, p = 0.015), albeit not significant after several multiple correlation tests. We also demonstrated a significant inverse correlation between T2 and tumor blood flow in enhancing tumor (R = −0.58, p = 0.0016) and tendency towards the same correlation in peritumoral zone (R = −0.42, p = 0.03, not significant after multiple test correction). We assume that regions with higher cellular proliferation may be less hydrated and thus have lower T2 values. ⁴⁷ The potential ability of relaxometry to predict histopathological data has importance for preoperative planning, especially for biopsies.

One of the newest relaxometry application fields, also involving neuropathology, is radiogenomics. As far as conventional MRI is concerned, molecular and genetic factors (firstly, IDH and 1p/19q status) may relate to tumor signal intensity, structure, and location. ⁴⁸ The most widely known visual radiogenomic feature is likely the “T2-FLAIR mismatch” sign, which depicts high signal on T2-weighted images with T2-FLAIR hypointensity (except for peripheral hyperintense rim) in IDH-mutant 1p/19q non-codeleted astrocytoma. Although its specificity tends to ultimate values (100%), the sign lacks sensitivity (51%, or 42% in a meta-analysis).^49–51 Broadening radiogenomics to increase its diagnostic accuracy may include the quantification of tissue relaxation parameters (as has been done with ADC). ⁴⁸ Kinoshita et al. ⁵² demonstrated the T2-FLAIR mismatch sign to be associated with longer T1 and T2 values. In a complex multimodal study by Haubold et al. ⁵³ water content-based M0 maps acquired by MRF could invigorate prediction of 1p19q codeletion (combination with T1-weighted images resulted in 94% sensitivity, 91.2% specificity) and IDH1 mutation (combination with enhanced T1-weighted images and FLAIR yielded 77.3% sensitivity, 87.2% specificity). 3D MRF-based IDH1-status discrimination was also feasible in a study of Tippareddy et al. ³⁶ Kikuchi et al. ⁵⁴ demonstrated a 100% sensitivity and specificity of T2 SyMRI relaxometry in differentiating between IDH-mutant astrocytomas and oligodendrogliomas.

Figures 2 and 3 (our results, informed consent was given by the patients) illustrate typical relaxometric properties of IDH1-mutant (higher values) and IDH1-wildtype (lower values) diffuse gliomas without contrast enhancement and perfusion elevation (relaxomertic maps were obtained with Magnetic Resonance Image Compilation (MAGiC) technology based on QRAPMASTER sequence).

Figure 2.

MR images and relaxometric maps of a 41 y.o. male with diffuse non-enhancing grade II astrocytoma, IDH1-mutant. (a) – T2-weighted image; (b) – T2-FLAIR; (c) – FSPGR BRAVO, non-enhanced; (d) – FSPGR BRAVO, enhanced; (e) − T1 map (ms), (f) – T2 map (ms), (g) – proton density map (p.u. – percentage units), (h) – blood flow map (pCASL perfusion, ml/100 g/min).

Figure 3.

MR images and relaxometric maps of a 57 y.o. female with diffuse non-enhancing grade II astrocytoma, IDH1-wildtype. (a) – T2-weighted image; (b) – T2-FLAIR; (c) – FSPGR BRAVO, non-enhanced; (d) – FSPGR BRAVO, enhanced; (e) − T1 map (ms), (f) – T2 map (ms), (g) – proton density map (p.u. – percentage units), (h) – blood flow map (pCASL perfusion, ml/100 g/min).

T2 and T2* MR relaxometry in glioma treatment response assessment

Relaxometric mapping may be useful in assessing glioma response to treatment, particularly antiangiogenic therapy. Radiological tumor response to the anti-vascular endothelial growth factor (VEGF) monoclonal antibody, bevacizumab, can be shown as decreased size and/or intensity of enhancing tumor on contrast-enhanced T1-weighted images. ⁵⁵ Bevacizumab may also have direct effects on blood vessels, decreasing vascular permeability and contrast enhancement without inducing tumor regression. This situation is called “pseudoresponse.” ⁵⁶ Some gliomas may also demonstrate non-enhancing progression. According to Hattingen et al. ⁴³ modern Response Assessment in Neuro-Oncology (RANO) criteria have several disadvantages, such as subjective determination of progression on FLAIR and lack of distinction power between glioma and peritumoral zone on conventional images. In addition, the volume of the FLAIR-hyperintense area and its changes after bevacizumab treatment are not predictive of overall or progression-free survival. ⁵⁷

Several studies have examined changes in T2 relaxometric values during antiangiogenic therapy. Such dynamics may be measured using differential quantitative T2 mapping, in which T2 maps are created before and during the therapy course. Ellingson et al. ⁵⁸ and Hattingen et al. ⁴³ showed that bevacizumab therapy led to reduction of T2 in gliomas. Ellingson et al. found that lower median T2 in non-enhancing lesion after bevacizumab treatment correlated with higher overall and progression-free survival with the cutoff value of 160 ms to predict a longer period without tumor progression. Meanwhile, the median voxelwise change in T2 (ΔT2), albeit correlating with patient overall survival, did not predict survival benefit (calculations made for ΔT2 of 25 ms). ⁵⁸ Hattingen et al. showed a ΔT2 cutoff value of 26 ms for a tumor enhancing area with ΔT2 lower than 26 ms, predicting longer overall survival. Such differences may be explained by the fact that Hattingen et al. ⁴³ applied tumor segmentation into subregions, while Ellingson et al. studied the behavior of the entire non-enhancing pathologic area.

When applying differential quantitative T2 mapping, ΔT2 may be explained by edema reduction. Thus, relapse of glioma with less permeable vessels may result in recurrent tumor with lower T2 at the former site of edema with higher values of T2. ⁴³ Therefore, glioma segmentation may be useful, as therapy-related changes in various regions may have different predictive power. In further research by Ellingson et al., T2 values higher than 125 ms and lower than 250 ms were chosen to separate non-enhancing tumor from normal-appearing white matter and edema, respectively. Reduction of non-enhancing tumor by ≥50% resulted in more prominent overall and progression-free survival on bevacizumab treatment. ⁵⁹ Bontempi et al. ⁶⁰ showed areas of increased intra/extracellular water T2 beyond the FLAIR-defined periphery of lower grade gliomas before proton therapy, which evolved in two directions towards resolution (T2 decrease) or extension of FLAIR-hyperintensity (T2 increase may result from radiation injury) at the end of the therapy, which needs further investigation.

Differentiation of glioma pseudoprogession from true progression is also necessary as these states demand different treatments. ⁶¹ Both phenomena emerge as new and/or increasing areas of contrast enhancement within the treated tumor site. Current clinical protocols utilize a variety of advanced imaging techniques for brain tumor evaluation, including MRI (perfusion imaging, diffusion imaging, and spectroscopy) and position emission tomography (PET).^62–64 Nevertheless, lesion characterization in many common clinical scenarios remains ambiguous. ⁴⁵ In pseudoprogression, enhancement may be related to inflammation and treatment-related disruption of the blood-brain barrier due to endothelial damage and inflammatory response. ⁶⁵ Progression pathology relates to increased perfusion and increased content of deoxyhemoglobin in the tumor bed, and thus may be detected using susceptibility-weighted sequences. Belliveau et al. ⁶⁶ established that true glioma progression is accompanied by a higher ratio of R2* between enhancing and non-enhancing lesions (>1.3) than are pseudoprogression cases, which show a ratio of 1. As MRF provides quantitative multiparameteric maps in a single fast acquisition, this technique should be able to simplify and shorten clinical protocols, thereby improving patient comfort and throughput while yielding valuable MRI-based biomarkers.

Non-enhanced and post-contrast T1 relaxometry

Measuring T1 values in glioma dynamics on therapy may relate to non-contrast or enhanced MRI. More frequently, T1 values are obtained as an equivalent of tumor contrast enhancement. As experimental results demonstrate, T1 mapping may be useful for the depiction of contrast agent concentration within studied tissue, as well as for the study small tumors. ⁶⁷ Like differential T2 maps, relaxometric maps may be applied to the assessment of T1 changes. Thus, given the disadvantages of conventional contrast-enhanced images in monitoring glioma dynamics, the question of T1 maps in evaluation of therapy response remains topical. Subtraction techniques (to compare pre- and post-contrast images) were first studied with T1-weighted images. They showed higher capability of differentiation between residual lesions after antiangiogenic therapy and more prominent survival prediction than did conventional T1-weighted image segmentation. ⁶⁸

Lescher et al. extended the principle to relaxometric studies, combining the acquisition of differential T1 and T2 maps (subtracting maps obtained at different time points) and subtraction T1 maps (subtracting enhanced maps from native ones) on bevacizumab therapy. Non-enhanced T1 and T2 differential maps provided earlier progression detection than did conventional images. In addition, subtraction maps showed very subtle contrast enhancement suggestive of glioblastoma progression that was verified during follow-up. This phenomenon may be caused by increased blood-brain barrier permeability, which is extensive enough to cause water leakage but not extensive enough to cause vast contrast enhancement. ⁶⁹ Müller et al. segmented the area of glioblastoma-associated “cloudy” enhancement on relaxometric maps. The region showed 10–50% post-contrast T1 reduction (in contrast to >50% reduction in the tumor core). This area also extended beyond the tumor border, had elevated perfusion compared with contralateral brain matter, and was invisible using subtraction of conventional T1-weighted images, thus representing a problem for modern conventional imaging with visual assessment. Cloudy enhancing volume decrease on therapy (>21.4%) was associated with longer progression-free survival. The authors demonstrated that the cloudy enhancement area was a separate and clinically significant phenomenon irrelevant to both late enhancement and contrast agent diffusion. ⁷⁰

One intriguing technology in enhanced relaxometry is the obtainment of native R1 maps based on post-contrast R1 maps and values of proton density according to the relationship between them. Warntjes et al. showed that this approach could help to obtain post-contrast-only synthetic R1 enhancement maps that correlated well with ordinary subtraction maps. Additionally, maps calculated with this technology could reveal tumor-suspicious foci beyond the estimated borderline. ⁷¹ Histologic verification is needed to define the true nature of the discrepancy between synthetic enhancement and subtraction maps.

Conclusion

Relaxometric methods of studying central nervous system tumors were among the first in the whole of quantitative MRI. Dissatisfactory results in the early steps of relaxometry, when the modality was tested for “histologic differentiation,” can be explained by overlapping relaxometric values in different tumors. Thus, segmentation of glioma peritumoral zone may be reasonable, as it has different structure than that of metastases. No less promising is relaxometry-based study of tumor infiltration to detect perfusion-negative sites in which tumor has already spread. MR relaxometry can help to monitor the tumor dynamic in therapy. Combined application of relaxometry and diffusion, perfusion, spectroscopic imaging, as well as positron emission tomography (or either of these methods) may add particular benefit to preoperative diagnosis and planning.

ORCID iD

Ivan V Chekhonin https://orcid.org/0000-0002-6652-2472