Radiomics Can Distinguish Pediatric Supratentorial Embryonal Tumors, High-Grade Gliomas, and Ependymomas.

BACKGROUND AND PURPOSE: Pediatric supratentorial tumors such as embryonal tumors, high-grade gliomas, and ependymomas are difficult to distinguish by histopathology and imaging because of overlapping features. We applied machine learning to uncover MR imaging-based radiomics phenotypes that can differentiate these tumor types. MATERIALS AND METHODS: Our retrospective cohort of 231 patients from 7 participating institutions had 50 embryonal tumors, 127 high-grade gliomas, and 54 ependymomas. For each tumor volume, we extracted 900 Image Biomarker Standardization Initiative-based PyRadiomics features from T2-weighted and gadolinium-enhanced T1-weighted images. A reduced feature set was obtained by sparse regression analysis and was used as input for 6 candidate classifier models. Training and test sets were randomly allocated from the total cohort in a 75:25 ratio. RESULTS: The final classifier model for embryonal tumor-versus-high-grade gliomas identified 23 features with an area under the curve of 0.98; the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 0.85, 0.91, 0.79, 0.94, and 0.89, respectively. The classifier for embryonal tumor-versus-ependymomas identified 4 features with an area under the curve of 0.82; the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 0.93, 0.69, 0.76, 0.90, and 0.81, respectively. The classifier for high-grade gliomas-versus-ependymomas identified 35 features with an area under the curve of 0.96; the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 0.82, 0.94, 0.82, 0.94, and 0.91, respectively. CONCLUSIONS: In this multi-institutional study, we identified distinct radiomic phenotypes that distinguish pediatric supratentorial tumors, high-grade gliomas, and ependymomas with high accuracy. Incorporation of this technique in diagnostic algorithms can improve diagnosis, risk stratification, and treatment planning.

Radiomic Phenotypes Distinguish Atypical Teratoid/Rhabdoid Tumors from Medulloblastoma.

BACKGROUND AND PURPOSE: Atypical teratoid/rhabdoid tumors and medulloblastomas have similar imaging and histologic features but distinctly different outcomes. We hypothesized that they could be distinguished by MR imaging-based radiomic phenotypes. MATERIALS AND METHODS: We retrospectively assembled T2-weighted and gadolinium-enhanced T1-weighted images of 48 posterior fossa atypical teratoid/rhabdoid tumors and 96 match-paired medulloblastomas from 7 institutions. Using a holdout test set, we measured the performance of 6 candidate classifier models using 6 imaging features derived by sparse regression of 900 T2WI and 900 T1WI Imaging Biomarker Standardization Initiative-based radiomics features. RESULTS: From the originally extracted 1800 total Imaging Biomarker Standardization Initiative-based features, sparse regression consistently reduced the feature set to 1 from T1WI and 5 from T2WI. Among classifier models, logistic regression performed with the highest AUC of 0.86, with sensitivity, specificity, accuracy, and F1 scores of 0.80, 0.82, 0.81, and 0.85, respectively. The top 3 important Imaging Biomarker Standardization Initiative features, by decreasing order of relative contribution, included voxel intensity at the 90th percentile, inverse difference moment normalized, and kurtosis-all from T2WI. CONCLUSIONS: Six quantitative signatures of image intensity, texture, and morphology distinguish atypical teratoid/rhabdoid tumors from medulloblastomas with high prediction performance across different machine learning strategies. Use of this technique for preoperative diagnosis of atypical teratoid/rhabdoid tumors could significantly inform therapeutic strategies and patient care discussions.

Deep Learning for Pediatric Posterior Fossa Tumor Detection and Classification: A Multi-Institutional Study.

BACKGROUND AND PURPOSE: Posterior fossa tumors are the most common pediatric brain tumors. MR imaging is key to tumor detection, diagnosis, and therapy guidance. We sought to develop an MR imaging-based deep learning model for posterior fossa tumor detection and tumor pathology classification. MATERIALS AND METHODS: The study cohort comprised 617 children (median age, 92 months; 56% males) from 5 pediatric institutions with posterior fossa tumors: diffuse midline glioma of the pons (n = 122), medulloblastoma (n = 272), pilocytic astrocytoma (n = 135), and ependymoma (n = 88). There were 199 controls. Tumor histology served as ground truth except for diffuse midline glioma of the pons, which was primarily diagnosed by MR imaging. A modified ResNeXt-50-32x4d architecture served as the backbone for a multitask classifier model, using T2-weighted MRIs as input to detect the presence of tumor and predict tumor class. Deep learning model performance was compared against that of 4 radiologists. RESULTS: Model tumor detection accuracy exceeded an AUROC of 0.99 and was similar to that of 4 radiologists. Model tumor classification accuracy was 92% with an F1 score of 0.80. The model was most accurate at predicting diffuse midline glioma of the pons, followed by pilocytic astrocytoma and medulloblastoma. Ependymoma prediction was the least accurate. Tumor type classification accuracy and F1 score were higher than those of 2 of the 4 radiologists. CONCLUSIONS: We present a multi-institutional deep learning model for pediatric posterior fossa tumor detection and classification with the potential to augment and improve the accuracy of radiologic diagnosis.

MR Imaging-Based Radiomic Signatures of Distinct Molecular Subgroups of Medulloblastoma.

BACKGROUND AND PURPOSE: Distinct molecular subgroups of pediatric medulloblastoma confer important differences in prognosis and therapy. Currently, tissue sampling is the only method to obtain information for classification. Our goal was to develop and validate radiomic and machine learning approaches for predicting molecular subgroups of pediatric medulloblastoma. MATERIALS AND METHODS: In this multi-institutional retrospective study, we evaluated MR imaging datasets of 109 pediatric patients with medulloblastoma from 3 children's hospitals from January 2001 to January 2014. A computational framework was developed to extract MR imaging-based radiomic features from tumor segmentations, and we tested 2 predictive models: a double 10-fold cross-validation using a combined dataset consisting of all 3 patient cohorts and a 3-dataset cross-validation, in which training was performed on 2 cohorts and testing was performed on the third independent cohort. We used the Wilcoxon rank sum test for feature selection with assessment of area under the receiver operating characteristic curve to evaluate model performance. RESULTS: Of 590 MR imaging-derived radiomic features, including intensity-based histograms, tumor edge-sharpness, Gabor features, and local area integral invariant features, extracted from imaging-derived tumor segmentations, tumor edge-sharpness was most useful for predicting sonic hedgehog and group 4 tumors. Receiver operating characteristic analysis revealed superior performance of the double 10-fold cross-validation model for predicting sonic hedgehog, group 3, and group 4 tumors when using combined T1- and T2-weighted images (area under the curve = 0.79, 0.70, and 0.83, respectively). With the independent 3-dataset cross-validation strategy, select radiomic features were predictive of sonic hedgehog (area under the curve = 0.70-0.73) and group 4 (area under the curve = 0.76-0.80) medulloblastoma. CONCLUSIONS: This study provides proof-of-concept results for the application of radiomic and machine learning approaches to a multi-institutional dataset for the prediction of medulloblastoma subgroups.

Casein kinase 2α regulates glioblastoma brain tumor-initiating cell growth through the ß-catenin pathway.

Glioblastoma (GBM) is the most common and fatal primary brain tumor in humans, and it is essential that new and better therapies are developed to treat this disease. Previous research suggests that casein kinase 2 (CK2) may be a promising therapeutic target for GBMs. CK2 has enhanced expression or activity in numerous cancers, including GBM, and it has been demonstrated that inhibitors of CK2 regressed tumor growth in GBM xenograft mouse models. Our studies demonstrate that the CK2 subunit, CK2α, is overexpressed in and has an important role in regulating brain tumor-initiating cells (BTIC) in GBM. Initial studies showed that two GBM cell lines (U87-MG and U138) transduced with CK2α had enhanced proliferation and anchorage-independent growth. Inhibition of CKα using siRNA or small-molecule inhibitors (TBBz, CX-4945) reduced cell growth, decreased tumor size, and increased survival rates in GBM xenograft mouse models. We also verified that inhibition of CK2α decreased the activity of a well-known GBM-initiating cell regulator, ß-catenin. Loss of CK2α decreased two ß-catenin-regulated genes that are involved in GBM-initiating cell growth, OCT4 and NANOG. To determine the importance of CK2α in GBM stem cell maintenance, we reduced CK2α activity in primary GBM samples and tumor spheres derived from GBM patients. We discovered that loss of CK2α activity reduced the sphere-forming capacity of BTIC and decreased numerous GBM stem cell markers, including CD133, CD90, CD49f and A2B5. Our study suggests that CK2α is involved in GBM tumorigenesis by maintaining BTIC through the regulation of ß-catenin.

Cerebrospinal fluid and plasma oxytocin concentrations are positively correlated and negatively predict anxiety in children.

The neuropeptide oxytocin (OXT) exerts anxiolytic and prosocial effects in the central nervous system of rodents. A number of recent studies have attempted to translate these findings by investigating the relationships between peripheral (e.g., blood, urinary and salivary) OXT concentrations and behavioral functioning in humans. Although peripheral samples are easy to obtain in humans, whether peripheral OXT measures are functionally related to central OXT activity remains unclear. To investigate a possible relationship, we quantified OXT concentrations in concomitantly collected cerebrospinal fluid (CSF) and blood samples from child and adult patients undergoing clinically indicated lumbar punctures or other CSF-related procedures. Anxiety scores were obtained in a subset of child participants whose parents completed psychometric assessments. Findings from this study indicate that plasma OXT concentrations significantly and positively predict CSF OXT concentrations (r=0.56, P=0.0064, N=27). Moreover, both plasma (r=-0.92, P=0.0262, N=10) and CSF (r=-0.91, P=0.0335, N=10) OXT concentrations significantly and negatively predicted trait anxiety scores, consistent with the preclinical literature. Importantly, plasma OXT concentrations significantly and positively (r=0.96, P=0.0115, N=10) predicted CSF OXT concentrations in the subset of child participants who provided behavioral data. This study provides the first empirical support for the use of blood measures of OXT as a surrogate for central OXT activity, validated in the context of behavioral functioning. These preliminary findings also suggest that impaired OXT signaling may be a biomarker of anxiety in humans, and a potential target for therapeutic development in individuals with anxiety disorders.

Hydrocephalus decreases arterial spin-labeled cerebral perfusion.

BACKGROUND AND PURPOSE: Reduced cerebral perfusion has been observed with elevated intracranial pressure. We hypothesized that arterial spin-labeled CBF can be used as a marker for symptomatic hydrocephalus. MATERIALS AND METHODS: We compared baseline arterial spin-labeled CBF in 19 children (median age, 6.5 years; range, 1-17 years) with new posterior fossa brain tumors and clinical signs of intracranial hypertension with arterial spin-labeled CBF in 16 age-matched controls and 4 patients with posterior fossa tumors without ventriculomegaly or signs of intracranial hypertension. Measurements were recorded in the cerebrum at the vertex, deep gray nuclei, and periventricular white matter and were assessed for a relationship to ventricular size. In 16 symptomatic patients, we compared cerebral perfusion before and after alleviation of hydrocephalus. RESULTS: Patients with uncompensated hydrocephalus had lower arterial spin-labeled CBF than healthy controls for all brain regions interrogated (P < .001). No perfusion difference was seen between asymptomatic patients with posterior fossa tumors and healthy controls (P = 1.000). The median arterial spin-labeled CBF increased after alleviation of obstructive hydrocephalus (P < .002). The distance between the frontal horns inversely correlated with arterial spin-labeled CBF of the cerebrum (P = .036) but not the putamen (P = .156), thalamus (P = .111), or periventricular white matter (P = .121). CONCLUSIONS: Arterial spin-labeled-CBF was reduced in children with uncompensated hydrocephalus and restored after its alleviation. Arterial spin-labeled-CBF perfusion MR imaging may serve a future role in the neurosurgical evaluation of hydrocephalus, as a potential noninvasive method to follow changes of intracranial pressure with time.

Hematopoietic stem cells give rise to perivascular endothelial-like cells during brain tumor angiogenesis.

Bone marrow (BM) cells have recently been shown to give rise to skeletal, hepatic, cardiac, neural, and vascular endothelial tissues. However, it has been shown that this is the result of cell fusion rather than transdifferentiation of hematopoietic stem cells (HSC). For this study, we established a mouse model of brain tumor growth to investigate the differentiation potential of HSC into endothelial cells during brain tumor-induced angiogenesis. Nontransgenic (GFP(neg)) recipient mice were lethally irradiated, and their hematopoietic cells were subsequently repopulated by transplantation of a single green fluorescent protein (GFP)-expressing HSC. Rat glioma (RT-2/RAG) cells were then injected into the striatum of the chimeric mice 6-8 weeks post-transplantation. The animals were sacrificed 3-9 days after tumor implantation, and the mobilization, temporal-spatial distribution, and lineage-specific marker expression profile of the GFP(+) cells within the growing tumor were analyzed. We saw that GFP(+) cells gave rise to elongated, CD34(+)/Flk-1(+) cells that incorporated into the endothelium of tumor blood vessels. However, all GFP(+) cells were also CD45(+), and the presence of CD45 on the HSC-derived endothelial-like cells supports the hypothesis that the hematopoietic cells were recruited into the tumor milieu. The fact that we failed to demonstrate the expression of von Willebrand factor in these cells argues against a true endothelial identity. Nevertheless, the recruitment of HSC-derived endothelial-like cells was an extremely rare event in normal brain parenchyma, and, thus, the permissive influence afforded by the growing tumor appeared to enhance the perivascular tropism and acquisition of an endothelial phenotypes by a population of HSC-derived cells.

Cyclophosphamide/granulocyte colony-stimulating factor causes selective mobilization of bone marrow hematopoietic stem cells into the blood after M phase of the cell cycle.

Cytokine-mobilized peripheral blood hematopoietic stem cells (MPB HSC) are widely used for transplantation in the treatment of malignancies, but the mechanism of HSC mobilization is unclear. Although many HSC in bone marrow (BM) cycle rapidly and expand their numbers in response to cytoreductive agents, such as cyclophosphamide (CY), and cytokines, such as granulocyte colony-stimulating factor (G-CSF), MPB HSC are almost all in the G(0) or G(1) phase of the cell cycle. This has raised the question of whether a subset of noncycling BM HSC is selectively released, or whether cycling BM HSC are mobilized after M phase, but before the next S phase of the cell cycle. To distinguish between these possibilities, mice were treated with one dose of CY followed by daily doses of G-CSF, and dividing cells were marked by administration of bromodeoxyuridine (BrdU) during the interval that BM HSC are expanding. After CY and 4 days of G-CSF, 98.5% of the 2n DNA content long-term repopulating MPB (LT)-HSC stained positively for BrdU, and therefore derived from cells that divided during the treatment interval. Next, LT-HSC from mice previously treated with a single dose of CY, which kills cycling cells, and 3 daily doses of G-CSF, were nearly all killed by a second dose of CY, suggesting that CY/G-CSF causes virtually all LT-HSC to cycle. Analysis of cyclin D2 messenger RNA (mRNA) expression and total RNA content of MPB HSC suggests that these cells are mostly in G(1) phase. After CY/G-CSF treatment, virtually all BM LT-HSC enter the cell cycle; some of these HSC then migrate into the blood, specifically after M phase, and are rapidly recruited to particular hematopoietic organs.

The role of apoptosis in the regulation of hematopoietic stem cells: Overexpression of Bcl-2 increases both their number and repopulation potential.

Hematopoietic stem cells (HSC) give rise to cells of all hematopoietic lineages, many of which are short lived. HSC face developmental choices: self-renewal (remain an HSC with long-term multilineage repopulating potential) or differentiation (become an HSC with short-term multilineage repopulating potential and, eventually, a mature cell). There is a large overcapacity of differentiating hematopoietic cells and apoptosis plays a role in regulating their numbers. It is not clear whether apoptosis plays a direct role in regulating HSC numbers. To address this, we have employed a transgenic mouse model that overexpresses BCL-2 in all hematopoietic cells, including HSC: H2K-BCL-2. Cells from H2K-BCL-2 mice have been shown to be protected against a wide variety of apoptosis-inducing challenges. This block in apoptosis affects their HSC compartment. H2K-BCL-2-transgenic mice have increased numbers of HSC in bone marrow (2.4x wild type), but fewer of these cells are in the S/G(2)/M phases of the cell cycle (0.6x wild type). Their HSC have an increased plating efficiency in vitro, engraft at least as well as wild-type HSC in vivo, and have an advantage following competitive reconstitution with wild-type HSC.

In vivo proliferation and cell cycle kinetics of long-term self-renewing hematopoietic stem cells.

A rare set of hematopoietic stem cells (HSC) must undergo a massive expansion to produce mature blood cells. The phenotypic isolation of HSC from mice offers the opportunity to determine directly their proliferation kinetics. We analyzed the proliferation and cell cycle kinetics of long-term self-renewing HSC (LT-HSC) in normal adult mice. At any one time, approximately 5% of LT-HSC were in S/G2/M phases of the cell cycle and another 20% were in G1 phase. BrdUrd incorporation was used to determine the rate at which different cohorts of HSC entered the cell cycle over time. About 50% of LT-HSC incorporated BrdUrd by 6 days and >90% incorporated BrdUrd by 30 days. By 6 months, 99% of LT-HSC had incorporated BrdUrd. We calculated that approximately 8% of LT-HSC asynchronously entered the cell cycle per day. Nested reverse transcription-PCR analysis revealed cyclin D2 expression in a high proportion of LT-HSC. Although approximately 75% of LT-HSC are quiescent in G0 at any one time, all HSC are recruited into cycle regularly such that 99% of LT-HSC divide on average every 57 days.

Hematopoietic stem cells: challenges to expectations.

The past year provided a number of challenges to our expectations regarding hematopoietic stem cell (HSC) biology. Evidence has emerged that HSCs arise intraembryonically before they can be detected in the yolk sac. A number of genes that may regulate the formation, self-renewal, or differentiation of HSC have been identified. New markers for purifying HSCs have also been described. Although different groups have attributed different properties to HSCs, it now appears that the differences may be explained by variations in assay conditions rather than by differences in the HSCs themselves. Finally, insights have emerged into the complexity of the regulation of HSC proliferation and adhesion properties.