Background
IDH-wildtype glioblastoma is the most lethal and common tumor of the central nervous system, resulting in a median prognosis of 12–14 months [
1,
2] and being characterized by its high and heterogeneus vascularity [
3‐
5]. Blood supply is required for the establishment, growth, and progression of the tumor; and several mechanisms are implicated in the formation of new vessels [
3‐
5]. One of the results of these mechanisms is microvascular proliferation (MVP), which generally occurs in the core of glioblastomas by sprouting new vascular microvessels from pre-existing ones, depending on the presence of hypoxia [
3].
These pathologic heterogeneity features, including vascular proliferation, robust angiogenesis and extensive microvasculature heterogeneity could vary depending on
IDH-mutation status in high-grade gliomas [
6]. In fact, the last update of 2020 CNS glioma classification and grading [
2] differentiates between
IDH-wildtype glioblastomas and
IDH-mutant astrocytomas (previously named as IDH-mutated glioblastoma) as different type of gliomas, with different prognosis and vascular characteristics.
MVP is together to necrosis, the first criterion in the last update of 2020 CNS glioma classification and grading [
2]. It is marked by two or more blood vessels sharing a common vessel wall [
5], and interactions between tumor cells and blood vessels during microvascular proliferation seem to facilitate tumor growth [
5‐
8]. The result of MVP is the formation of large-lumen microvessels, usually with a glomeruloid appearance, that represent one of the main histopathologic hallmark of glioblastoma [
9].
Considering the relevance of this vascular process, the microvessel area (MVA), i.e., the total area covered by the microvessels in the tumor sample, and microvessel density (MVD), i.e., the number of microvessels per volume unit, have been previously investigated [
9‐
16]. Different studies suggest that MVD poorly describes the morphometric diversity of these microvessels in high-grade gliomas [
9‐
11]. However, MVA may provide a more robust clinical biomarker, useful for prognosis and grading [
9‐
11,
17‐
21]. Regardless of this evidence, the histopathological quantification of MVA is still used exclusively in the research setting. Relevant limitations, including time- and cost-expending, labor intensity, and invasiveness make it challenging for routine clinical practice.
A complementary approach to overcom
1e the limitations in MVA quantification is perfusion MRI [
9,
12]. Some studies found that measures of relative cerebral blood volume (rCBV) positively correlate with microvascular structures in different glioma tumors [
9,
11,
13‐
16]. However, these studies are few and present important limitations such as animal-based studies [
11,
13], small cohorts of glioblastoma patients [
9,
14‐
16], low number of analyzed histopathological specimens [
9,
14‐
16], or analysis with non-spatial coregistered data [
13,
20].
The integration of advanced and automatic techniques capable of calculating robust imaging markers, including rCBV, could help in high-grade glioma classification, including the diagnosis of
IDH-wildtype glioblastoma and
IDH-mutant astrocytoma. Besides, the complementary use of rCBV would involve important advantages since calculations derived from routine presurgical MRIs can be performed through automatic and robust methodologies, such as ONCOhabitats methodology [
22‐
24].
In this context, we hypothesize that MVA could be directly associated with the rCBV in
IDH-wildtype glioblastoma and this correlation can be measured using a robust MRI processing service. The areal density of microvessels on sections is an unbiased estimator of the volume density of microvessels according to the Delesse principle [
25], and we hypothesize that the volume of microvessels can be related to the rCBV. Since the typical spatial resolution, DSC sequences is 2-mm in-plane × 5-mm slices [
26], the calculation of rCBV would be reliable when it is calculated in areas larger than 2 mm.
In addition, we hypothesize that rCBV could be useful to find vascular differences between
IDH-wildtype glioblastomas and
IDH-mutant astrocytomas [
6], and therefore, supporting the new glioma classification, which differentiates between these two tumors, providing an imaging method based on routinary presurgical MRI and Artificial Intelligence techniques.
The general purpose of our study is to evaluate the potential use of rCBV, calculated with the ONCOhabitats methodology, to detect the presence or absence of microvessels in different regions of IDH- wildtype glioblastoma, and to find differences in vascularity between IDH-wildtype glioblastoma and IDH-mutant astrocytoma. The study’s specific objectives are 1) to analyze the histopathologic and radiologic correlation between the imaging markers (rCBVmean and rCBVmax) with the local MVA in IDH-wildtype glioblastoma; 2) to study whether these imaging markers can differentiate regions of the tumor with presence or absence of microvessels 3) to analyze the capacity of the rCBV to differentiate between IDH-wildtype glioblastoma and IDH-mutant astrocytoma samples.
Discussion
Microvascular proliferation is one of the main histopathologic hallmarks of glioblastomas, being key for the current glioma classification [
1,
2]. In addition, microvessel area can be considered as an independent prognostic biomarker according to previous results, in which authors reported significant longer survival in patients with glioblastoma tumors lacking the presence of new microvessels [
11,
17]. Those works suggested that tumoral microvasculature is associated with survival differences among tumors with identical histologic grades [
9,
11,
17]. However, despite its clinical potential value, direct MVA quantification is clinically unfeasible due to its time-consuming and labor-intensive nature.
By contrast, imaging markers derived from routinary MRI protocols, such as rCBV, present several advantages since they are fast to calculate, it does not represent any extra cost, and it is non-invasive compared with MVA. However, although rCBV is used for the assessment of brain tumors, it is not widely considered as a biomarker for clinical decision-making yet, probably due to the difficulty to normalize the rCBV values, which can generate confusion about prospective clinical guidelines, but also due to the lack of robust studies using spatially localized histologic correlations [
12].
In this sense, here we investigated the association between the imaging markers rCBV
mean and rCBV
max, calculated with the validated method ONCOhabitats [
22‐
24], and the MVA in
IDH-wildtype glioblastoma samples. Moreover, we analyzed the differences of rCBV between those areas of the tumor with the presence of microvessels from those regions of the tumor without evidence of microvessels.
We found significant correlations between rCBV
mean and rCBV
max and MVA when analyzing 73 tissue samples derived from 17 human
IDH-wildtype glioblastoma tumors. Also, we found significant results when we evaluated the differences of both rCBV
mean and rCBV
max between tissue blocks with presence of microvessels from those blocks defined by the absence of microvessels. Microvascular proliferation is, together to necrosis, the first criterion in the last update of 2020 CNS glioma classification and grading, since it is considered as one of the main hallmarks of high grade-gliomas (including
IDH-wildtype glioblastoma and
IDH-mutant astrocytoma). Therefore, its correlation with rCBV, makes it also a potential candidate useful in glioma classification. Despite the few similar studies conducted [
9‐
16], our results are consistent with those previously reported, finding a significant positive correlation between rCBV and MVA and significant differences in rCBV in those regions of the tumor with presence or absence of microvessels. In previous studies developed with human data and which analyze continuous variables (MVA or MV) [
14,
15], similar correlation coefficients were found (
ρ = 0.42 [
14];
ρ = 0.46 [
15] vs.
ρ = 0.43 in the present study). However, in previous studies, only 2 and 4 glioblastoma patients were enrrolled, versus the 73 samples derived from 17 patients used in the current study. Increasing the interpatient heterogeneity in our study resulted in not higher correlation coefficients. Nonetheless, the analyses are more robust and
p-values more significant. A more detailed comparison with previous studies can be found in Table S
3 of the Supporting Information.
Furthermore, in this study we have investigated the differences in rCBVmean and rCBVmax between IDH-wildtype glioblastoma and IDH-mutant astrocytoma samples. We found that blocks from IDH-wildtype glioblastoma present almost 2.5 times higher rCBV values than blocks from IDH-mutant astrocytomas. These represent promising preliminary results to propose the rCBV, calculated with ONCOhabitats, to predict with a non-invasive method the IDH status in these gliomas and a complementary method for diagnosis.
This study has some limitations. Firstly, the manual registration between morphologic MRI images and the resected tumor image could be affected by deformations of the tumor tissue morphology when resected and/or the difficulty of finding matchings between both image features. Also, the number of independent analyzed samples is not much higher despite it being higher than in other previous studies. In addition, the results derived from the comparation between IDH-wildtype and IDH-mutant samples should be considered with caution, since only 7 blocks from 2 patients were included for the IDH-mutant group.
The results derived from this work suggest the potential of imaging vascular markers calculated with the ONCOhabitats platform for helping in unmet challenges in high-grade glioma management, including glioma classification and prediction of IDH mutation status, with a non-invasive method and from the initial stage of diagnosis. We consider that the rCBV is a clinically relevant option for decision making in glioblastoma [
12,
29,
30], since it could be a complementary tool to histopathology for analyzing intratumor vascular heterogeneity at both temporal and spatial levels in a non-invasive way [
23]. This marker could be especially relevant for inoperable tumors, for which an exhaustive histopathological analysis cannot be performed. An early diagnosis, a correct classification and a more precise and personalized analysis of the glioma will have a positive impact on the patient’s treatment. Furthermore, this study opens up the possibility of evaluating tumor vascularity more correctly after antiangiogenic treatments, in addition to other prognostic/predictive markers related to tumor vascularization.
In addition, we consider useful to provide the ONCOhabitats results for Ivy GAP dataset with the purpose of enabling researchers investigating other relevant correlations between imaging-based biomarkers and histopathology for prognostic/predictive applications in glioblastoma. These results are publicly available for viewing and downloading in Zenodo (
https://zenodo.org/record/4704106#.YJu8GagzY2w) [
31].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.