Effects of hydroxyurea on brain function in children with sickle cell anemia.

INTRODUCTION: Sickle cell anemia (SCA) results in numerous adverse effects on the brain, including neurocognitive dysfunction. Hydroxyurea has been utilized extensively for management of SCA, but its effects on brain function have not been established. METHODS: We examined prospectively the effects of 1 year of treatment with hydroxyurea on brain function in children with SCA (HbSS/HbSß0 -thalassemia) by baseline and exit evaluations, including comprehensive neurocognitive testing, transcranial Doppler ultrasound (TCD), and brain MRI (silent cerebral infarcts [SCI], gray matter cerebral blood flow [GM-CBF], and blood oxygen level-dependent [BOLD] signal from visual stimulation). RESULTS: Nineteen patients with SCA, mean age 12.4 years (range 7.2-17.8), were evaluated. At baseline, subjects had these mean values: full-scale IQ (FSIQ) 82.8, TCD velocity 133 cm/s, GM-CBF 64.4 ml/100 g/min, BOLD signal 2.34% increase, and frequency of SCI 47%. After 1 year of hydroxyurea, there were increases in FSIQ (+2, p = .059) and reading passage comprehension (+4, p = .033), a significant decrease in TCD velocity (-11 cm/s, p = .007), and no significant changes in GM-CBF, BOLD, or SCI frequency. Hemoglobin F (HbF) was associated with passage comprehension, hemoglobin with lower TCD velocity, and lower GM-CBF with greater working memory. Higher BOLD signal was associated with higher processing speed and lower TCD velocity with higher math fluency. DISCUSSION: Improvements in neurocognition and decreased TCD velocity following 1 year of treatment support hydroxyurea use for improving neurocognitive outcomes in SCA. Understanding the mechanisms of benefit, as indicated by relationships of neurocognitive function with HbF, hemoglobin, and CBF, requires further evaluation.

Successive distinct high-grade gliomas in L-2-hydroxyglutaric aciduria.

Patients with L-2-hydroxyglutaric aciduria are at risk for developing cerebral neoplasms, particularly gliomas, as one of the optical isomers of the known oncometabolite, 2-hydroxyglutarate is produced in L-2-hydroxyglutaric aciduria. To illustrate the concept of sustained oncogenic potential in permanent exposure to L-2-hydroxyglutarate in brain tissue, we present the medical history of a patient with L-2-hydroxyglutaric aciduria who underwent surgery to remove a right temporal anaplastic astrocytoma and developed an anatomically distinct, but histopathologically similar, tumor in the left frontal region 40 months later. This is the first reported case of successive distinct gliomas in a patient with L-2-hydroxyglutaric aciduria. While this implies a significant, cumulative lifetime risk for cerebral neoplasms in patients with this rare organic aciduria, it also allows further insight into a unique mechanism of tumorigenesis in the brain.

Addition of MR imaging features and genetic biomarkers strengthens glioblastoma survival prediction in TCGA patients.

PURPOSE: The purpose of our study was to assess whether a model combining clinical factors, MR imaging features, and genomics would better predict overall survival of patients with glioblastoma (GBM) than either individual data type. METHODS: The study was conducted leveraging The Cancer Genome Atlas (TCGA) effort supported by the National Institutes of Health. Six neuroradiologists reviewed MRI images from The Cancer Imaging Archive (http://cancerimagingarchive.net) of 102 GBM patients using the VASARI scoring system. The patients' clinical and genetic data were obtained from the TCGA website (http://www.cancergenome.nih.gov/). Patient outcome was measured in terms of overall survival time. The association between different categories of biomarkers and survival was evaluated using Cox analysis. RESULTS: The features that were significantly associated with survival were: (1) clinical factors: chemotherapy; (2) imaging: proportion of tumor contrast enhancement on MRI; and (3) genomics: HRAS copy number variation. The combination of these three biomarkers resulted in an incremental increase in the strength of prediction of survival, with the model that included clinical, imaging, and genetic variables having the highest predictive accuracy (area under the curve 0.679±0.068, Akaike's information criterion 566.7, P<0.001). CONCLUSION: A combination of clinical factors, imaging features, and HRAS copy number variation best predicts survival of patients with GBM.

Outcome prediction in patients with glioblastoma by using imaging, clinical, and genomic biomarkers: focus on the nonenhancing component of the tumor.

PURPOSE: To correlate patient survival with morphologic imaging features and hemodynamic parameters obtained from the nonenhancing region (NER) of glioblastoma (GBM), along with clinical and genomic markers. MATERIALS AND METHODS: An institutional review board waiver was obtained for this HIPAA-compliant retrospective study. Forty-five patients with GBM underwent baseline imaging with contrast material-enhanced magnetic resonance (MR) imaging and dynamic susceptibility contrast-enhanced T2*-weighted perfusion MR imaging. Molecular and clinical predictors of survival were obtained. Single and multivariable models of overall survival (OS) and progression-free survival (PFS) were explored with Kaplan-Meier estimates, Cox regression, and random survival forests. RESULTS: Worsening OS (log-rank test, P = .0103) and PFS (log-rank test, P = .0223) were associated with increasing relative cerebral blood volume of NER (rCBVNER), which was higher with deep white matter involvement (t test, P = .0482) and poor NER margin definition (t test, P = .0147). NER crossing the midline was the only morphologic feature of NER associated with poor survival (log-rank test, P = .0125). Preoperative Karnofsky performance score (KPS) and resection extent (n = 30) were clinically significant OS predictors (log-rank test, P = .0176 and P = .0038, respectively). No genomic alterations were associated with survival, except patients with high rCBVNER and wild-type epidermal growth factor receptor (EGFR) mutation had significantly poor survival (log-rank test, P = .0306; area under the receiver operating characteristic curve = 0.62). Combining resection extent with rCBVNER marginally improved prognostic ability (permutation, P = .084). Random forest models of presurgical predictors indicated rCBVNER as the top predictor; also important were KPS, age at diagnosis, and NER crossing the midline. A multivariable model containing rCBVNER, age at diagnosis, and KPS can be used to group patients with more than 1 year of difference in observed median survival (0.49-1.79 years). CONCLUSION: Patients with high rCBVNER and NER crossing the midline and those with high rCBVNER and wild-type EGFR mutation showed poor survival. In multivariable survival models, however, rCBVNER provided unique prognostic information that went above and beyond the assessment of all NER imaging features, as well as clinical and genomic features.

Genomic mapping and survival prediction in glioblastoma: molecular subclassification strengthened by hemodynamic imaging biomarkers.

PURPOSE: To correlate tumor blood volume, measured by using dynamic susceptibility contrast material-enhanced T2*-weighted magnetic resonance (MR) perfusion studies, with patient survival and determine its association with molecular subclasses of glioblastoma (GBM). MATERIALS AND METHODS: This HIPAA-compliant retrospective study was approved by institutional review board. Fifty patients underwent dynamic susceptibility contrast-enhanced T2*-weighted MR perfusion studies and had gene expression data available from the Cancer Genome Atlas. Relative cerebral blood volume (rCBV) (maximum rCBV [rCBV(max)] and mean rCBV [rCBV(mean)]) of the contrast-enhanced lesion as well as rCBV of the nonenhanced lesion (rCBV(NEL)) were measured. Patients were subclassified according to the Verhaak and Phillips classification schemas, which are based on similarity to defined genomic expression signature. We correlated rCBV measures with the molecular subclasses as well as with patient overall survival by using Cox regression analysis. RESULTS: No statistically significant differences were noted for rCBV(max), rCBV(mean) of contrast-enhanced lesion or rCBV(NEL) between the four Verhaak classes or the three Phillips classes. However, increased rCBV measures are associated with poor overall survival in GBM. The rCBV(max) (P = .0131) is the strongest predictor of overall survival regardless of potential confounders or molecular classification. Interestingly, including the Verhaak molecular GBM classification in the survival model clarifies the association of rCBV(mean) with patient overall survival (hazard ratio: 1.46, P = .0212) compared with rCBV(mean) alone (hazard ratio: 1.25, P = .1918). Phillips subclasses are not predictive of overall survival nor do they affect the predictive ability of rCBV measures on overall survival. CONCLUSION: The rCBV(max) measurements could be used to predict patient overall survival independent of the molecular subclasses of GBM; however, Verhaak classifiers provided additional information, suggesting that molecular markers could be used in combination with hemodynamic imaging biomarkers in the future.

MR imaging predictors of molecular profile and survival: multi-institutional study of the TCGA glioblastoma data set.

PURPOSE: To conduct a comprehensive analysis of radiologist-made assessments of glioblastoma (GBM) tumor size and composition by using a community-developed controlled terminology of magnetic resonance (MR) imaging visual features as they relate to genetic alterations, gene expression class, and patient survival. MATERIALS AND METHODS: Because all study patients had been previously deidentified by the Cancer Genome Atlas (TCGA), a publicly available data set that contains no linkage to patient identifiers and that is HIPAA compliant, no institutional review board approval was required. Presurgical MR images of 75 patients with GBM with genetic data in the TCGA portal were rated by three neuroradiologists for size, location, and tumor morphology by using a standardized feature set. Interrater agreements were analyzed by using the Krippendorff α statistic and intraclass correlation coefficient. Associations between survival, tumor size, and morphology were determined by using multivariate Cox regression models; associations between imaging features and genomics were studied by using the Fisher exact test. RESULTS: Interrater analysis showed significant agreement in terms of contrast material enhancement, nonenhancement, necrosis, edema, and size variables. Contrast-enhanced tumor volume and longest axis length of tumor were strongly associated with poor survival (respectively, hazard ratio: 8.84, P = .0253, and hazard ratio: 1.02, P = .00973), even after adjusting for Karnofsky performance score (P = .0208). Proneural class GBM had significantly lower levels of contrast enhancement (P = .02) than other subtypes, while mesenchymal GBM showed lower levels of nonenhanced tumor (P < .01). CONCLUSION: This analysis demonstrates a method for consistent image feature annotation capable of reproducibly characterizing brain tumors; this study shows that radiologists' estimations of macroscopic imaging features can be combined with genetic alterations and gene expression subtypes to provide deeper insight to the underlying biologic properties of GBM subsets.

Early prediction of response to Vorinostat in an orthotopic rat glioma model.

Glioblastoma is the most common primary brain tumor and is uniformly fatal despite aggressive surgical and adjuvant therapy. As survival is short, it is critical to determine the value of therapy early on in treatment. Improved early predictive assessment would allow neuro-oncologists to personalize and adjust or change treatment sooner to maximize the use of efficacious therapy. During carcinogenesis, tumor suppressor genes can be silenced by aberrant histone deacetylation. This epigenetic modification has become an important target for tumor therapy. Suberoylanilide hydroxamic acid (SAHA, Vorinostat, Zolinza) is an orally active, potent inhibitor of histone deacetylase (HDAC) activity. A major shortcoming of the use of HDAC inhibitors in the treatment of patients with brain tumors is the lack of reliable biomarkers to predict and determine response. Histological evaluation may reflect tumor viability following treatment, but is an invasive procedure and impractical for glioblastoma. Another problem is that response to SAHA therapy is associated with tumor redifferentiation and cytostasis rather than tumor size reduction, thus limiting the use of traditional imaging methods. A noninvasive method to assess drug delivery and efficacy is needed. Here, we investigated whether changes in (1)H MRS metabolites could render reliable biomarkers for an early response to SAHA treatment in an orthotopic animal model for glioma. Untreated tumors exhibited significantly elevated alanine and lactate levels and reduced inositol, N-acetylaspartate and creatine levels, typical changes reported in glioblastoma relative to normal brain tissues. The (1)H MRS-detectable metabolites of SAHA-treated tumors were restored to those of normal-like brain tissues. In addition, reduced inositol and N-acetylaspartate were found to be potential biomarkers for mood alteration and depression, which may also be alleviated with SAHA treatment. Our study suggests that (1)H MRS can provide reliable metabolic biomarkers at the earliest stage of SAHA treatment to predict the therapeutic response.

The use of neuroimaging to guide the histologic diagnosis of central nervous system lesions.

Recent advances in neuroimaging techniques, particularly in magnetic resonance imaging, have led to substantially improved spatial anatomic resolution such that subtle or small central nervous system lesions, which could go undetected on gross examination of brain sections, are now readily identified on imaging. Although neuroimaging is generally considered the surrogate of gross neuropathology, it is still not a substitute for tissue diagnosis. Rather, it can be a valuable tool for the surgical pathologist in the process of formulating a differential diagnosis based on location and imaging features, as well as in identifying radiologic/pathologic discordance, such as the possible undersampling of a heterogenous glioma, which could lead to underestimation of the tumor grade. The following review focuses on the application of neuroimaging techniques, mainly magnetic resonance imaging, to the histologic diagnosis of central nervous system lesions, and the correlation of imaging features of infiltrative gliomas with histologic findings pertinent to tumor grading. The use of advanced functional magnetic resonance methods, specifically diffusion-weighted imaging, perfusion-weighted imaging, and magnetic resonance spectroscopy is also discussed, as well as the common pitfalls in imaging interpretation.

Assessment of Fat Fractions in the Tongue, Soft Palate, Pharyngeal Wall, and Parapharyngeal Fat Pad by the GOOSE and DIXON Methods.

OBJECTIVE: The 2-point DIXON method is widely used to assess fat fractions (FFs) in magnetic resonance images (MRIs) of the tongue, pharyngeal wall, and surrounding tissues in patients with obstructive sleep apnea (OSA). However, the method is semiquantitative and is susceptible to B0 field inhomogeneities and R2* confounding factors. Using the method, although several studies have shown that patients with OSA have increased fat deposition around the pharyngeal cavity, conflicting findings was also reported in 1 study. This discrepancy necessitates that we examine the FF estimation method used in the earlier studies and seek a more accurate method to measure FFs. MATERIALS AND METHODS: We examined the advantages of using the GOOSE (globally optimal surface estimation) method to replace the 2-point DIXON method for quantifying fat in the tongue and surrounding tissues on MRIs. We first used phantoms with known FFs (true FFs) to validate the GOOSE method and examine the errors in the DIXON method. Then, we compared the 2 methods in the tongue, soft palate, pharyngeal wall, and parapharyngeal fat pad of 63 healthy participants to further assess the errors caused by the DIXON method. Six participants were excluded from the comparison of the tongue FFs because of technical failures. Paired Student t tests were performed on FFs to detect significant differences between the 2 methods. All measures were obtained using 3 T Siemens MRI scanners. RESULTS: In the phantoms, the FFs measured by GOOSE agreed with the true FF, with only a 1.2% mean absolute error. However, the same measure by DIXON had a 10.5% mean absolute error. The FFs obtained by DIXON were significantly lower than those obtained by GOOSE (P < 0.0001) in the human participants. We found strong correlations between GOOSE and DIXON in the tongue (R2 = 0.90), soft palate (R2 = 0.66), and parapharyngeal fat pad (R2 = 0.88), but the correlation was weaker in the posterior pharyngeal walls (R2 = 0.32) in participants. CONCLUSIONS: The widely used 2-point DIXON underestimated FFs, relative to GOOSE, in phantom measurements and tissues studied in vivo. Thus, an advanced method, such as GOOSE, that uses multiecho complex data is preferred for estimating FF.

11C-Methionine PET for Identification of Pediatric High-Grade Glioma Recurrence.

Differentiating tumor recurrence or progression from pseudoprogression during surveillance of pediatric high-grade gliomas (PHGGs) using MRI, the primary imaging modality for evaluation of brain tumors, can be challenging. The aim of this study was to evaluate whether 11C-methionine PET, a molecular imaging technique that detects functionally active tumors, is useful for further evaluating MRI changes concerning for tumor recurrence during routine surveillance. Methods: Using 11C-methionine PET during follow-up visits, we evaluated 27 lesions in 26 patients with new or worsening MRI abnormalities for whom tumor recurrence was of concern. We performed quantitative and qualitative assessments of both 11C-methionine PET and MRI data to predict the presence of tumor recurrence. Further, to assess for an association with overall survival (OS), we plotted the time from development of the imaging changes against survival. Results: Qualitative evaluation of 11C-methionine PET achieved 100% sensitivity, 60% specificity, and 93% accuracy to correctly predict the presence of tumors in 27 new or worsening MRI abnormalities. Qualitative MRI evaluation achieved sensitivity ranging from 86% to 95%, specificity ranging from 40% to 60%, and accuracy ranging from 85% to 89%. The interobserver agreement for 11C-methionine PET assessment was 100%, whereas the interobserver agreement was only 50% for MRI (P < 0.01). Quantitative MRI and 11C-methionine PET evaluation using receiver-operating characteristics demonstrated higher specificity (80%) than did qualitative evaluations (40%-60%). Postcontrast enhancement volume, metabolic tumor volume, tumor-to-brain ratio, and presence of tumor as determined by consensus MRI assessment were inversely associated with OS. Conclusion:11C-methionine PET has slightly higher sensitivity and accuracy for correctly predicting tumor recurrence, with excellent interobserver agreement, than does MRI. Quantitative 11C-methionine PET can also predict OS. These findings suggest that 11C-methionine PET can be useful for further evaluation of MRI changes during surveillance of previously treated PHGGs.

Progression of central nervous system disease from pediatric to young adulthood in sickle cell anemia.

Silent cerebral infarcts and arteriopathy are common and progressive in individuals with sickle cell anemia. However, most data describing brain lesions in sickle cell anemia are cross-sectional or derive from pediatric cohorts with short follow-up. We investigated the progression of silent cerebral infarct and cerebral vessel stenosis on brain MRI and MRA, respectively, by describing the incidence of new or worsening lesions over a period of up to 25 years among young adults with sickle cell anemia and explored risk factors for progression. Forty-four adults with sickle cell anemia (HbSS or HbSß0thalassemia), exposed to chronic transfusions (n = 12) or hydroxyurea (n = 32), median age 19.2 years (range 18.0-31.5), received a screening brain MRI/MRA and their results were compared with a clinical exam performed during childhood and adolescence. We used exact log-rank test to compare MRI and MRA progression among any two groups. The hazard ratio (HR) and 95% confidence interval (CI) were calculated from Cox regression analyses. Progression of MRI and MRA occurred in 12 (27%) and 4 (9%) young adults, respectively, relative to their pediatric exams. MRI progression risk was high among participants with abnormal pediatric exams (HR: 11.6, 95% CI: 2.5-54.7) and conditional or abnormal transcranial Doppler ultrasound velocities (HR: 3.9, 95% CI: 1.0-15.1). Among individuals treated with hydroxyurea, high fetal hemoglobin measured in childhood was associated with lower hazard of MRI progression (HR: 0.86, 95% CI: 0.76-0.98). MRA progression occurred more frequently among those with prior stroke (HR: 8.6, 95% CI: 1.2-64), abnormal pediatric exam (P = 0.00084), and elevated transcranial Doppler ultrasound velocities (P = 0.004). Brain MRI/MRA imaging in pediatrics can identify high-risk patients for CNS disease progression in young adulthood, prompting consideration for early aggressive treatments.

Bithalamic gliomas may be molecularly distinct from their unilateral high-grade counterparts.

Bithalamic gliomas are rare cancers diagnosed based on poorly defined radiologic criteria. Infiltrative astrocytomas account for most cases. While some previous studies reported dismal outcomes for patients with bithalamic gliomas irrespective of therapy and histologic grade, others described better prognoses even without anticancer therapy. Little is known about their molecular characteristics. We reviewed clinical, radiologic, and histologic features of patients with bithalamic gliomas treated at our institution over 15 years. Targeted sequencing of mutational hotspots in H3F3A, HIST1H3B, IDH1/2, and BRAF, and genome-wide analysis of DNA methylation and copy number abnormalities was performed in available tumors. Eleven patients with bithalamic gliomas were identified. Their median age at diagnosis was 4.8 years (range: 1-15.7). Additional involvement of the brainstem, basal ganglia, and cerebral lobes occurred in 11, 9, and 3 cases, respectively. All patients presented with hydrocephalus. Two-thirds of the patients had a histologic diagnosis of anaplastic astrocytoma. Despite aggressive therapy, our youngest patient, the only one diagnosed before 1 year of age, is the sole long-term survivor. DNA methylation could be performed in seven tumors, all of which clustered with the RTK I 'PDGFRA' subgroup by unsupervised hierarchical analysis of methylation array against a previously published cohort of 59 pediatric high-grade gliomas. Sequencing of hotspots mutations could be done in 10 tumors, none of which harbored H3F3A p.K27 and/or the respective DNA methylation signature, and any other hotspot mutations. Amplification of MDM4 (n = 2), PDGFRA (n = 2), and ID2 combined with MYCN (n = 1) were observed in 7 tumors available for analysis. In comparison with the previously published experience with unilateral high-grade thalamic astrocytomas where H3F3A p.K27 was present in two-thirds of cases, the absence of this molecular subgroup in bithalamic gliomas was striking. This finding suggests that unilateral and bithalamic high-grade gliomas may represent two distinct molecular entities.

Posttreatment DSC-MRI is Predictive of Early Treatment Failure in Children with Supratentorial High-Grade Glioma Treated with Erlotinib.

BACKGROUND AND PURPOSE: The role of perfusion imaging in the management of pediatric high grade glioma is unclear. We evaluated the ability of dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) to determine grade, evaluate post-treatment response and predict treatment failure. MATERIAL AND METHODS: In this study 22 patients with high-grade glioma underwent biopsy and were treated with concurrent and sequential radiotherapy and erlotinib as part of a phase I/II clinical trial (NCT00124657). Preradiotherapy, immediate postradiotherapy, 6­month and treatment failure DSC MR images were reviewed, registered, and processed for the ratio of cerebral blood flow (CBF) and cerebral blood volume (CBV). Processed, derived perfusion, and T1-weighted images (T1WI), T2WI, and fluid attenuation inversion recovery (FLAIR) MRI sequences were used for segmentation and extraction of tumor perfusion parameters at all time points. Patient, tumor, treatment, and outcome data were summarized and related to perfusion data. RESULTS: Regional CBF in tumors increased from diagnosis to postradiotherapy, while they decreased to levels below those at diagnosis from postradiotherapy to 6­month follow-up. At 6 months, the median regional CBF was higher in tumors that progressed (median 1.16) than in those that did not (median, 0.95; P < 0.05). Patients with regional CBF ratios above 1.4 at diagnosis had shorter survival times than did those with regional CBF ratios below 1.4 (P = 0.77). Tumors with a regional CBV above 1.15 at the postradiotherapy (1-3 months) follow-up scan were associated with an earlier time to death than that of tumors with a regional CBV below 1.15 (P < 0.05). CONCLUSION: Posttreatment perfusion characteristics are prognostic and may help predict survival. Overall, perfusion MRI is useful for managing pediatric high-grade glioma and should be incorporated into future clinical trials.

Multiple-response regression analysis links magnetic resonance imaging features to de-regulated protein expression and pathway activity in lower grade glioma.

BACKGROUND AND PURPOSE: Lower grade gliomas (LGGs), lesions of WHO grades II and III, comprise 10-15% of primary brain tumors. In this first-of-a-kind study, we aim to carry out a radioproteomic characterization of LGGs using proteomics data from the TCGA and imaging data from the TCIA cohorts, to obtain an association between tumor MRI characteristics and protein measurements. The availability of linked imaging and molecular data permits the assessment of relationships between tumor genomic/proteomic measurements with phenotypic features. MATERIALS AND METHODS: Multiple-response regression of the image-derived, radiologist scored features with reverse-phase protein array (RPPA) expression levels generated correlation coefficients for each combination of image-feature and protein or phospho-protein in the RPPA dataset. Significantly-associated proteins for VASARI features were analyzed with Ingenuity Pathway Analysis software. Hierarchical clustering of the results of the pathway analysis was used to determine which feature groups were most strongly correlated with pathway activity and cellular functions. RESULTS: The multiple-response regression approach identified multiple proteins associated with each VASARI imaging feature. VASARI features were found to be correlated with expression of IL8, PTEN, PI3K/Akt, Neuregulin, ERK/MAPK, p70S6K and EGF signaling pathways. CONCLUSION: Radioproteomics analysis might enable an insight into the phenotypic consequences of molecular aberrations in LGGs.

A combinatorial radiographic phenotype may stratify patient survival and be associated with invasion and proliferation characteristics in glioblastoma.

OBJECTIVE: Individual MRI characteristics (e.g., volume) are routinely used to identify survival-associated phenotypes for glioblastoma (GBM). This study investigated whether combinations of MRI features can also stratify survival. Furthermore, the molecular differences between phenotype-induced groups were investigated. METHODS: Ninety-two patients with imaging, molecular, and survival data from the TCGA (The Cancer Genome Atlas)-GBM collection were included in this study. For combinatorial phenotype analysis, hierarchical clustering was used. Groups were defined based on a cutpoint obtained via tree-based partitioning. Furthermore, differential expression analysis of microRNA (miRNA) and mRNA expression data was performed using GenePattern Suite. Functional analysis of the resulting genes and miRNAs was performed using Ingenuity Pathway Analysis. Pathway analysis was performed using Gene Set Enrichment Analysis. RESULTS: Clustering analysis reveals that image-based grouping of the patients is driven by 3 features: volume-class, hemorrhage, and T1/FLAIR-envelope ratio. A combination of these features stratifies survival in a statistically significant manner. A cutpoint analysis yields a significant survival difference in the training set (median survival difference: 12 months, p = 0.004) as well as a validation set (p = 0.0001). Specifically, a low value for any of these 3 features indicates favorable survival characteristics. Differential expression analysis between cutpoint-induced groups suggests that several immune-associated (natural killer cell activity, T-cell lymphocyte differentiation) and metabolism-associated (mitochondrial activity, oxidative phosphorylation) pathways underlie the transition of this phenotype. Integrating data for mRNA and miRNA suggests the roles of several genes regulating proliferation and invasion. CONCLUSIONS: A 3-way combination of MRI phenotypes may be capable of stratifying survival in GBM. Examination of molecular processes associated with groups created by this combinatorial phenotype suggests the role of biological processes associated with growth and invasion characteristics.