RPS 1011a - Brain tumours

Purpose:

Isocitrate dehydrogenase (IDH) mutation has a predictive prognostic value in high grade gliomas (HGG). The aim of this study was to investigate if dynamic susceptibility contrast perfusion (DSC) MRI is able to differentiate IDH mutationstatus in HGGs.

Methods and materials:

60 patients with HGG (21 glioblastomas, 17 anaplasticastrositomas; 14 IDH-mutant, 24 IDH-wild) who underwent 3T conventional MRI and DSC-MRI prior to a biopsy or surgery were included retrospectively. Perfusion (rCBV, PSR) and diffusion metrics (mean ADC, minimum ADC) in the tumour core and peritumoral non-enhancing area were measured. A Mann Whitney U test was used to detect statistically significant differences in metrics between IDH-mutant and IDH-wild type tumours. P>0.05 was considered significant.

Results:

rCBV in the tumour core was the only parameter which showed statistically significant differences between IDH-mutant (meanrCBV=3.14) and IDH-wild tumour types (meanrCBV=5.29, p<0.001). The AUC value of rCBV was 0.86 with a sensitivity of 0.83 and a specificity of 0.79 at the best cut-off point of 3.26. There were no significant differences for other metrics for both the tumour core and peritumoural area between IDH-mutant and IDH-wild tumours.

Conclusion:

rCBV's measurement of the tumour core may help to differentiate IDH mutation status in HGGs but it is not possible to differentiate by using other DSC perfusion metrics and DWI metrics of the tumour core and peritumoural non-enhancing area.

Limitations:

The inadequate number of cases.

Ethics committee approval

The local institutional review board approved this retrospective study.

Funding:

No funding was received for this work.

Purpose:

Cytoreductive surgery, radiation, and chemotherapy are the cornerstones of glioblastoma (GB) therapy. Currently, response to therapy is assessed by MRI based on two-dimensional tumour measurements, which have inherent limitations around the tumour-cavity. Manual segmentation of the lesions is time-consuming and suffers from high inter-rater-variability. To overcome these limitations and allow for a user-independent volume assessment, this study aimed to train and evaluate a state-of-the-art deep learning model (DLM) for fully-automatic GB segmentation of contrast-enhancing tumours (CET), oedema (ED), and tumour cavities (TC) on clinical routine follow-up data.

Methods and materials:

94 subjects undergoing primary GB resection were included. T1, T2, FLAIR, and contrast-enhancing (CE) T1 sequences were obtained during the course of treatment follow-up. Ground truth segmentations for the tumour compartments were obtained by consensus reading. Data was split into a training set (80 subjects) and test set (14 subjects). A multi-parametric, multi-label DLM for GB segmentation was trained on postoperative scans of the training set. A transfer learning approach was also used by including a preoperative model trained on an additional 220 GB subjects.

Results:

The DLM was tested on 85 longitudinal follow-up scans from the test set. Volume correlations (R) of DLM to manual segmentations were 0.89 (CET), 0.95 (ED), and 0.96 (TC). Automatic segmentation achieved Dice coefficients (±SD) of 0.35±0.33 (CET), 0.64±0.22 (ED), and 0.57±0.30 (TC). For CET>2 ml, the median Dice was 0.61.

Conclusion:

The proposed fully-automatic approach is robust in detecting the different tumour compartments on routine follow-up data. A DLM which is sensitive to small CET’s (<2 ml) is currently being developed. DLM’s show a strong potential to provide automated deep-learning-based GB segmentation in longitudinal GB analysis that may facilitate RANO assessment and improve the radiologist's workflow.

Limitations:

n/a

Ethics committee approval

The local ethics committee approved this retrospective study.

Funding:

No funding was received for this work.

Purpose:

No data is currently available on postoperative haemodynamic variations after neurosurgery, while these presumed variations could have imaging, clinical, and physiopathological implications. Here, we aim to characterise perioperative brain perfusion modifications after awake surgery in a cohort of patients with diffuse low-grade gliomas using MRI.

Methods and materials:

53 patients with diffuse low-grade gliomas who underwent awake surgery were included in this study. MRI acquisitions on a 3.0T magnet including perfusion-weighted imaging GE-EPI with contrast agent were performed before, within 36 h, and three months after surgery. Mean perfusion parameter values, as well as their asymmetry index, have been estimated for each hemisphere and each ROI and then compared between the three time-points. ROIs were extracted from the AAL atlas.

Results:

We observed haemodynamic alterations ipsilateral to the tumour resection, marked by a stronger asymmetry index between hemispheres and on all perfusion parameters. Regional differences in the evolution of the haemodynamic variations were found, including both transient postoperative modifications (e.g. in the putamen) and persistent postoperative alterations (e.g. in the precuneus). Finally, permanent haemodynamic impairments at 3 months post-surgery existed not only in specific regions but also translated to a general ipsilateral hemispheric impairment.

Conclusion:

The observed post-operative regional differences and persisting modifications of the cerebral vascular autoregulatory processes ask for caution in the analysis and interpretation of perfusion and BOLD-based imaging immediately and long-term after brain surgery.

Limitations:

A main limitation concerns the impact of post-operative brainshift that could induce misalignement between AAL maps and perfusion volumes between MRI sessions.

Ethics committee approval

The study was approved by the local ethical committee. Participants gave informed consent before inclusion.

Funding:

No funding was received for this work.

Purpose:

O⁶-methylguanine-methyltransferase (MGMT) promoter methylation status is an important prognostic factor for patients with glioblastomas (GBM). There are conflicting reports about the topographical distribution of GBM with versus without MGMT promotor methylation, possibly caused by molecular heterogeneity in GBM populations. We initiated this study to re-evaluate the topographical distribution of GBM with versus without MGMT promoter methylation.

Methods and materials:

Preoperative T2-weighted/FLAIR and post-contrast T1-weighted MRI scans of patients aged 18 years or older with IDH wildtype GBM were collected. Tumours were semi-automatically segmented and the topographical distribution between GBM with versus without MGMT promoter methylation presented using frequency heatmaps. Voxel-wise differences were analysed using permutation testing with threshold-free cluster enhancement.

Results:

436 IDH wildtype GBM patients were included, 211 with and 225 without MGMT promoter methylation. Visual examination suggested that when compared with MGMT unmethylated GBM, MGMT methylated GBM were more frequently located near the bifrontal and left occipital periventricular area and less frequently near the right occipital periventricular area. Statistical analyses, however, showed no significant difference in topographical distribution between MGMT methylated versus MGMT unmethylated GBM.

Conclusion:

This study re-evaluated the topographical distribution of MGMT promoter methylation in 436 newly diagnosed IDH wildtype GBM, which is the largest homogenous IDH wildtype GBM population to date, in light of the updated WHO 2016 classification. There was no statistically significant difference in anatomical localisation between MGMT methylated versus unmethylated IDH wildtype GBM.

Limitations:

The retrospective design, potential selection and confounding biases.

Ethics committee approval

The study design was approved by the Medical Ethical Committee of Erasmus MC and HMC. (IRB: no informed consent needed due to retrospective nature of study).

Funding:

The Dutch Cancer Society (KWF project number EMCR 2015-7859).

Purpose:

Glioblastomas and metastatic brain tumours are the most common malignant brain tumours. It is often difficult to differentiate between these two neoplasms on conventional MR imaging. Our goal is to evaluate the role of the T1-weighted dynamic contrast-enhanced (DCE) MRI perfusion technique to differentiate between glioblastomas and metastatic lesions.

Methods and materials:

DCE-MRI was retrospectively analysed in 19 patients with unique metastatic brain lesions and 22 patients with glioblastoma wild-type prior to any therapeutic intervention. DCE perfusion parameters calculated for the region of maximal tumoural enhancement were volume transfer constant (K^trans), efflux rate constant (K^ep), fractional volume of extravascular extracellular space (V_e), initial area under the gadolinium contrast agent concentration-time curve (IAUGC), contrast enhancement ratio (CER), and maximum slope of increase (Max Slope). Mann-Whitney U-tests were applied to these parameters.

Results:

CER was significantly higher in metastasis (0.94) than in glioblastoma (0.67) (P=0.032). V_e was marginally higher in metastasis (0.37) than in glioblastoma (0.25) (P=0.055). The remaining perfusion parameters did not show statistically significant differences.

Conclusion:

DCE-MRI showed that it has a limited potential to differentiate glioblastoma wild-type and metastases brain tumours. V_e and CER were the best parameters to discriminate between glioblastomas and metastases.

Limitations:

It is a retrospective study with a small series of cases.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

To assess whether magnetic resonance imaging (MRI) can predict the histological WHO grading of meningiomas and if it is able to differentiate WHOI from WHOII and WHOIII.

Methods and materials:

76 patients with 77 histologically proven meningiomas were enrolled (43 WHOI, 27 WHOII, and 7 WHOIII). Data was derived from multiparametric MRI sequences: pre- and post-contrast T1, T2, FLAIR, and diffusion (DWI). T1 post-contrast was used to obtain lesion volume, which was rigidly registered on the ADC map obtained from DWI in order to derive multi-parametric imaging features based on conventional ADC and an intra-voxel-incoherent-motion (IVIM) model (i.e. ADC, D, D*, and f) within the whole tumour volume. Qualitative features from morphological images were also evaluated: tumour localisation, T1- and T2- signal intensity relative to grey matter, shape, tumour-brain interface, peritumoural oedema, capsular enhancement, and tumour enhancement (homogeneous vs inhomogeneous).

Results:

Regarding DWI parameters, median values of ADC, D, and D* were higher for WHOI than WHOII and WHOIII meningiomas (p<0.05) in univariate analysis. Statistically significant differences (p<0.05) were also found in univariate analysis among qualitative features such as tumour localisation, T1- and T2- signal intensity relative to grey matter, tumour shape, peritumoural oedema, capsular enhancement, and tumour enhancement. Combining DWI and qualitative parameters, the median ADC and tumour enhancement resulted statistically different in multivariate analysis between the two groups (p<0.05), with an AUC of 0.98, a specificity of 0.93, and sensitivity of 0.97.

Conclusion:

Both DWI and qualitative features could be potential predictors of meningiomas histological grade. Lower ADC and inhomogeneous tumour enhancement suggest a high-grade meningioma. In future, these features may be useful to characterise lesions with no histological diagnosis and for the application in tailored therapy treatments.

Limitations:

n/a

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

Glioblastomas (GBM) are the most common malignant primary brain tumour and methods to improve early detection and evaluate the treatment response are highly desirable. We explored changes in whole-brain apparent diffusion coefficient (ADC) values with respect to survival (progression-free [PFS], overall [OS]) in a cohort of GBM patients followed at regular intervals until disease progression.

Methods and materials:

A total of 43 subjects met the inclusion criteria and were analysed retrospectively. Histogram data was extracted from whole-brain ADC maps including skewness, kurtosis, entropy, median, mode, and 15^th percentile (p15) and 85^th percentile (p85) values, and linear regression slopes (versus time) were fitted. Regression slope directionality (positive/negative) was subjected to univariate Cox regression. The final model was determined by a LASSO on metrics above the threshold.

Results:

Skewness, kurtosis, median, p15, and p85 were all below the threshold for both PFS and OS and were analysed further. The median regression slope directionality best modelled PFS (p=0.001; HR 3.3; 95% CI 1.6-6.7), while p85 was selected for OS (p=0.002; HR 0.29; 95% CI 0.13-0.64).

Conclusion:

Our data showed tantalising potential in the use of whole-brain ADC measurements in following GBM patients, specifically serial median ADC values which correlated with PFS, and p85 values which correlated with OS. Whole-brain ADC measurements are fast and easy to perform and free of ROI-placement bias.

Limitations:

Due to the retrospective design, detailed clinical information was not available and the assessment of potential prognostic benefit at individual time points was not possible.

Ethics committee approval:

The study was approved by the institutional review boards or ethics committees of all participating centres, and all patients provided written informed consent.

Funding:

Supported by MH CZ-DRO (NHH, 00023884). IG174301 and AZV MZ CR NV18-04-00457.