Grading of gliomas: the road from eminence to evidence

Proteomic Analysis of Spatial Heterogeneity Identifies HMGB2 as Putative Biomarker of Tumor Progression in Adult-Type Diffuse Astrocytomas

Although grading is defined by the highest histological grade observed in a glioma, most high-grade gliomas retain areas with histology reminiscent of their low-grade counterparts. We sought to achieve the following: (i) identify proteins and molecular pathways involved in glioma evolution; and (ii) validate the high mobility group protein B2 (HMGB2) as a key player in tumor progression and as a prognostic/predictive biomarker for diffuse astrocytomas. We performed liquid chromatography tandem mass spectrometry (LC-MS/MS) in multiple areas of adult-type astrocytomas and validated our finding in multiplatform-omics studies and high-throughput IHC analysis. LC-MS/MSdetected proteomic signatures characterizing glioma evolution towards higher grades associated with, but not completely dependent, on IDH status. Spatial heterogeneity of diffuse astrocytomas was associated with dysregulation of specific molecular pathways, and HMGB2 was identified as a putative driver of tumor progression, and an early marker of worse overall survival in grades 2 and 3 diffuse gliomas, at least in part regulated by DNA methylation. In grade 4 astrocytomas, HMGB2 expression was strongly associated with proliferative activity and microvascular proliferation. Grounded in proteomic findings, our results showed that HMGB2 expression assessed by IHC detected early signs of tumor progression in grades 2 and 3 astrocytomas, as well as identified GBMs that had a better response to the standard chemoradiation with temozolomide.

Molecular diversity in isocitrate dehydrogenase-wild-type glioblastoma

In the dynamic landscape of glioblastoma, the 2021 World Health Organization Classification of Central Nervous System tumours endeavoured to establish biological homogeneity, yet isocitrate dehydrogenase-wild-type (IDH-wt) glioblastoma persists as a tapestry of clinical and molecular diversity. Intertumoural heterogeneity in IDH-wt glioblastoma presents a formidable challenge in treatment strategies. Recent strides in genetics and molecular biology have enhanced diagnostic precision, revealing distinct subtypes and invasive patterns that influence survival in patients with IDH-wt glioblastoma. Genetic and molecular biomarkers, such as the overexpression of neurofibromin 1, phosphatase and tensin homolog and/or cyclin-dependent kinase inhibitor 2A, along with specific immune cell abundance and neurotransmitters, correlate with favourable outcomes. Conversely, increased expression of epidermal growth factor receptor tyrosine kinase, platelet-derived growth factor receptor alpha and/or vascular endothelial growth factor receptor, coupled with the prevalence of glioma stem cells, tumour-associated myeloid cells, regulatory T cells and exhausted effector cells, signifies an unfavourable prognosis. The methylation status of O6-methylguanine-DNA methyltransferase and the influence of microenvironmental factors and neurotransmitters further shape treatment responses. Understanding intertumoural heterogeneity is complemented by insights into intratumoural dynamics and cellular interactions within the tumour microenvironment. Glioma stem cells and immune cell composition significantly impact progression and outcomes, emphasizing the need for personalized therapies targeting pro-tumoural signalling pathways and resistance mechanisms. A successful glioblastoma management demands biomarker identification, combination therapies and a nuanced approach considering intratumoural variability. These advancements herald a transformative era in glioblastoma comprehension and treatment.

Multiparametric MRI and T2/FLAIR mismatch complements the World Health Organization 2021 classification for the diagnosis of IDH-mutant 1p/19q non-co-deleted/ATRX-mutant astrocytoma

AIM

To investigate whether T2-weighted imaging-fluid-attenuated inversion recovery (T2/FLAIR) mismatch, T2∗ dynamic susceptibility contrast (DSC) perfusion, and magnetic resonance spectroscopy (MRS) correlated with the histological diagnosis and grading of IDH (isocitrate dehydrogenase)-mutant, 1p/19q non-co-deleted/ATRX (alpha-thalassemia mental retardation X-linked)-mutant astrocytoma.

MATERIALS

Imaging of 101 IDH-mutant diffuse glioma cases of histological grades 2-3 (2019-2021) were analysed retrospectively by two neuroradiologists blinded to the molecular diagnosis. T2/FLAIR mismatch sign is used for radio-phenotyping, and pre-biopsy multiparametric MRI images were assessed for grading purposes. Cut-off values pre-determined for radiologically high-grade lesions were relative cerebral blood volume (rCBV) ≥2, choline/creatine ratio (Cho/Cr) ≥1.5 (30 ms echo time [TE]), Cho/Cr ≥1.8 (135 ms TE).

RESULTS

Sixteen of the 101 cases showed T2/FLAIR mismatch, all of which were histogenetically confirmed IDH-mutant 1p/19q non-co-deleted/ATRX mutant astrocytomas; 50% were grade 3 (8/16) and 50% grade 2 (8/16). None showed contrast enhancement. Nine of the 16 had adequate multiparametric MRI for analysis. Any positive value by combining rCBV ≥2 with Cho/Cr ≥1.5 (30 ms TE) or Cho/Cr ≥1.8 (135 ms TE) predicted grade 3 histology with sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 100%.

CONCLUSION

The T2/FLAIR mismatch sign detected diffuse astrocytomas with 100% specificity. When combined with high Cho/Cr and raised rCBV, this predicted histological grading with high accuracy. The future direction for imaging should explore a similar integrated layered approach of 2021 classification of central nervous system (CNS) tumours combining radio-phenotyping and grading from structural and multiparametric imaging.

UnMICST: Deep learning with real augmentation for robust segmentation of highly multiplexed images of human tissues

Upcoming technologies enable routine collection of highly multiplexed (20-60 channel), subcellular resolution images of mammalian tissues for research and diagnosis. Extracting single cell data from such images requires accurate image segmentation, a challenging problem commonly tackled with deep learning. In this paper, we report two findings that substantially improve image segmentation of tissues using a range of machine learning architectures. First, we unexpectedly find that the inclusion of intentionally defocused and saturated images in training data substantially improves subsequent image segmentation. Such real augmentation outperforms computational augmentation (Gaussian blurring). In addition, we find that it is practical to image the nuclear envelope in multiple tissues using an antibody cocktail thereby better identifying nuclear outlines and improving segmentation. The two approaches cumulatively and substantially improve segmentation on a wide range of tissue types. We speculate that the use of real augmentations will have applications in image processing outside of microscopy.

Bioimaging Nucleic-Acid Aptamers with Different Specificities in Human Glioblastoma Tissues Highlights Tumoral Heterogeneity

Nucleic-acid aptamers are of strong interest for diagnosis and therapy. Compared with antibodies, they are smaller, stable upon variations in temperature, easy to modify, and have higher tissue-penetration abilities. However, they have been little described as detection probes in histology studies of human tissue sections. In this study, we performed fluorescence imaging with two aptamers targeting cell-surface receptors EGFR and integrin α5β1, both involved in the aggressiveness of glioblastoma. The aptamers’ cell-binding specificities were confirmed using confocal imaging. The affinities of aptamers for glioblastoma cells expressing these receptors were in the 100–300 nM range. The two aptamers were then used to detect EGFR and integrin α5β1 in human glioblastoma tissues and compared with antibody labeling. Our aptafluorescence assays proved to be able to very easily reveal, in a one-step process, not only inter-tumoral glioblastoma heterogeneity (differences observed at the population level) but also intra-tumoral heterogeneity (differences among cells within individual tumors) when aptamers with different specificities were used simultaneously in multiplexing labeling experiments. The discussion also addresses the strengths and limitations of nucleic-acid aptamers for biomarker detection in histology.

Diagnostic Accuracy of Magnetic Resonance Spectroscopy in Predicting the Grade of Glioma Keeping Histopathology as the Gold Standard

Background

Gliomas are the most prevalent intrinsic tumors of the central nervous system and are categorized from grade I to grade IV. Magnetic resonance imaging (MRI) provides exact diagnosis, prognosis, and assessment of tumor response to current chemotherapy/immunotherapy and radiation therapy. With histopathology serving as the gold standard, we aimed to assess the diagnostic accuracy of magnetic resonance spectroscopy (MRS) in predicting glioma grade.

Methodology

This cross-sectional study was conducted in the Department of Radiology, KRL Hospital, Islamabad, from December 15, 2019, to September 30, 2021. After providing written consent, 80 patients with untreated gliomas were included in this study. The voxel of interest was identified using MRI brain conventional contrast-enhanced sequences to assess the grade of the gliomas and link it to the histology report. Following this identification, tissue metabolites were calculated using MRS.

Results

The patients’ age ranged from 13 to 80 years, with a mean age of 49.5 years. Male patients comprised 57.5% of the total study population, while female patients comprised 42.5%. Overall, 23.75% of patients had low-grade tumors, while 76.25% had high-grade tumors. Low-grade tumors had a choline (Cho)/creatine (Cr) metabolite ratio of 1.7421, whereas high-grade tumors had an average Cho/Cr metabolite ratio of 2.5575. N-acetyl aspartate (NAA)/Cr ratio was 1.6368 in low grade and 0.6734 in high-grade tumors. Sensitivity of 77% and specificity of 84.2% were noted, with 78.75% diagnostic accuracy for the Cho/Cr ratio.

Conclusions

Multivoxel MRS has been shown to reliably predict the grade of gliomas despite its non-invasive nature and lack of procedural challenges. When used together Cho/Cr and NAA/Cr ratios and histopathology can accurately determine tumor grade and can be used as a supplementary non-invasive technique.

Prognostic and predictive impact of MGMT promoter methylation status in high risk grade II glioma

BACKGROUND

MGMT promoter methylation has been associated with favorable prognosis and survival outcomes in patients with glioblastoma and WHO grade III glioma. However, the effects of promoter methylation of MGMT in patients with WHO grade II gliomas have not been established. The purpose of the current study is to evaluate the prognostic impact and predictive values of MGMT methylation in patients with grade II glioma.

METHODS

The National Cancer Database (NCDB) was queried (2004-2016) for patients with newly diagnosed grade II glioma. Demographics and clinical characteristics of these patients were examined. Statistics included Kaplan-Meier overall survival (OS) analysis alongside Cox proportional hazards modeling.

RESULTS

A total of 11,223 patients met the selection criteria; 1252 patients (11%) had MGMT testing. Of the patients who had MGMT testing, 58.5% were MGMT methylated (mMGMT), and 43.5% were MGMT unmethylated (uMGMT). mMGMT patients had greater median overall survival (77.3 months) than both uMGMT patients (42.6 months) and patients with no MGMT status reported (61.9 months (p < 0.001 for both). mMGMT was also associated with improved OS, when compared to patients with uMGMT, for patients receiving adjuvant chemoradiation or adjuvant radiation therapy.

CONCLUSIONS

This is the largest study to date demonstrating both the prognostic and predictive impact of MGMT methylation on patients with grade II glioma. The current results show that mMGMT is a prognostic factor and possibly a predictive biomarker for grade II glioma patients. MGMT methylation status can be used to determine and stratify patients by risk levels, and thus select patients for treatment intensification.

IMPORTANCE OF STUDY

The present study is the largest to date examining the prognostic and predictive significance of MGMT methylation (mMGMT) in patients with WHO grade II glioma. The results suggest that mMGMT is prognostic with increasing overall survival rates for patients with mMGMT compared to uMGMT patients. The results also suggest that mMGMT is predictive as shown by improved overall survival in patients receiving gross total resection, adjuvant chemoradiation or adjuvant radiation therapy, but no difference was observed in patients receiving adjuvant chemotherapy or no adjuvant treatment.

Prognostic risk stratification of gliomas using deep learning in digital pathology images

Background

Evaluation of tumor-tissue images stained with hematoxylin and eosin (H&E) is pivotal in diagnosis, yet only a fraction of the rich phenotypic information is considered for clinical care. Here, we propose a survival deep learning (SDL) framework to extract this information to predict glioma survival.

Methods

Digitized whole slide images were downloaded from The Cancer Genome Atlas (TCGA) for 766 diffuse glioma patients, including isocitrate dehydrogenase (IDH)-mutant/1p19q-codeleted oligodendroglioma, IDH-mutant/1p19q-intact astrocytoma, and IDH-wildtype astrocytoma/glioblastoma. Our SDL framework employs a residual convolutional neural network with a survival model to predict patient risk from H&E-stained whole-slide images. We used statistical sampling techniques and randomized the transformation of images to address challenges in learning from histology images. The SDL risk score was evaluated in traditional and recursive partitioning (RPA) survival models.

Results

The SDL risk score demonstrated substantial univariate prognostic power (median concordance index of 0.79 [se: 0.01]). After adjusting for age and World Health Organization 2016 subtype, the SDL risk score was significantly associated with overall survival (OS; hazard ratio = 2.45; 95% CI: 2.01 to 3.00). Four distinct survival risk groups were characterized by RPA based on SDL risk score, IDH status, and age with markedly different median OS ranging from 1.03 years to 14.14 years.

Conclusions

The present study highlights the independent prognostic power of the SDL risk score for objective and accurate prediction of glioma outcomes. Further, we show that the RPA delineation of patient-specific risk scores and clinical prognostic factors can successfully demarcate the OS of glioma patients.

Somatic mitochondrial DNA mutations in different grades of glioma

Aim: Mitochondrial DNA (mtDNA) alterations play an important role in the multistep processes of cancer development. Gliomas are among the most diagnosed brain cancer. The relationship between mtDNA alterations and different grades of gliomas are still elusive. This study aimed to elucidate the profile of somatic mtDNA mutations in different grades of gliomas and correlate it with clinical phenotype. Materials & methods: Forty histopathologically confirmed glioma tissue samples and their matched blood were collected and subjected for mtDNA sequencing. Results & conclusion: About 75% of the gliomas harbored at least one somatic mutation in the mtDNA gene, and 45% of these mutations were pathogenic. Mutations were scattered across the mtDNA genome, and the commonest nonsynonymous mutations were located at complex I and IV of the mitochondrial respiratory chain. These findings may have implication for future research to determine the mitochondrial energetics and its downstream metabolomics on gliomas.

MRI Features May Predict Molecular Features of Glioblastoma in Isocitrate Dehydrogenase Wild-Type Lower-Grade Gliomas

BACKGROUND AND PURPOSE

Isocitrate dehydrogenase (IDH) wild-type lower-grade gliomas (histologic grades II and III) with epidermal growth factor receptor (EGFR) amplification or telomerase reverse transcriptase (TERT) promoter mutation are reported to behave similar to glioblastoma. We aimed to evaluate whether MR imaging features could identify a subset of IDH wild-type lower-grade gliomas that carry molecular features of glioblastoma.

MATERIALS AND METHODS

In this multi-institutional retrospective study, pathologically confirmed IDH wild-type lower-grade gliomas from 2 tertiary institutions and The Cancer Genome Atlas constituted the training set (institution 1 and The Cancer Genome Atlas, 64 patients) and the independent test set (institution 2, 57 patients). Preoperative MRIs were analyzed using the Visually AcceSAble Rembrandt Images and radiomics. The molecular glioblastoma status was determined on the basis of the presence of EGFR amplification and TERT promoter mutation. Molecular glioblastoma was present in 73.4% and 56.1% in the training and test sets, respectively. Models using clinical, Visually AcceSAble Rembrandt Images, and radiomic features were built to predict the molecular glioblastoma status in the training set; then they were validated in the test set.

RESULTS

In the test set, a model using both Visually AcceSAble Rembrandt Images and radiomic features showed superior predictive performance (area under the curve = 0.854) than that with only clinical features or Visually AcceSAble Rembrandt Images (areas under the curve = 0.514 and 0.648, respectively; P < . 001, both). When both Visually AcceSAble Rembrandt Images and radiomics were added to clinical features, the predictive performance significantly increased (areas under the curve = 0.514 versus 0.863, P < .001).

CONCLUSIONS

MR imaging features integrated with machine learning classifiers may predict a subset of IDH wild-type lower-grade gliomas that carry molecular features of glioblastoma.

Glioma Grading via Analysis of Digital Pathology Images Using Machine Learning

Cancer pathology reflects disease progression (or regression) and associated molecular characteristics, and provides rich phenotypic information that is predictive of cancer grade and has potential implications in treatment planning and prognosis. According to the remarkable performance of computational approaches in the digital pathology domain, we hypothesized that machine learning can help to distinguish low-grade gliomas (LGG) from high-grade gliomas (HGG) by exploiting the rich phenotypic information that reflects the microvascular proliferation level, mitotic activity, presence of necrosis, and nuclear atypia present in digital pathology images. A set of 735 whole-slide digital pathology images of glioma patients (median age: 49.65 years, male: 427, female: 308, median survival: 761.26 days) were obtained from TCGA. Sub-images that contained a viable tumor area, showing sufficient histologic characteristics, and that did not have any staining artifact were extracted. Several clinical measures and imaging features, including conventional (intensity, morphology) and advanced textures features (gray-level co-occurrence matrix and gray-level run-length matrix), extracted from the sub-images were further used for training the support vector machine model with linear configuration. We sought to evaluate the combined effect of conventional imaging, clinical, and texture features by assessing the predictive value of each feature type and their combinations through a predictive classifier. The texture features were successfully validated on the glioma patients in 10-fold cross-validation (accuracy = 75.12%, AUC = 0.652). The addition of texture features to clinical and conventional imaging features improved grade prediction compared to the models trained on clinical and conventional imaging features alone (p = 0.045 and p = 0.032 for conventional imaging features and texture features, respectively). The integration of imaging, texture, and clinical features yielded a significant improvement in accuracy, supporting the synergistic value of these features in the predictive model. The findings suggest that the texture features, when combined with conventional imaging and clinical markers, may provide an objective, accurate, and integrated prediction of glioma grades. The proposed digital pathology imaging-based marker may help to (i) stratify patients into clinical trials, (ii) select patients for targeted therapies, and (iii) personalize treatment planning on an individual person basis.

Evolving Insights into the Molecular Neuropathology of Diffuse Gliomas in Adults

Recent advances in molecular analysis and genome sequencing have prompted a paradigm shift in neuropathology. This article discusses the discovery and clinical relevance of molecular biomarkers in diffuse gliomas in adults and how these biomarkers led to revision of the World Health Organization classification of these tumors. We relate progress in clinical classification to an overview of studies using molecular profiling to study gene expression and DNA methylation to categorize diffuse gliomas in adults and issues dealing with intratumoral heterogeneity. These efforts will refine the taxonomy of diffuse gliomas, facilitate selection of appropriate treatment regimens, and ultimately improve patient's lives.

Prediction of Overall Survival Based on Isocitrate Dehydrogenase 1 Mutation and 18F-FDG Uptake on PET/CT in Patients With Cerebral Gliomas

PURPOSE

This retrospective study aimed to correlate F-FDG uptake on PET/CT with isocitrate dehydrogenase enzyme isoform 1 (IDH1) mutation in patients with cerebral gliomas. Hierarchical interactions between factors affecting overall survival (OS) were also examined.

METHODS

In 59 patients with glioma, the ratio of the SUVmax of a glioma to the SUVmean of the contralateral cortex (G/C ratio) on F-FDG PET/CT and the presence of IDH1 mutation were correlated. The prognostic value of clinicopathologic factors and G/C ratio for OS were assessed using a Cox proportional hazards model and classification and regression tree models.

RESULTS

The mean G/C ratio of IDH1-mutant tumors was significantly lower than that of IDH1 wild-type tumors (0.73 vs 1.14, P = 0.004). In multivariate analysis, IDH1-mutant and G/C ratio were significant for OS. The classification and regression tree modeling identified 3 risk groups for OS (group 1: IDH1 mutant [hazard ratio, 0.2]; group 2: G/C ratio ≤0.8 with IDH1 wild type [hazard ratio, 0.83]; group 3: G/C ratio >0.8 with IDH1 wild type [hazard ratio, 1.9]) (overall P < 0.001). The mean OS was 37.0 months in group 1, 28.6 months in group 2, and 20.7 months in group 3, respectively, showing significant differences among the groups (group 1 vs group 2: P = 0.023, group 2 vs group 3: P = 0.049, group 1 vs group3: P < 0.001).

CONCLUSIONS

F-FDG uptake of IDH1-mutant gliomas was significantly lower than that of IDH1 wild-type gliomas. IDH1 mutation was the most important factor in identifying patients with the best prognosis, whereas increased F-FDG uptake provided additional prognostic information for predicting poor OS among patients with IDH1 wild-type gliomas.

Diffusion tensor image features predict IDH genotype in newly diagnosed WHO grade II/III gliomas

We hypothesized that machine learning analysis based on texture information from the preoperative MRI can predict IDH mutational status in newly diagnosed WHO grade II and III gliomas. This retrospective study included in total 79 consecutive patients with a newly diagnosed WHO grade II or III glioma. Local binary pattern texture features were generated from preoperative B0 and fractional anisotropy (FA) diffusion tensor imaging. Using a training set of 59 patients, a single hidden layer neural network was then trained on the texture features to predict IDH status. The model was validated based on the prediction accuracy calculated in a previously unseen set of 20 gliomas. Prediction accuracy of the generated model was 92% (54/59 cases; AUC = 0.921) in the training and 95% (19/20; AUC = 0.952) in the validation cohort. The ten most important features were comprised of tumor size and both B0 and FA texture information, underlining the joint contribution of imaging data to classification. Machine learning analysis of DTI texture information and tumor size reliably predicts IDH status in preoperative MRI of gliomas. Such information may increasingly support individualized surgical strategies, supplement pathological analysis and highlight the potential of radiogenomics.

Correlative analyses of the expression levels of PIAS3, p-SHP2, SOCS1 and SOCS3 with STAT3 activation in human astrocytomas

The importance of signal transducer and activator of transcription 3 (STAT3) signaling in the growth and survival of glioblastoma cells has been well documented, while the reasons leading to STAT3 activation remains to be elucidated. Suppressors of cytokine signaling (SOCS) 1 and SOCS3, SH2 domain‑containing phosphatase (SHP2) and protein inhibitors of activated STAT3 (PIAS3) are known to inhibit STAT3 signal transduction, while their expression statuses in the four grades of astrocytomas and relevance with STAT3 activation remain to be described. The present study aimed to address these issues by tissue microarray‑based immunohistochemical profiling the expression levels of phosphorylated (p)‑STAT3, SOCS1, SOCS3, PIAS3 and p‑SHP2. The results revealed that p‑STAT3 nuclear translocation was rarely observed in non‑cancerous brain tissues and its frequencies were increased in a tumor grade‑associated manner (65.2, 77.1, 81.8 and 85.7% for grade I‑IV, respectively). PIAS3, p‑SHP2, SOCS1 and SOCS3 were expressed in higher levels (++ and +++) in 63.6, 90, 87.5 and 81.8% of tumor surrounding brain tissues, which reduced to 13.1, 47.8, 33.3 and 50% in grade I, 11.4, 65.7, 58.3 and 77.1% in grade II, 9.1, 63.6, 38.1 and 31.8% in grade III and 7.1, 66.7, 30.8 and 7.1% in grade IV astrocytomas. The above results revealed that although the expression levels of SOCS1, SOCS3 and, in particular, p‑SHP2, tend to decrease in the four types of astrocytomas, PIAS3 downregulation is more negatively correlated with STAT3 activation in the stepwise progress of astrocytomas and would indicate an unfavorable outcome.

Histologic classification of gliomas

Gliomas form a heterogeneous group of tumors of the central nervous system (CNS) and are traditionally classified based on histologic type and malignancy grade. Most gliomas, the diffuse gliomas, show extensive infiltration in the CNS parenchyma. Diffuse gliomas can be further typed as astrocytic, oligodendroglial, or rare mixed oligodendroglial-astrocytic of World Health Organization (WHO) grade II (low grade), III (anaplastic), or IV (glioblastoma). Other gliomas generally have a more circumscribed growth pattern, with pilocytic astrocytomas (WHO grade I) and ependymal tumors (WHO grade I, II, or III) as the most frequent representatives. This chapter provides an overview of the histology of all glial neoplasms listed in the WHO 2016 classification, including the less frequent "nondiffuse" gliomas and mixed neuronal-glial tumors. For multiple decades the histologic diagnosis of these tumors formed a useful basis for assessment of prognosis and therapeutic management. However, it is now fully clear that information on the molecular underpinnings often allows for a more robust classification of (glial) neoplasms. Indeed, in the WHO 2016 classification, histologic and molecular findings are integrated in the definition of several gliomas. As such, this chapter and Chapter 6 are highly interrelated and neither should be considered in isolation.

Recent advances in subtyping tumors of the central nervous system using molecular data

INTRODUCTION

Primary brain tumors account for substantial morbidity and mortality. They often infiltrate the brain diffusely, continue growing, and cause adverse events, such as headaches, seizures, and neurological deficits. The classification of primary brain tumors, based for decades on histology, has been fundamentally changed by the World Health Organization in 2016 by incorporation of molecular data. Areas covered: Literature from glioblastomas, high- and low-grade astrocytic, oligodendroglial, glioneuronal and ependymal tumors from the last five years were reviewed. Results from comprehensive molecular profiling of neoplasms and impact of recent molecular subtyping on neuropathological diagnosis are presented. Expert commentary: The identification of frequent acquired mutations shows that adult and pediatric glioblastomas have divergent biology with differing prognoses. Astrocytoma and oligodendroglioma are more closely related than previously thought. Molecular profiling now enables the precise classification of most diffuse gliomas into three clinically and therapeutically different subtypes according to the presence or absence of IDH mutation and 1p/19q codeletion. New subgroups with different clinical outcomes and anatomic locations have emerged in ependymomas and pediatric embryonal tumors.

Multiparametric MRI-based differentiation of WHO grade II/III glioma and WHO grade IV glioblastoma

Non-invasive, imaging-based examination of glioma biology has received increasing attention in the past couple of years. To this end, the development and refinement of novel MRI techniques, reflecting underlying oncogenic processes such as hypoxia or angiogenesis, has greatly benefitted this research area. We have recently established a novel BOLD (blood oxygenation level dependent) based MRI method for the measurement of relative oxygen extraction fraction (rOEF) in glioma patients. In a set of 37 patients with newly diagnosed glioma, we assessed the performance of a machine learning model based on multiple MRI modalities including rOEF and perfusion imaging to predict WHO grade. An oblique random forest machine learning classifier using the entire feature vector as input yielded a five-fold cross-validated area under the curve of 0.944, with 34/37 patients correctly classified (accuracy 91.8%). The most important features in this classifier as per bootstrapped feature importance scores consisted of standard deviation of T1-weighted contrast enhanced signal, maximum rOEF value and cerebral blood volume (CBV) standard deviation. This study suggests that multimodal MRI information reflects underlying tumor biology, which is non-invasively detectable through integrative data analysis, and thus highlights the potential of such integrative approaches in the field of radiogenomics.

Compliance with reporting guidelines by Australian pathologists: an audit of the quality of histopathology reporting in high-grade glioma

BACKGROUND

Diagnostic pathology reports inform management plans for patients with glioma, and there is an increasing clinical need for molecular testing. We assessed the quality of histopathology reports of grade III/IV gliomas.

METHODS

Reports were obtained as part of a tumor biobank. From 720 pathology reports, 594 eligible reports were assessed for 28 elements derived from published checklists. A summary quality score incorporated 9 critical parameters for clinical decision making: diagnosis using World Health Organization 2007 criteria; cell type; grade; narrative supporting cell type and grade; absence of equivocal language; conclusion reporting cell type and grade; and conclusion aligned with report narrative.

RESULTS

Of 594 eligible reports, the final conclusion was not supported by the report narrative in 122 (21%). Tumor classification and grade were not supported by the narrative in 105 (18%) and 36 (6%) reports, respectively. Only 145 (24%) reports fulfilled all 9 quality criteria, while 25% contained 6 or fewer key quality indices. Report quality was higher when pathologists had neuropathology subspecialization, when a grade IV tumor was reported, and when the specimen was from an initial resection or grade-progressed tumor rather than recurrent high-grade glioma. Use of molecular testing increased over time, from 29% to 48% over four quartiles of the study. Molecular testing was more frequently done where oligodendroglial elements were reported.

CONCLUSION

A significant proportion of reports failed to meet key indicators of report quality. Pathology reporting is critical in communicating between pathologists and treating clinicians. Clinicians should be aware of reporting quality and seek clarification when required.

5-hydroxymethylcytosine loss is associated with poor prognosis for patients with WHO grade II diffuse astrocytomas

Currently, the reliable prognostic biomarkers for WHO grade II diffuse astrocytomas (DA) are still limited. We investigated the relations between the level of 5-Hydroxymethylcytosine (5hmC), an oxidated production of 5-methylcytosine (5mC) by the ten eleven translocated (TET) enzymes, and clinicopathological features of glioma patients. With an identified anti-5hmC antibody, we performed immunohistochemistry in 287 glioma cases. We detected that 5hmC variably reduced in most gliomas and 5hmC reduction was closely associated with higher pathological grades and shortened survival of glioma patients. In multivariate analysis, 5hmC had no independent prognostic value in the entire patient cohort. However, multivariate analysis within subtypes of gliomas revealed that 5hmC was still a prognostic marker confined to DA. In addition, we detected that IDH1 mutation by DNA sequencing was associated with favorable survival within DA. Lastly, we detected that the combination of 5hmC/KI67 was a useful prognostic marker for restratification of DA.

Brain Tumor Epidemiology - A Hub within Multidisciplinary Neuro-oncology. Report on the 15th Brain Tumor Epidemiology Consortium (BTEC) Annual Meeting, Vienna, 2014

The Brain Tumor Epidemiology Consortium (BTEC) is an open scientific forum, which fosters the development of multi-center, international and inter-disciplinary collaborations. BTEC aims to develop a better understanding of the etiology, outcomes, and prevention of brain tumors (http://epi.grants.cancer.gov/btec/). The 15th annual Brain Tumor Epidemiology Consortium Meeting, hosted by the Austrian Societies of Neuropathology and Neuro-oncology, was held on September 9 - 11, 2014 in Vienna, Austria. The meeting focused on the central role of brain tumor epidemiology within multidisciplinary neuro-oncology. Knowledge of disease incidence, outcomes, as well as risk factors is fundamental to all fields involved in research and treatment of patients with brain tumors; thus, epidemiology constitutes an important link between disciplines, indeed the very hub. This was reflected by the scientific program, which included various sessions linking brain tumor epidemiology with clinical neuro-oncology, tissue-based research, and cancer registration. Renowned experts from Europe and the United States contributed their personal perspectives stimulating further group discussions. Several concrete action plans evolved for the group to move forward until next year's meeting, which will be held at the Mayo Clinic at Rochester, MN, USA.

Oligodendroglioma: pathology, molecular mechanisms and markers

For nearly a century, the diagnosis and grading of oligodendrogliomas and oligoastrocytomas has been based on histopathology alone. Roughly 20 years ago, the first glioma-associated molecular signature was found with complete chromosome 1p and 19q codeletion being particularly common in histologically classic oligodendrogliomas. Subsequently, this codeletion appeared to not only carry diagnostic, but also prognostic and predictive information, the latter aspect only recently resolved after carefully constructed clinical trials with very long follow-up times. More recently described biomarkers, including the non-balanced translocation leading to 1p/19q codeletion, promoter hypermethylation of the MGMT gene, mutations of the IDH1 or IDH2 gene, and mutations of FUBP1 (on 1p) or CIC (on 19q), have greatly enhanced our understanding of oligodendroglioma biology, although their diagnostic, prognostic, and predictive roles are less clear. It has therefore been suggested that complete 1p/19q codeletion be required for the diagnosis of 'canonical oligodendroglioma'. This transition to an integrated morphological and molecular diagnosis may result in the disappearance of oligoastrocytoma as an entity, but brings new challenges as well. For instance it needs to be sorted out how (histopathological) criteria for grading of 'canonical oligodendrogliomas' should be adapted, how pediatric oligodendrogliomas (known to lack codeletions) should be defined, which platforms and cut-off levels should ideally be used for demonstration of particular molecular aberrations, and how the diagnosis of oligodendroglioma should be made in centers/countries where molecular diagnostics is not available. Meanwhile, smart integration of morphological and molecular information will lead to recognition of biologically much more uniform groups within the spectrum of diffuse gliomas and thereby facilitate tailored treatments for individual patients.

IDH mutations: genotype-phenotype correlation and prognostic impact

IDH1/2 mutation is the most frequent genomic alteration found in gliomas, affecting 40% of these tumors and is one of the earliest alterations occurring in gliomagenesis. We investigated a series of 1305 gliomas and showed that IDH mutation is almost constant in 1p19q codeleted tumors. We found that the distribution of IDH1^R132H, IDH1^nonR132H, and IDH2 mutations differed between astrocytic, mixed, and oligodendroglial tumors, with an overrepresentation of IDH2 mutations in oligodendroglial phenotype and an overrepresentation of IDH1^nonR132H in astrocytic tumors. We stratified grade II and grade III gliomas according to the codeletion of 1p19q and IDH mutation to define three distinct prognostic subgroups: 1p19q and IDH mutated, IDH mutated—which contains mostly TP53 mutated tumors, and none of these alterations. We confirmed that IDH mutation with a hazard ratio = 0.358 is an independent prognostic factor of good outcome. These data refine current knowledge on IDH mutation prognostic impact and genotype-phenotype associations.

Involvement of urokinase-type plasminogen activator system in cancer: an overview

Currently, there are several studies supporting the role of urokinase-type plasminogen activator (uPA) system in cancer. The association of uPA to its receptor triggers the conversion of plasminogen into plasmin. This process is regulated by the uPA inhibitors (PAI-1 and PAI-2). Plasmin promotes degradation of basement membrane and extracellular matrix (ECM) components as well as activation of ECM latent matrix metalloproteases. Degradation and remodeling of the surrounding tissues is crucial in the early steps of tumor progression by facilitating expansion of the tumor mass, release of tumor growth factors, activation of cytokines as well as induction of tumor cell proliferation, migration, and invasion. Hence, many tumors showed a correlation between uPA system component levels and tumor aggressiveness and survival. Therefore, this review summarizes the structure of the uPA system, its contribution to cancer progression, and the clinical relevance of uPA family members in cancer diagnosis. In addition, the review evaluates the significance of uPA system in the development of cancer-targeted therapies.

A novel semi-supervised methodology for extracting tumor type-specific MRS sources in human brain data

Background

The clinical investigation of human brain tumors often starts with a non-invasive imaging study, providing information about the tumor extent and location, but little insight into the biochemistry of the analyzed tissue. Magnetic Resonance Spectroscopy can complement imaging by supplying a metabolic fingerprint of the tissue. This study analyzes single-voxel magnetic resonance spectra, which represent signal information in the frequency domain. Given that a single voxel may contain a heterogeneous mix of tissues, signal source identification is a relevant challenge for the problem of tumor type classification from the spectroscopic signal.

Methodology/Principal Findings

Non-negative matrix factorization techniques have recently shown their potential for the identification of meaningful sources from brain tissue spectroscopy data. In this study, we use a convex variant of these methods that is capable of handling negatively-valued data and generating sources that can be interpreted as tumor class prototypes. A novel approach to convex non-negative matrix factorization is proposed, in which prior knowledge about class information is utilized in model optimization. Class-specific information is integrated into this semi-supervised process by setting the metric of a latent variable space where the matrix factorization is carried out. The reported experimental study comprises 196 cases from different tumor types drawn from two international, multi-center databases. The results indicate that the proposed approach outperforms a purely unsupervised process by achieving near perfect correlation of the extracted sources with the mean spectra of the tumor types. It also improves tissue type classification.

Conclusions/Significance

We show that source extraction by unsupervised matrix factorization benefits from the integration of the available class information, so operating in a semi-supervised learning manner, for discriminative source identification and brain tumor labeling from single-voxel spectroscopy data. We are confident that the proposed methodology has wider applicability for biomedical signal processing.

Correlation of diffusion tensor, dynamic susceptibility contrast MRI and (99m)Tc-Tetrofosmin brain SPECT with tumour grade and Ki-67 immunohistochemistry in glioma

OBJECTIVE

Assessment of the grade and type of glioma is of paramount importance for prognosis. Tumour proliferative potentials may provide additional information on the behaviour of the tumour, its response to treatment and prognosis. The purpose of this study was to investigate the correlation between diffusion tensor imaging (DTI), dynamic susceptibility contrast (DSC) magnetic resonance imaging (MRI) and (99m)Tc-Tetrofosmin brain single-photon emission computed tomography (SPECT), and the tumour grade and Ki-67 labelling index in newly diagnosed gliomas.

METHODS

Study was made of patients with suspected glioma on brain MRI between December 2010 and January 2012, by DTI, DSC MRI and (99m)Tc-Tetrofosmin brain SPECT. The proliferative activity of each tumour was measured by deriving the Ki-67 proliferation index from immunohistochemical staining of tumour specimens.

RESULTS

Glioma was newly diagnosed in 25 patients (17 men, 8 women, aged 19-79 years, median 55 years). The Ki-67 index ranged from 1% to 80% (mean 19.4%). On evaluation of the relationship between the (99m)Tc-Tetrofosmin tumour uptake by gliomas was found to be significantly correlated with cellular proliferation (rho=0.924, p<0.0001). Regarding DTI, significant negative correlation was demonstrated between the apparent diffusion coefficient (ADC) ratio and the Ki-67 index (rho=-0.545, p=0.0087). Significant correlation was also observed between the fractional anisotropy (FA) ratio and the Ki-67 index (rho=0.489, p=0.02). Strong correlation was found between relative cerebral blood volume (rCBV) and Ki-67 index (rho=0.853, p<0.0001), and between the (99m)Tc-Tetrofosmin lesion-to-normal (L/N) uptake ratio and rCBV (rho=0.808, p ≤ 0.0001). Significant negative correlation was demonstrated between the (99m)Tc-Tetrofosmin L/N ratio and ADC ratio (rho=-0.513, p=0.014). These imaging techniques were able to distinguish between low-grade and high-grade gliomas.

CONCLUSIONS

Findings on DSC MRI and brain SPECT with (99m)Tc-Tetrofosmin metrics were more closely correlated with glioma cellular proliferation.

The preoperative use of navigated transcranial magnetic stimulation facilitates early resection of suspected low-grade gliomas in the motor cortex

BACKGROUND

Resection is recommended for low-grade gliomas, but often it is not performed if the tumor is suspected of invading the primary motor cortex. The study aim is to assess what influence preoperative navigated transcranial magnetic stimulation (nTMS) has on the treatment strategy and clinical outcome for suspected low-grade gliomas in presumed motor eloquent location.

METHODS

This paper reports on all our patients with gliomas in the primary motor cortex that were non-enhancing on MRI, since we began using nTMS (n = 11). For the comparison group, we identified the 11 most recent such patients just before we started using nTMS.

RESULTS

Exact delineation of motor functional versus non-functional cortical tissue was provided by nTMS in all cases, also within the area of altered FLAIR signal. In 6 out of 11 cases, the nTMS mapping result changed the treatment plan towards early and more extensive resection. Only one nTMS patient had another seizure within the follow-up period, whereas four patients in the comparison group had further seizures. In the nTMS group, 1 of 4 patients with pre-op neurological deficits improved by one year; whereas the comparison group had increased neurological deficits in 3 of the 8 patients not having surgery. The median (range) change of tumor volume from baseline to 1 year was -83 % (-67 % to -100 %) in the nTMS group, but +12 % (+40 % to -56 %) in the comparison group (p < 0.001).

CONCLUSIONS

nTMS provides accurate motor mapping results also in infiltrative gliomas and enables more frequent and more extensive surgical resection of non-enhancing gliomas in or near the primary motor cortex. The substantial differences observed here in neurological and oncological outcomes suggest that further comparative research is warranted.

Implanted carmustine wafers followed by concomitant radiochemotherapy to treat newly diagnosed malignant gliomas: prospective, observational, multicenter study on 92 cases

OBJECTIVES

Study the feasibility and effectiveness of a treatment associated surgery, intraoperative chemotherapy (carmustine wafers), and concomitant radiochemotherapy (temozolomide) for the management of newly diagnosed, high-grade gliomas.

METHODS

Prospective multicenter study conducted in 17 French centers with a total of 92 patients with newly diagnosed malignant glioma treated by surgery, implanted Carmustine wafers (Gliadel(®)) followed by concomitant radiochemotherapy by temozolomide (Temodar(®)). Clinical, imaging, and survival data were collected to study toxicity-induced adverse events and efficacy.

RESULTS

A total of 20.6 % presented with adverse events during surgery, potentially attributable to carmustine, including 5 severe infections. Afterwards, 37.2 % of patients showed adverse events during radiochemotherapy and 40 % during adjuvant chemotherapy by temozolomide. We report a 10.5-month, median, progression-free survival and an 18.8-month median overall survival. No significant statistical difference was observed according to age, Karnofsky Performance Scale, or grade of the tumor. A prognostic difference at the limit of the significance threshold was observed according to the extent of the resection.

CONCLUSIONS

Multimodal treatment associating implanted carmustine chemotherapy and concomitant radiochemotherapy with temozolomide seems to yield better survival rates than those usually described when carmustine or temozolomide are used alone independently from one another. These interesting results were obtained without increased adverse events and would need to be validated during a phase 3 study.

Convex non-negative matrix factorization for brain tumor delimitation from MRSI data

Background

Pattern Recognition techniques can provide invaluable insights in the field of neuro-oncology. This is because the clinical analysis of brain tumors requires the use of non-invasive methods that generate complex data in electronic format. Magnetic Resonance (MR), in the modalities of spectroscopy (MRS) and spectroscopic imaging (MRSI), has been widely applied to this purpose. The heterogeneity of the tissue in the brain volumes analyzed by MR remains a challenge in terms of pathological area delimitation.

Methodology/Principal Findings

A pre-clinical study was carried out using seven brain tumor-bearing mice. Imaging and spectroscopy information was acquired from the brain tissue. A methodology is proposed to extract tissue type-specific sources from these signals by applying Convex Non-negative Matrix Factorization (Convex-NMF). Its suitability for the delimitation of pathological brain area from MRSI is experimentally confirmed by comparing the images obtained with its application to selected target regions, and to the gold standard of registered histopathology data. The former showed good accuracy for the solid tumor region (proliferation index (PI)>30%). The latter yielded (i) high sensitivity and specificity in most cases, (ii) acquisition conditions for safe thresholds in tumor and non-tumor regions (PI>30% for solid tumoral region; ≤5% for non-tumor), and (iii) fairly good results when borderline pixels were considered.

Conclusions/Significance

The unsupervised nature of Convex-NMF, which does not use prior information regarding the tumor area for its delimitation, places this approach one step ahead of classical label-requiring supervised methods for discrimination between tissue types, minimizing the negative effect of using mislabeled voxels. Convex-NMF also relaxes the non-negativity constraints on the observed data, which allows for a natural representation of the MRSI signal. This should help radiologists to accurately tackle one of the main sources of uncertainty in the clinical management of brain tumors, which is the difficulty of appropriately delimiting the pathological area.

Plexin-B1: a potential diagnostic biomarker for glioma and a future target for glioma immunotherapy

Gliomas are the most common tumors in the central nervous system. Plexin-B1 is abundantly expressed in the nervous system as an axonal guidance molecule during neuronal development. However, the correlation between its expression and the clinical characteristics of gliomas, and its therapeutic significance, remain largely unexplored. In this study, we detected the expression of Plexin-B1 in clinical glioma tissue samples. Plexin-B1 was highly expressed in the cytoplasm and on the membrane of glioma tissues, while only trace levels of Plexin-B1 were present in normal brain tissue. The expression level of Plexin-B1 in glioma tissue was associated with the pathological grade of the glioma. In addition, we used flow cytometry to analyze the expression of Plexin-B1 in glioma cell lines and its ligand, semaphorin 4D (Sema4D), in natural killer (NK) cell lines. Cytotoxicity assays showed cytolysis of the U251 glioma cell line by the NK cell line, NK92, and this was markedly downregulated when the neutralizing antibody to Plexin-B1 was added. This study demonstrates that Plexin-B1 could be used as a diagnostic biomarker, and also suggests that it may be involved in the cytotoxicity of NK cells to glioma cells. Plexin-B1 could be a useful future target for glioma immunotherapy.

Gliosis versus glioma?: don't grade until you know

A major challenge in the routine practice of surgical neuropathology is the distinction between reactive astrocytosis, which may be because of non-neoplastic and neoplastic conditions, and a low-grade infiltrating diffuse astrocytoma [World Health Organization (WHO) grade II]. This can be particularly challenging with small biopsies that often yield limited amounts of tissue for pathologic study, especially considering the marked differences in prognosis and therapy after a pathologic diagnosis. This paper will review some basic principles of gliosis as an astrocytic reaction to a wide range of central nervous system insults and focus on some common diagnostic pitfalls such as (1) gliosis associated with brain tumor mimics, including demyelinating disease and infections, (2) gliosis associated with nonglial tumors such as craniopharyngioma, hemangioblastoma, metastases, and central nervous system lymphoma. New diagnostic methods have facilitated the differentiation between reactive astrocytosis and the diffuse gliomas. Of these, the use of mutated isocitrate dehydrogenase-1 (IDH-1) as a marker of diffuse infiltrating astroctomas, oligodendrogliomas, and a subset of glioblastomas (secondary glioblastomas) is particularly exciting for tissue diagnosis and patient prognosis. In addition IDH-1 may be useful to distinguish a diffuse infiltrating glioma from low-grade "focal" neoplasms such as the pilocytic astocytoma in histologically ambiguous cases. The discovery of BRAF mutations as molecular signatures of some pilocytic astrocytomas, gangliogliomas, and pleomorphic xanthoastrocytomas has provided another diagnostic tool for the pathologist. Only after a definitive diagnosis of a diffuse infiltrating glioma or a focal glioma is made should a tumor grade be applied and some practical issues in current glioma grading are provided.

Distinct sulfonation activities in resveratrol-sensitive and resveratrol-insensitive human glioblastoma cells

Glioblastoma multiforme (GBM) cells show different responses to resveratrol, for unknown reasons. Our data from human medulloblastoma cells and primary cultures of rat brain cells revealed an inverse correlation of sulfonation activity with resveratrol sensitivities, providing a clue to the underlying mechanisms of the variable sensitivities of GBM cells to resveratrol. In this study, we found that U251 cells were sensitive and LN229 cells were insensitive to resveratrol. Thus, these two cell lines were taken as comparable models for elucidating the influence of sulfonation activities on resveratrol sensitivity. HPLC showed identical resveratrol metabolic patterns in both cell lines. LC/MS and high-resolution mass MS analyses further demonstrated that resveratrol monosulfate generated by sulfotransferases (SULTs) was the major metabolite of human GBM cells. The levels of brain-associated SULT (SULT1A1, SULT1C2, and SULT4A1) expression in U251 cells were lower than those in LN229 cells, suggesting the inverse relationship of SULT-mediated sulfonation activity with high intracellular resveratrol bioavailability and resveratrol sensitivity of human GBM cells. Furthermore, immunohistochemical staining revealed reductions in expression of the three brain-associated SULTs in 72.8%, 47.5% and 66.3% of astrocytomas, respectively. Therefore, the levels of brain-associated SULTs and sulfonation activity mediated by them could be important parameters for evaluating the potential response of human GBM cells to resveratrol, and may have value in the personalized treatment of GBMs with resveratrol.

Neuropathological diagnosis of brain tumours

With recent progress in radiological, pathological, immunohistochemical, molecular and genetic diagnoses, the characterisation of brain tumours has improved. The last World Health Organization (WHO) Classification of Tumours of the Central Nervous System was done in 2007, based on morphological features, growth pattern and molecular profile of neoplastic cells, defined malignancy grade. The neuropathological diagnosis and the grading of each histotype are based on identification of histopathological criteria and immunohistochemical data. Molecular and genetic profiles may identify different tumour subtypes varying in biological and clinical behaviour, indicating prognostic and predictive factors. In order to investigate new therapeutic approaches, it is important to study the molecular pathways responsible for proliferation, invasion, angiogenesis, and anaplastic transformation. Different prognostic and predictive factors for glioma patients were identified by genetic studies, such as the loss of heterozygosis on chromosome 1p and 19q for oligodendrogliomas, proangiogenic factors such as Vascular Endothelial Growth Factor for glioblastomas and the methylation status of gene promoter of MethylGuanine-MethylTransferase. In conclusion, the prognostic evaluation and the therapeutic strategies for patients depend on the synthesis of histological diagnosis, malignancy grade, gene-molecular profile, radiological images, surgical resection and clinical findings (age, tumour location, and "performance status").

Non-negative matrix factorisation methods for the spectral decomposition of MRS data from human brain tumours

BACKGROUND

In-vivo single voxel proton magnetic resonance spectroscopy (SV 1H-MRS), coupled with supervised pattern recognition (PR) methods, has been widely used in clinical studies of discrimination of brain tumour types and follow-up of patients bearing abnormal brain masses. SV 1H-MRS provides useful biochemical information about the metabolic state of tumours and can be performed at short (< 45 ms) or long (> 45 ms) echo time (TE), each with particular advantages. Short-TE spectra are more adequate for detecting lipids, while the long-TE provides a much flatter signal baseline in between peaks but also negative signals for metabolites such as lactate. Both, lipids and lactate, are respectively indicative of specific metabolic processes taking place. Ideally, the information provided by both TE should be of use for clinical purposes. In this study, we characterise the performance of a range of Non-negative Matrix Factorisation (NMF) methods in two respects: first, to derive sources correlated with the mean spectra of known tissue types (tumours and normal tissue); second, taking the best performing NMF method for source separation, we compare its accuracy for class assignment when using the mixing matrix directly as a basis for classification, as against using the method for dimensionality reduction (DR). For this, we used SV 1H-MRS data with positive and negative peaks, from a widely tested SV 1H-MRS human brain tumour database.

RESULTS

The results reported in this paper reveal the advantage of using a recently described variant of NMF, namely Convex-NMF, as an unsupervised method of source extraction from SV1H-MRS. Most of the sources extracted in our experiments closely correspond to the mean spectra of some of the analysed tumour types. This similarity allows accurate diagnostic predictions to be made both in fully unsupervised mode and using Convex-NMF as a DR step previous to standard supervised classification. The obtained results are comparable to, or more accurate than those obtained with supervised techniques.

CONCLUSIONS

The unsupervised properties of Convex-NMF place this approach one step ahead of classical label-requiring supervised methods for the discrimination of brain tumour types, as it accounts for their increasingly recognised molecular subtype heterogeneity. The application of Convex-NMF in computer assisted decision support systems is expected to facilitate further improvements in the uptake of MRS-derived information by clinicians.