Introduction
The noninvasive differentiation between glioblastoma (GBM) and brain metastases is challenging in conventional MRI. Both entities comprise comparable imaging features such as peritumoral T2-hyperintense signal alterations around a contrast-enhancing tumor. Whereas the primarily expansive growth pattern of metastases leads to peritumoral edema, GBM exhibits tumor infiltration into and even beyond peritumoral signal alterations [
1,
2]. Due to the infiltrative growth pattern of GBM, recurrences frequently occur at the margins of the resection cavity [
3]. Thus supratotal tumor resection beyond the contrast-enhanced tumor is increasingly considered [
4]. Therefore, a more detailed investigation of the extent of tumor infiltration within the peritumoral zone is paramount.
Several studies have investigated the peritumoral T2 hyperintense zone in these entities seeking for imaging correlates of diffuse peritumoral infiltration in GBM: FET-PET can be used to visualize metabolically active tumor in GBM beyond the contrast-enhancing tumor [
5], and widespread MR perfusion imaging provides evidence of possible tumor neoangiogenesis within the peritumoral T2 changes in GBM, which in comparison to metastatic tumors may present with increased peritumoral rCBV [
6‐
8]. Comparative studies using diffusion-weighted MRI have revealed lower immediate peritumoral ADC-values in GBM [
9,
10] and also, in general, a relatively increased mean diffusivity within peritumoral signal alterations in metastases [
11]. In contrast to functional information on metabolism and vascularization, recent diffusion-weighted MRI techniques allow for the approximation of microstructural information in vivo from sequence parameters of millimeter-sized voxels in a mesoscopic approach [
12]. The use of diffusion tensor imaging (DTI) and its parameters fractional anisotropy (FA) was frequently investigated [
13], however, with varying results [
14]. Several studies describe increased FA in peritumoral edema in GBM/high-grade gliomas compared to metastases [
15] but there are also studies that find no significant differences in FA [
16,
17].
Novel, biophysically motivated diffusion-weighted MRI techniques such as Diffusion Microstructure Imaging (DMI) are based on a multicompartment model, which disentangles the distribution of microstructural compartments per voxel [
18]. The DMI model differentiates the three volume fractions V‑CSF (free water compartment), V‑intra (intraaxonal compartment) and V-extra (extraaxonal compartment, representing the extra-axonal cellular compartment and the extracellular matrix), as well as the corresponding diffusivities Dax-intra (axial intraaxonal diffusivity) and Dax-extra (axial extraaxonal diffusivity).
Unlike conventional MRI sequences, DMI allows for detecting previously masked information even in brain tissue with supposedly homogeneous signal representation in both inconspicuous and pathologically signal-altered structures. Compared to DTI, biophysically motivated multicompartment models like neurite orientation dispersion and density imaging (NODDI) and diffusion microstructure imaging (DMI) provide a more specific and interpretable approximation of the microstructure. For example, in T2 hyperintense lesions such as periventricular caps in NPH [
19] or peritumoral edema around intracranial masses [
11,
20] the proportionate free water content can be calculated. Moreover, structural alterations can be detected even in conventionally inconspicuous brain parenchymal structures in patients with mesial temporal sclerosis, subacute COVID-19 or Parkinson’s disease [
21‐
23].
Regarding the peritumoral zone, DMI revealed significant differences in the free water content in GBM and metastases with elevated free water around metastases, matching histopathological considerations [
20]. There were signs of increased, primarily vasogenic edema formation around metastases, while in GBM this was significantly less accentuated. Though the peritumoral T2 signal elevations were examined in whole, not taking into account the possibility of spatial gradual variations and heterogeneity. We hypothesize that gradual zone-based analysis using novel diffusion-weighted imaging metrics such as Dax-extra may detect differences between GBM and metastases that indicate an infiltrative tumor component in GBM.
Thus, we investigated the microstructure within the peritumoral zones of GBM and brain metastases using DTI and DMI in a mesoscopic approach.
Discussion
Advanced diffusion-weighted imaging revealed gradual microstructural differences within the peritumoral edema of GBM and metastases, which may be attributed to differing histopathological characteristics of these zones.
Although there was no significant difference in free water between the inner and outer peritumoral zones, there is evidence that subtle effects of free water decreasing peripherally is present at least in GBM, where we found a trend towards declining MD and V-CSF from the IZ to the OZ but without reaching significance. Beyond this, FA significantly increased outward in metastases and, on the other hand, in both entities V‑intra increased outward. According to the three-compartment theory, this is indirectly linked to a decrease in the other components [
18], even if a level of significance in those was not reached. The larger FA gradient in metastases also distinguished them from GBM, even after controlling for edema size. Previous studies focusing on peritumoral diffusion imaging corroborate our findings, as Latysheva et al. [
26] did not observe a significant difference between median FA or MD values between inner and outer peritumoral zones in GBM. Wang et al. [
17] did not observe a significant between-group difference regarding GBM and metastases groups considering inner or outer peritumoral ADC (Apparent Diffusion Coefficient), but also not for FA. Thus, the available data hint at the limited use of MD and V-CSF to further stratify the peritumoral infiltration zone in these entities, even if metastases and GBM differ in terms of mean global peritumoral water content, as previously demonstrated using V‑CSF [
20]. This primarily corresponds to the different edema characteristics with a predominantly vasogenic edema for metastases and an additional infiltrative edema component for GBM.
The discrepancy of FA between inner and outer zones in metastases suggests an additional tumor related effect in GBM beyond the vasogenic edema component. This may be related to peripheral tumor infiltration on the one hand, and indirect effects alternating the axonal microstructure on the other, ultimately “preventing” peripheral normalization of FA levels. The significantly decreased V‑intra in the inner zone in GBM is also indicative of axonal damage. This is plausible as GBM commonly present with extensive necrosis affecting both gray and white matter [
27] leading to secondary wallerian degeneration of the axons involved. Metastases commonly exhibit expansional growth, even though tumor infiltration within the immediate inner peritumoral areas has been reported [
28]. This is supported by the finding that Dax-intra was relatively increased in GBM in the inner vs. the outer zone: It is known that structural harm of the axon may not only lead to axonal loss but also an increase of the axons’ inner diameter, which has been indicated by histopathological studies in multiple sclerosis [
29], in acute ischemic stroke [
30], and may also occur in chronic traumatic brain injury [
31].
In contrast, no significant difference was detectable in metastases, which suits well with a pure vasogenic edema component with no or minimal impairment of the axons passing through.
The most interesting finding ultimately emerges in the analysis of the Dax-extra: On the one hand, Dax-extra was reduced in both groups in the inner zone, but on the other hand, a divergent effect emerged (similar to that related to Dax-intra) when considering the difference of the inner vs. outer zones. This divergence might fit to the fact that in the inner zone of metastases, a predominant space-occupying/compressing effect is present, whereas in primarily infiltrating glioblastomas a more pronounced Dax-extra reduction in the peritumoral inner zone leads to a negative gradient (Out-In). This may be caused by a pronounced mass-effect. However, as GBM usually represent diffusely infiltrating lesions, these findings may rather be explained by infiltrating elements along the axon/axonal sheath [
27], limiting axial diffusion along those. This finding is supported by previous studies indicating a gradually decreasing cellularity from in- to outside in GBMs in contrast to metastases [
32]. Besides the perivascular and leptomeningeal spaces, the brain parenchyma and white matter tracts represent main routes of invasion in GBM [
33] with molecular mechanisms so far being poorly understood. Since structural MRI biomarkers of axonal tumor spread are largely lacking, Dax-extra might be an interesting candidate warranting further investigations.
Compared to PWI and PET, diffusion imaging represents a less invasive approach, as it eliminates the need for contrast agent administration and does not involve radiation exposure. Additionally, advanced diffusion-weighted imaging is part of standard cerebral MRI protocols in primary care centers, and the sequence, as outlined in the methods section and Supplementary Table
1, does not substantially extend the overall acquisition time. Furthermore, for tumors located in eloquent areas, presurgical visualization of fiber tracts is crucial to avoid damaging critical structures such as the optic tract or corticospinal tract (CST) during resection. Thus, both microstructural diffusion MRI and tractography may be acquired within one sequence. From a therapeutic point of view, it would be conceivable to adapt surgical and/or postoperative radiotherapy with the aim of treating suspicious areas more aggressively. From a diagnostic point of view, an adjustment of the follow-up intervals would also be conceivable, depending on the quantitative findings.
A novel approach of our study is also that we captured and segmented the entirety of the peritumoral edema, while for example the DWI-based work of Lemercier et al. [
9] manually selected three ROIs and the perfusion study by Aparici-Robles et al. [
6] defined three adjacent zones of 3 mm diameter each around the contrast-enhancing tumor and thus did not account for the outer parts, in particular in extensive peritumoral signal alterations.
The study’s limitations certainly include undersampling, necessitating confirmation of results in a larger patient population. The drawn conclusions primarily rely on theoretical considerations, emphasizing the need for histopathologic correlation. However, obtaining tissue samples from peritumoral areas poses ethical challenges and additional procedural risks. The observed effects on microstructural diffusion imaging metrics encompass various peritumoral compartments with tumors of diverse genetic signatures and metastases exhibiting varying histopathology. Future investigations should address and potentially correct these heterogeneities. Notably, the measured values lack a connection to underlying white matter connectivity. The study acknowledges the possibility of distant axonal damage influencing distant white matter microstructure, highlighting the need for dedicated studies to explore these relationships. Additionally, the impact of connectivity on the presented outcomes, whether positive or negative, remains unclear and warrants further investigation.
In summary, our results indicate microstructural discrepancies in peritumoral brain tissue between GBM and metastases, with evidence of altered intraaxonal diffusivity and tumor cell infiltrates in GBM.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.