Introduction
Magnetic resonance imaging (MRI) is the gold standard to diagnose meningioma and represents an important imaging tool for surgical as well as radiation treatment planning, monitoring and follow-up after treatment [
1]. Following meningioma surgery, conventional neuroimaging with MRI has limitations in distinguishing between tumour remnants and adjacent anatomical structures, postoperative changes (e.g., scars) [
2] and/or bone involvement [
3]. This is particularly important for subsequent treatment planning such as (re-)resection or radiation therapy (i.e., definition of the target volume). A further challenge in meningioma management is the early prediction of tumour recurrence or progression. Studies have shown that positron emission tomography (PET) imaging can overcome some of these challenges.
Somatostatin receptors (SSTR) are one of the main targets for PET imaging of meningiomas. High levels of SSTR subtype 2 expression were found in meningioma compared to a very low expression in adjacent structures like brain tissue or bone [
4‐
6]. Gallium-68 [68Ga]Ga–labeled SSTR ligands (DOTATOC, DOTATATE, DOTANOC) with high affinity to these receptors have therefore been shown to add valuable diagnostic information during meningioma management [
2,
7‐
9]. [68Ga]Ga-DOTATOC PET has the ability to differentiate between tumorous and non-tumorous areas in regions with low MRI contrast [
2,
8]. Due to the good tumour/non-tumour contrast, [68Ga]Ga-DOTATOC PET has also been used for radiation planning [
10‐
13] with the goal to spare as much critical tissue as possible without missing tumour. It was also shown that [68Ga]Ga-DOTATOC PET maximum standardized uptake value (SUV
max) predicted faster growth in World Health Organization (WHO) grades I and II meningioma [
9]. To determine tracer uptake intensity in PET imaging, SUV
max is used to supplement visual interpretation and it represents the tissue radioactivity concentration [
14]. A correlation between SSTR2 expression and corresponding SUV
max was found [
15] in neuronavigated tissue biopsies. No correlation of SSTR subtypes (especially 2A und 2B) with SUV
max from [68Ga]Ga-DOTATOC PET has been done so far in meningioma patients. Although, SSTR-directed PET provides additional diagnostic information, it is not routinely integrated into the first-line diagnostic evaluation of meningiomas as not every neuro-oncologic center has the availability of a PET scanner. Hence, obtaining maximum information from MRI images which are acquired in clinical routine, is desirable.
Diffusion-weighted imaging (DWI) is a broadly available MRI sequence used to provide quantitative information on the diffusion of water molecules within the brain tissue and is an integral part of standard brain tumour imaging [
16]. Radiomics is a method introduced to characterize complex structural properties from imaging data such as texture, shape, or decencies among neighbouring voxels. Radiomics has been shown to have numerous applications in neuroradiology [
17] and could help the assessment of tumour phenotypes from routine medical images by providing additional quantitative information. Indeed, several studies investigating radiomics features derived from DWI MRI in meningioma patients already exist [
18].
In this study, we aim to investigate the pathophysiological background of the SUV
max signal by comparing it to the expression of SSTR subtypes in meningioma tissue. As DWI MRI and ADC maps have been associated to information on cellular density [
19] and properties of the extracellular matrix [
20,
21] we hypothesize that the complex information described by radiomic features may contain signal related to [68Ga]Ga-DOTATOC PET/CT SUV
max values. For this purpose, we trained and evaluated a predictive model for inferring SUV
max values from radiomic features of the meningioma derived from ADC maps.
Discussion
In this retrospective study we immunohistochemically quantified SSTR subtypes in patients with resected meningioma and showed that SSTR subtypes 2A, 2B and 5 correlate significantly with SUVmax signal in [68Ga]Ga-DOTATOC PET/CT. In a second step we showed the potential of radiomic features derived from ADC maps from DWI MRI to model the [68Ga]Ga-DOTATOC PET/CT SUVmax signal in meningioma patients of different grades. The features with high explanatory value (selected in > 50% of all models) were dominated by first order, GLDM, GLCM, NGTDM and 3D Shape features.
In the first part of our study, we aimed to provide a pathophysiological background for SUV
max signal in [68Ga]Ga-DOTATOC PET. In PET, SUV came to be used as a tool to supplement visual interpretation and measures relative tissue uptake in comparison to other structures considering an optimal diagnostic threshold [
37] thereby gaining additional information on tumour margins and tumour volume for possible radiotherapy or radionuclide therapy [
13,
38].
So far, only one study [
8] investigated the correlation between SSTR expression and SUV signal in [68Ga]Ga-DOTATOC PET in patients with meningioma. In 21 meningioma patients the authors found a significant positive correlation between SUV
max and SSTR2 expression and by analysing locally different biopsies a SUV
max cut off value of 2.3 was set to define tumorous tissue. A correlation subtype analysis in meningioma patients however has not been done so far. In our study we could confirm the correlation between SSTR2 and SUV
max signal. We furthermore could show that different subtypes correlate differently with SUV
max signal (SSTR2A correlated best followed by 2B and 5).
Only recently, a comprehensive analysis from 726 tumour samples showed a clear distinction of SSTR expression in meningioma subgroups. Especially, SSTR1, 2A, and 5 showed high expression rates [
39]. The expression of SSTR2A has also shown to be an independent prognostic value regarding meningioma recurrence [
40]. This finding is also important for further therapeutic consideration, as it relates to SSTR-targeted peptide receptor radionuclide therapy (PRRT), which represents a promising approach for treating refractory meningiomas that progress after surgery and radiotherapy [
41,
42]. A deeper understanding on the distribution and role of somatostatin receptors in meningiomas is essential to further develop and refine a differentiated targeted application. PET with [68Ga]Ga-labelled somatostatin analogues has shown to assess the tumour radionuclide uptake in PRRT of meningioma prior to treatment and serves as an estimate of the achievable dose [
38]. It has been demonstrated that a lesion-based analysis of SUV
max and SUV
mean in [68Ga]Ga-DOTATOC could predict response to PRRT [
43] making [68Ga]Ga-DOTATOC PET an important predictive biomarker for PRRT. By showing that not only SSTR2 but especially SSTR2A, 2B and SSTR5 are highly correlated with SUV
max signal from [68Ga]Ga-DOTATOC we also provide more insight into the pathophysiology of the SUV
max signal and refine this pretherapeutically used diagnostic tool.
In the second part of the study, we established a predictive model to infer SUV
max values from radiomic features derived from ADC maps. Besides semantic or standard feature like tumour volume and signal intensity, radiomics has the ability to generate many more parameters that have been linked to specific tumour characteristics. In our study, 13 top ranked features which have been selected in the MRI model, were classified into five groups, as shown in Table
2. First-order statistics describe the distribution of voxel intensities within the VOI and showed to be a helpful tool in identifying brain invasion in meningiomas [
44]. GLCM, GLDM and NGTDM are examples of textural features that are computed from gray level matrices extracted from a pre-segmented tumour. These features are then organized into groups based on the respective gray level matrices used in their extraction [
45]. GLCM and GLDM provide valuable information on determining the optimal width for analysing invasiveness and peritumoural regions in meningioma [
46]. GLCM features are utilized as biomarkers of heterogeneity, offering valuable insights into the tumour microenvironment [
47]. In the case of meningiomas, NGTDM features, along with the other textural features, have demonstrated their usefulness in predicting Ki-67 and p53 status, as well as showing good performance in predicting progesterone receptor expression in high-grad meningiomas [
48,
49]. Shape features, including 3D shape features, consist of descriptors that characterize the three-dimensional size and shape of ROI. These features are independent of the gray level intensity distribution of ROI. Several clinical trial have demonstrated that shape features extracted from MRI serve as informative imaging biomarkers for predicting high WHO grade and histological brain invasion in meningioma [
50,
51].
To date, ADC radiomics in meningioma have only been investigated to predict meningioma grade [
52] and outcome [
53]. In a study of 71 meningioma patients, four statistically independent radiomic features derived from FLAIR, T1 contrast enhanced MRI and DWI MRI showed strong association with meningioma grades [
52]. Using a decision forest classifier in 152 meningioma patients, built with 23 selected texture features and the ADC value an accuracy of 79.51% to predict meningioma grade was found [
54]. Morin et al. [
53] analysed prognostic models using clinical, radiologic (including ADC maps), and radiomic features to preoperatively identify meningiomas at risk for poor outcomes. Investigating 314 meningioma patients (57% WHO grade I, 35% grade II, and 8% grade III) at two independent institutions, they found that low ADC values were associated with high-grade meningioma, and low sphericity was associated with increased local failure and worse overall survival and the prediction of meningioma grading from preoperative brain MRI demonstrated good results in a meta-analysis [
55].
Our results show that radiometric features derived from ADC maps can be significantly linked to the SUV
max signal. Therefore, our MR-based methodology could be of particular value for centers with limited access to PET imaging. Based on our findings, radiomics of ADC maps could be utilized in further studies to predict response to PRRT, similar to how it has been done by Park et al. in selecting radiotherapy for meningioma WHO grade II [
56]. Certainly, as a limitation of this study, prospective studies are needed to show the full clinical utility of our model e.g. to detect tumour, for radiation planning, and to predict tumour growth. As a further benefit we state that diagnostic based on radiomics has the advantage of being reproducible. By now, depending on the physician doing the contouring of the tumour from lower resolution PET scans, the volume and the size of a meningioma can vary depending on the SUV threshold setting as there is no standardized procedure for selecting the intensity level [
57]. This could be valuable for follow-up investigations.
In conclusion, in this study we could show that SSTR subtypes 2A, 2B and 5 correlate highly significantly with SUVmax. We developed a radiomic model based on ADC maps derived from DWI MRI to model SUVmax from [68Ga]Ga-DOTATOC PET in meningiomas. Findings that may aid to increase the diagnostic as well as therapeutic accuracy in meningioma management.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.