Purpose: Wavelet-based reconstructions of dynamic susceptibility contrast (DSC) perfusion weighted imaging (wavelet-PWI) are a new and elegant way of vascular visualization. Wavelet-PWI yields maps with a clear depiction of hypervascular tumour, as recently shown. The aim of this study was to show if the wavelet-PWI power spectrum signal in tumour tissue is associated with tumour vascularity and cell proliferation in glioblastoma multiforme (GBM).
Methods and Materials: For this IRB-approved study 12 subjects (63.0+/-14.9y; 7m) with histologically confirmed GBM were included. Target regions for biopsies were prospectively marked on contrast-enhancing GBM regions as seen on preoperative 3T MRI T1-weighted images. During subsequent neurosurgical tumour resection 27 targeted biopsies were taken intraoperatively from these target regions. All specimens were analysed for the endothelial cell marker CD31 and the proliferation marker MIB-1 and correlated to the wavelet-PWI power spectrum signal derived from dynamic susceptibility contrast (DSC)-MRI.
Results: Wavelet-PWI maps could be successfully calculated in 12/12 patients within less than 3 minutes. There was a strong correlation between wavelet-PWI power spectrum signal (median=4.41) and conventional relative cerebral blood volume (median=5.97 ml/100g) in Spearman's rank-order correlation (κ=.83, p<.05). The wavelet-PWI power spectrum signal showed a significant correlation to CD31 dichotomized to no or present staining in a logistic regression model (p<.05) and a significant correlation to MIB-1 in a nonlinear generalized model (p<.05).
Conclusion: The wavelet-PWI power spectrum signal derived from existing DSC-MRI data might be a promising new surrogate for tumour vascularity and cell proliferation in GBM.
Purpose: To assess the tumour blood flow (TBF) in the supratentorial brain gliomas by ASL-perfusion in comparison with the histopathological characteristics and 5-ALA intraoperative fluorescence.
Methods and Materials: The study group included 186 patients (97 female, 89 male, avg. 45 ± 15 years) with primary supratentorial gliomas: 47 - low-grade (LGG) and 139 - high-grade (HGG: 50-grade III, 89-grade IV). Patients were examined on a 3T MR-scanner. The pseudo-continuous ASL (pcASL) technique was used to determine TBF. TBF was normalized regarding to intact white matter (nTBF). In 66 of 186 patients tumour resection was performed using fluorescence-guided technology. All diagnoses were confirmed histopathlogically.
Results: TBF and nTBF in the groups of LGG and HGG were significantly different (p<0.001). TBF in the group of LGG was 31.55±15.75 ml/100g/min, nTBF was 1.69±0.82. In the group of HGG, the TBF was 154.63±92.86 ml/100g/min, nTBF was 8.36±5.30. The sensitivity and the specificity of ASL in diagnosis of LGG and HGG were 82.7% and 95.7%, accordingly, AUC 0.946, cutoff-64 ml/100g/min. TBF and nTBF in groups of GrIII and GrIV didn’t show statistically significant difference (p=0.08). Nevertheless, we found a significant difference in TBF and nTBF between fluorescent and non-fluorescent gliomas (p<0.05). The sensitivity and specificity of pcASL in predicting fluorescence in gliomas was 77% and 71%, accordingly, AUC-0.722, сutoff-3.3.
Conclusion: pcASL is a reliable quantitative technique for the differential diagnosis between LGG and HGG. pcASL can be used in predicting intraoperative fluorescence in gliomas. The study was supported by RFBR №18-29-01018
Purpose: The purpose of the study is to grade gliomas and predict 1p19 q deletion status on the basis of MRI characteristics. Hypothesis: low-grade gliomas are more heterogeneous, less well circumscribed, have low mean ADC values, usually in frontal and parietal lobes and are 1p/19q codeleted.
Methods and Materials: Retrospective analysis of 97 patients (53 male, 44 female; age 20-70 years) with oligodendroglioma grade II and anaplastic oligodendroglioma grade III. Associations of 1p/19q deletion status and grading with imaging characteristics were assessed. Multivariable logistic regression models were used to assess the association of grading and 1p/19q deletion with imaging characteristics and results were summarised with odds ratios and 95% confidence intervals. P values of .005 were considered statistically significant. Statistical analysis was done using SPSS® v. 19 (SPSS Inc., Chicago, IL).
Results: Thirty-one of 97 patients had 1p/19q codeleted tumours (21= Grade II, 10= Grade III). They were mostly in frontal and parietal lobes showing heterogeneity, ill-defined margins and low ADC values (mean/SD: 1185.3/283.4). 66 patients did not show 1p/19q codeletion. They had circumscribed borders, mostly in temporal lobe and insula showing homogeneity and relative high ADC values (mean/SD: 1784.2/648.7).
Conclusion: Heterogeneous signal characteristics and less well circumscribed borders on T1, T2 and SWI with lower mean ADC values are features of low-grade gliomas and these are usually 1p19q codeleted tumours. Dealing with gliomas, MRI characteristics can predict tumour grade and 1p19q deletion status. In addition, this evaluation is recommended for better treatment planning and effective patient management and prognostication.
Purpose: Identification of consistent patterns for accurate MRI-based assessment of recurrence in patients with high-grade gliomas (HGG) treated with radiotherapy, chemotherapy and bevacizumab, taking IDH1 status into account.
Methods and Materials: 31 patients with HGG (Group A included 18 patients with IDH1-mutant tumour, Group B included 13 patients with IDH1-wild-type) treated with radiotherapy, chemotherapy (temozolomide and subsequently irinotecan) in combination with bevacizumab were prospectively studied between 2016 and 2018. Changes on T1CE, T2, flair, PWI and 11C-methionine PET/CT before radiotherapy, and after 4, 8, 12 and 16 months were evaluated. Tumour volume estimation was performed with GammaPlan workstation 10.1.
Results: The earliest features of recurrence were registered at 8 months after radiotherapy for IDH1-wild-type and 12 months for IDH1-mutant tumours during continued systemic therapy, and in 96.8% of cases were in a form of an increase in T2/Flair anomaly with unconvincing methionine uptake (1.4 or less). At the same time, T1CE showed no signs of pathological contrasting and PWI with no increase in CBV also. Only subsequent observations of T2/Flair hyperintense volumes revealed higher methionine uptake values (1.8 to 4.0) and increase in CBV.
Conclusion: Wild-type gliomas are characterized by an earlier onset of progression as opposed to IDH-mutant gliomas. Well-timed identification of recurrence during bevacizumab-containing therapy is challenging due to decrease in contrasting, blood flow and moderate methionine uptake, and can be suspected by T2/Flair. To identify continued tumour growth, it is necessary to take into account all available data to start anti-relapse therapy as early as possible.
Purpose: To assess the predictive value of preoperatively assessed diffusion kurtosis imaging (DKI) metrics as prognostic factors in the 2016 World Health Organization Classification of Tumors of the Central Nervous System integrated glioma groups.
Methods and Materials: Seventy-seven patients with histopathologically confirmed treatment-naïve glioma were retrospectively assessed between 08/2013 and 10/2017 using mean kurtosis (MK) and mean diffusivity (MD) histogram parameters from DKI, overall and progression-free survival, and relevant prognostic molecular data (isocitrate dehydrogenase [IDH]; alpha-thalassemia/mental retardation syndrome X-linked, [ATRX]; chromosome 1p/19q loss of heterozygosity). The optimal cutoff-values of the metric variables were determined using receiver operating characteristic (ROC) analysis. Univariate survival data were assessed using the Kaplan-Meier method. A multivariate Cox proportional hazards model was performed on significant results from univariate analysis.
Results: There were significant differences in overall and progression-free survival between patient age (p=0.001), resection statuses (p=0.002), WHO glioma grades (p<0.0001), and integrated molecular profiles (p<0.0001). Survival was significantly better in patients with lower MK and higher MD values (p=0.009) in gliomas without chromosome 1p/19q LOH (p<0.0001) and those with retained ATRX expression (p=0.008).
Conclusion: Patient age and MK from DKI from DKI are relevant factors for preoperatively predicting overall and progression-free survival. Regarding the molecular subgroups, they are unfavourable prognostic factors in gliomas without chromosome 1p/19q LOH and those with ATRX retention.
Purpose: In glioma patients tumour cells spread far beyond the lesion that is detected by conventional MRI. Assessing non-lesional brain in glioma patients could provide crucial information about disease burden. We developed an individual measure of altered functional connectivity based on resting-state functional MRI (rsfMRI) and related this marker to WHO grade, IDH mutation status, neurocognitive performance and overall survival.
Methods and Materials: 40 patients with suspected de novo glioma were prospectively included and rsfMRI data were obtained. We evaluated the abnormality of functional connectivity by comparing each patient’s data to normative data obtained from 1000 healthy individuals. Abnormality was quantified at each voxel, resulting in an individual measure for abnormality (abnormality index, ABI). Statistical analysis was conducted adjusting for tumour volume, age and SNR of rsfMRI data.
Results: ABI maps reflected the macroscopic tumor, but also displayed alterations in non-lesional brain tissue. On a quantitative level, ABI was associated with WHO grade, both when summarized in the lesional and non-lesional hemisphere (<0.01,respectively). ABI was increased in patients with IDH-wildtype gliomas, with strongest effects in the non-lesional hemisphere (p<0.01). Association with neurocognitive performance was strongest in the lesional hemisphere (p<0.005). 8/38 patients died within the follow-up period, showing a trend towards association of overall survival with ABI (p=0.12).
Conclusion: ABI captures widespread connectivity changes in glioma patients on an individual level. ABI reflects tumour biology and correlates with neurocognitive performance and might be associated with overall survival. Individual ABI maps might therefore proof as a useful complement to conventional structural MRI.
Purpose: Dynamic glucose enhanced (DGE) CEST imaging has almost only been shown at ultra-high field (UHF) due to low effect size. First results in brain tumour patients of a DGE CEST method with fast 3D imaging developed for clinical field strength are shown herein.
Methods and Materials: CEST saturated images at different frequency offsets were acquired at 160 time points before, during and after a glucose injection (0.3 mg/kg) with 6.3s temporal resolution (Total: 16:45 min) to detect accumulation in the brain. Two glioblastoma (IDH wild-type, unmethylated MGMT promoter) patients (1: male, 70y, 2: female, 75y) and 3 healthy controls were scanned at a clinical 3T System. DGE contrast images were analysed by subtracting each image from a pre-injection baseline image: ∆DGE(t)=DGEbaseline - DGE(t).
Results: In the high-grade glioma (1), glucose uptake in the Gadolinium enhancing region could be detected approximately 4 minutes after injection with a maximum increase of ∆DGE=0.51±0.078, whereas a contralateral white matter ROI was barely affected (∆DGE=0.07±0.085) at the same time point. The second glioma (2), with the same histology and grading, showed very little gadolinium enhancement as well as no significant detectable DGE effect. Healthy controls did not show any significant DGE contrast.
Conclusion: We demonstrated that stable dynamic glucose enhanced imaging can be accomplished at clinical field strength using optimized saturation and readout parameters. First results are promising, and indicate that glucoCEST corresponds more to the disruptions of the blood-brain-barrier with Gadolinium uptake than to the molecular tumour profile or tumour grading.
Purpose: Preoperative interpretation of resectability of diffuse non-enhancing glioma is primarily based on individual surgical expertise, to identify patients who benefit most from resective surgery. Here, we compare the agreement between observed resections and preoperative estimates of neurosurgeons and a resection probability map (RPM).
Methods and Materials: 234 consecutive patients were included from two neuro-oncological centres, who had resective surgery with functional mapping between 2006 and 2012 for a supra-tentorial non-enhancing glioma. Extent of resection (EOR) and residual tumour volume (RTV) were segmented and an RPM was constructed in standard brain space. Three junior and three senior neurosurgeons estimated EOR and RTV, blinded for postoperative results. We used Bland-Altman analysis to determine the agreement between the estimations and receiver operating characteristic analysis to calculate the diagnostic accuracy of the neurosurgeons and the RPM to predict observed resections.
Results: Preoperative estimates of resection results by junior and senior neurosurgeons were significantly biased towards overestimation of EOR (4.2% and 11.2%) and underestimation of RTV (4.3 and 9.0 mL), whereas estimates of the RPM were unbiased (-2.6% and -0.2 mL, respectively). The limits of agreement were wide for neurosurgeons and for the RPM. The RPM was significantly more accurate in identifying patients in whom an EOR>40% was observed than neurosurgeons.
Conclusion: Neurosurgeons estimate preoperative resectability before surgery of a non-enhancing glioma rather accurate - with a small bias, and imprecise - with wide limits of agreement. An RPM provides unbiased resectability estimates, which can be useful for surgical decision-making, planning and education.
Purpose: To assess whether radiomics features derived from multiparametric MRI can predict the tumour grade of lower grade gliomas (LGGs; WHO grade II and grade III) and the non-enhancing LGG subgroup.
Methods and Materials: Two-hundred and four patients with LGGs from our institutional cohort were allocated to training (n = 136) and test (n = 68) sets. Postcontrast T1-weighted images, T2-weighted images, and FLAIR images were analysed to extract 250 radiomics features. Various machine learning classifiers were trained using the radiomics features to predict the glioma grade. The trained classifiers were internally validated on the institutional test set, and externally validated on a separate cohort (n = 99) from The Cancer Genome Atlas (TCGA). Classifier performance was assessed by determining the area under the curve (AUC) from receiver operating characteristic analysis. An identical process was performed in the non-enhancing LGG subgroup (institutional training set, n = 73; institutional test set, n = 37; and TCGA cohort, n = 37) to predict the glioma grade.
Results: The performance of the best classifier was good in the internal validation (AUC 0.85) and fair in the external validation (AUC 0.72) to predict the LGG grade. For the non-enhancing LGG subgroup, the performance of the best classifier was good (AUC 0.82) in the internal validation, but poor in the external validation (AUC 0.68).
Conclusion: Radiomics feature-based classifiers may be useful to predict LGG grades. However, radiomics classifiers may have limited value when applied to the non-enhancing LGG subgroup in an external cohort.