An international study presenting a federated learning AI platform for pediatric brain tumors

While multiple factors impact disease, artificial intelligence (AI) studies in medicine often use small, non-diverse patient cohorts due to data sharing and privacy issues. Federated learning (FL) has emerged as a solution, enabling training across hospitals without direct data sharing. Here, we present FL-PedBrain, an FL platform for pediatric posterior fossa brain tumors, and evaluate its performance on a diverse, realistic, multi-center cohort. Pediatric brain tumors were targeted due to the scarcity of such datasets, even in tertiary care hospitals. Our platform orchestrates federated training for joint tumor classification and segmentation across 19 international sites. FL-PedBrain exhibits less than a 1.5% decrease in classification and a 3% reduction in segmentation performance compared to centralized data training. FL boosts segmentation performance by 20 to 30% on three external, out-of-network sites. Finally, we explore the sources of data heterogeneity and examine FL robustness in real-world scenarios with data imbalances.

CSPG4-targeting CAR-macrophages inhibit melanoma growth

Chimeric antigen receptor (CAR) T-cell therapy has revolutionized the treatment of hematological malignancies but has been clinically less effective in solid tumors. Engineering macrophages with CARs has emerged as a promising approach to overcome some of the challenges faced by CAR-T cells due to the macrophage's ability to easily infiltrate tumors, phagocytose their targets, and reprogram the immune response. We engineered CAR-macrophages (CAR-Ms) to target chondroitin sulfate proteoglycan 4 (CSPG4), an antigen expressed in melanoma, and several other solid tumors. CSPG4-targeting CAR-Ms exhibited specific phagocytosis of CSPG4-expressing melanoma cells. Combining CSPG4-targeting CAR-Ms with CD47 blocking antibodies synergistically enhanced CAR-M-mediated phagocytosis and effectively inhibited melanoma spheroid growth in 3D. Furthermore, CSPG4-targeting CAR-Ms inhibited melanoma tumor growth in mouse models. These results suggest that CSPG4-targeting CAR-M immunotherapy is a promising solid tumor immunotherapy approach for treating melanoma.

STATEMENT OF SIGNIFICANCE

We engineered macrophages with CARs as an alternative approach for solid tumor treatment. CAR-macrophages (CAR-Ms) targeting CSPG4, an antigen expressed in melanoma and other solid tumors, phagocytosed melanoma cells and inhibited melanoma growth in vivo . Thus, CSPG4-targeting CAR-Ms may be a promising strategy to treat patients with CSPG4-expressing tumors.

Morphological and ultrastructural investigation of the posterior atlanto-occipital membrane: Comparing children with Chiari malformation type I and controls

INTRODUCTION

The fibrous posterior atlanto-occipital membrane (PAOM) at the craniocervical junction is typically removed during decompression surgery for Chiari malformation type I (CM-I); however, its importance and ultrastructural architecture have not been investigated in children. We hypothesized that there are structural differences in the PAOM of patients with CM-I and those without.

METHODS

In this prospective study, blinded pathological analysis was performed on PAOM specimens from children who had surgery for CM-I and children who had surgery for posterior fossa tumors (controls). Clinical and radiographic data were collected. Statistical analysis included comparisons between the CM-I and control cohorts and correlations with imaging measures.

RESULTS

A total of 35 children (mean age at surgery 10.7 years; 94.3% white) with viable specimens for evaluation were enrolled: 24 with CM-I and 11 controls. There were no statistical demographic differences between the two cohorts. Four children had a family history of CM-I and five had a syndromic condition. The cohorts had similar measurements of tonsillar descent, syringomyelia, basion to C2, and condylar-to-C2 vertical axis (all p>0.05). The clival-axial angle was lower in patients with CM-I (138.1 vs. 149.3 degrees, p = 0.016). Morphologically, the PAOM demonstrated statistically higher proportions of disorganized architecture in patients with CM-I (75.0% vs. 36.4%, p = 0.012). There were no differences in PAOM fat, elastin, or collagen percentages overall and no differences in imaging or ultrastructural findings between male and female patients. Posterior fossa volume was lower in children with CM-I (163,234 mm3 vs. 218,305 mm3, p<0.001), a difference that persisted after normalizing for patient height (129.9 vs. 160.9, p = 0.028).

CONCLUSIONS

In patients with CM-I, the PAOM demonstrates disorganized architecture compared with that of control patients. This likely represents an anatomic adaptation in the presence of CM-I rather than a pathologic contribution.

A likely HOXC4 predisposition variant for Chiari malformations

OBJECTIVE

Inherited variants predisposing patients to type 1 or 1.5 Chiari malformation (CM) have been hypothesized but have proven difficult to confirm. The authors used a unique high-risk pedigree population resource and approach to identify rare candidate variants that likely predispose individuals to CM and protein structure prediction tools to identify pathogenicity mechanisms.

METHODS

By using the Utah Population Database, the authors identified pedigrees with significantly increased numbers of members with CM diagnosis. From a separate DNA biorepository of 451 samples from CM patients and families, 32 CM patients belonging to 1 or more of 24 high-risk Chiari pedigrees were identified. Two high-risk pedigrees had 3 CM-affected relatives, and 22 pedigrees had 2 CM-affected relatives. To identify rare candidate predisposition gene variants, whole-exome sequence data from these 32 CM patients belonging to 24 CM-affected related pairs from high-risk pedigrees were analyzed. The I-TASSER package for protein structure prediction was used to predict the structures of both the wild-type and mutant proteins found here.

RESULTS

Sequence analysis of the 24 affected relative pairs identified 38 rare candidate Chiari predisposition gene variants that were shared by at least 1 CM-affected pair from a high-risk pedigree. The authors found a candidate variant in HOXC4 that was shared by 2 CM-affected patients in 2 independent pedigrees. All 4 of these CM cases, 2 in each pedigree, exhibited a specific craniocervical bony phenotype defined by a clivoaxial angle less than 125°. The protein structure prediction results suggested that the mutation considered here may reduce the binding affinity of HOXC4 to DNA.

CONCLUSIONS

Analysis of unique and powerful Utah genetic resources allowed identification of 38 strong candidate CM predisposition gene variants. These variants should be pursued in independent populations. One of the candidates, a rare HOXC4 variant, was identified in 2 high-risk CM pedigrees, with this variant possibly predisposing patients to a Chiari phenotype with craniocervical kyphosis.

Diaphragm Pacing: A Safety, Appropriateness, Financial Neutrality, and Efficacy Analysis of Treating Chronic Respiratory Insufficiency

OBJECTIVES

This study aimed to evaluate the safety and applicability of treating chronic respiratory insufficiency with diaphragm pacing relative to mechanical ventilation.

MATERIALS AND METHODS

A literature review and analysis were conducted using the safety, appropriateness, financial neutrality, and efficacy principles.

RESULTS

Although mechanical ventilation is clearly indicated in acute respiratory failure, diaphragm pacing improves life expectancy, increases quality of life, and reduces complications in patients with chronic respiratory insufficiency.

CONCLUSION

Diaphragm pacing should be given more consideration in appropriately selected patients with chronic respiratory insufficiency.

Tacedinaline (CI-994), a class I HDAC inhibitor, targets intrinsic tumor growth and leptomeningeal dissemination in MYC-driven medulloblastoma while making them susceptible to anti-CD47-induced macrophage phagocytosis via NF-kB-TGM2 driven tumor inflammation

BACKGROUND

While major advances have been made in improving the quality of life and survival of children with most forms of medulloblastoma (MB), those with MYC-driven tumors (Grp3-MB) still suffer significant morbidity and mortality. There is an urgent need to explore multimodal therapeutic regimens which are effective and safe for children. Large-scale studies have revealed abnormal cancer epigenomes caused by mutations and structural alterations of chromatin modifiers, aberrant DNA methylation, and histone modification signatures. Therefore, targeting epigenetic modifiers for cancer treatment has gained increasing interest, and inhibitors for various epigenetic modulators have been intensively studied in clinical trials. Here, we report a cross-entity, epigenetic drug screen to evaluate therapeutic vulnerabilities in MYC amplified MB, which sensitizes them to macrophage-mediated phagocytosis by targeting the CD47-signal regulatory protein α (SIRPα) innate checkpoint pathway.

METHODS

We performed a primary screen including 78 epigenetic inhibitors and a secondary screen including 20 histone deacetylase inhibitors (HDACi) to compare response profiles in atypical teratoid/rhabdoid tumor (AT/RT, n=11), MB (n=14), and glioblastoma (n=14). This unbiased approach revealed the preferential activity of HDACi in MYC-driven MB. Importantly, the class I selective HDACi, CI-994, showed significant cell viability reduction mediated by induction of apoptosis in MYC-driven MB, with little-to-no activity in non-MYC-driven MB, AT/RT, and glioblastoma in vitro. We tested the combinatorial effect of targeting class I HDACs and the CD47-SIRPa phagocytosis checkpoint pathway using in vitro phagocytosis assays and in vivo orthotopic xenograft models.

RESULTS

CI-994 displayed antitumoral effects at the primary site and the metastatic compartment in two orthotopic mouse models of MYC-driven MB. Furthermore, RNA sequencing revealed nuclear factor-kB (NF-κB) pathway induction as a response to CI-994 treatment, followed by transglutaminase 2 (TGM2) expression, which enhanced inflammatory cytokine secretion. We further show interferon-γ release and cell surface expression of engulfment ('eat-me') signals (such as calreticulin). Finally, combining CI-994 treatment with an anti-CD47 mAb targeting the CD47-SIRPα phagocytosis checkpoint enhanced in vitro phagocytosis and survival in tumor-bearing mice.

CONCLUSION

Together, these findings suggest a dynamic relationship between MYC amplification and innate immune suppression in MYC amplified MB and support further investigation of phagocytosis modulation as a strategy to enhance cancer immunotherapy responses.

MRI Radiogenomics of Pediatric Medulloblastoma: A Multicenter Study

Background Radiogenomics of pediatric medulloblastoma (MB) offers an opportunity for MB risk stratification, which may aid therapeutic decision making, family counseling, and selection of patient groups suitable for targeted genetic analysis. Purpose To develop machine learning strategies that identify the four clinically significant MB molecular subgroups. Materials and Methods In this retrospective study, consecutive pediatric patients with newly diagnosed MB at MRI at 12 international pediatric sites between July 1997 and May 2020 were identified. There were 1800 features extracted from T2- and contrast-enhanced T1-weighted preoperative MRI scans. A two-stage sequential classifier was designed-one that first identifies non-wingless (WNT) and non-sonic hedgehog (SHH) MB and then differentiates therapeutically relevant WNT from SHH. Further, a classifier that distinguishes high-risk group 3 from group 4 MB was developed. An independent, binary subgroup analysis was conducted to uncover radiomics features unique to infantile versus childhood SHH subgroups. The best-performing models from six candidate classifiers were selected, and performance was measured on holdout test sets. CIs were obtained by bootstrapping the test sets for 2000 random samples. Model accuracy score was compared with the no-information rate using the Wald test. Results The study cohort comprised 263 patients (mean age ± SD at diagnosis, 87 months ± 60; 166 boys). A two-stage classifier outperformed a single-stage multiclass classifier. The combined, sequential classifier achieved a microaveraged F1 score of 88% and a binary F1 score of 95% specifically for WNT. A group 3 versus group 4 classifier achieved an area under the receiver operating characteristic curve of 98%. Of the Image Biomarker Standardization Initiative features, texture and first-order intensity features were most contributory across the molecular subgroups. Conclusion An MRI-based machine learning decision path allowed identification of the four clinically relevant molecular pediatric medulloblastoma subgroups. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Chaudhary and Bapuraj in this issue.

Targeted Next-Generation Sequencing Reveals Divergent Clonal Evolution in Components of Composite Pleomorphic Xanthoastrocytoma-Ganglioglioma

Composite pleomorphic xanthoastrocytoma-ganglioglioma (PXA-GG) is an extremely rare central nervous system neoplasm with 2 distinct but intermingled components. Whether this tumor represents a "collision tumor" of separate neoplasms or a monoclonal neoplasm with divergent evolution is poorly understood. Clinicopathologic studies and capture-based next generation sequencing were performed on extracted DNA from all available PXA-GG at 2 medical centers. Five PXA-GG were diagnosed in 1 male and 4 female patients ranging from 13 to 25 years in age. Four arose within the cerebral hemispheres; 1 presented in the cerebellar vermis. DNA was sufficient for analysis in 4 PXA components and 3 GG components. Four paired PXA and GG components harbored BRAF p.V600E hotspot mutations. The 4 sequenced PXA components demonstrated CDKN2A homozygous deletion by sequencing with loss of p16 (protein product of CDKN2A) expression by immunohistochemistry, which was intact in all assessed GG components. The PXA components also demonstrated more frequent copy number alterations relative to paired GG components. In one PXA-GG, shared chromosomal copy number alterations were identified in both components. Our findings support divergent evolution of the PXA and GG components from a common BRAF p.V600E-mutant precursor lesion, with additional acquisition of CDKN2A homozygous deletion in the PXA component as is typically seen in conventional PXA.

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 Signatures of Posterior Fossa Ependymoma: Molecular Subgroups and Risk Profiles

BACKGROUND

The risk profile for posterior fossa ependymoma (EP) depends on surgical and molecular status [Group A (PFA) versus Group B (PFB)]. While subtotal tumor resection is known to confer worse prognosis, MRI-based EP risk-profiling is unexplored. We aimed to apply machine learning strategies to link MRI-based biomarkers of high-risk EP and also to distinguish PFA from PFB.

METHODS

We extracted 1800 quantitative features from presurgical T2-weighted (T2-MRI) and gadolinium-enhanced T1-weighted (T1-MRI) imaging of 157 EP patients. We implemented nested cross-validation to identify features for risk score calculations and apply a Cox model for survival analysis. We conducted additional feature selection for PFA versus PFB and examined performance across three candidate classifiers.

RESULTS

For all EP patients with GTR, we identified four T2-MRI-based features and stratified patients into high- and low-risk groups, with 5-year overall survival rates of 62% and 100%, respectively (p < 0.0001). Among presumed PFA patients with GTR, four T1-MRI and five T2-MRI features predicted divergence of high- and low-risk groups, with 5-year overall survival rates of 62.7% and 96.7%, respectively (p = 0.002). T1-MRI-based features showed the best performance distinguishing PFA from PFB with an AUC of 0.86.

CONCLUSIONS

We present machine learning strategies to identify MRI phenotypes that distinguish PFA from PFB, as well as high- and low-risk PFA. We also describe quantitative image predictors of aggressive EP tumors that might assist risk-profiling after surgery. Future studies could examine translating radiomics as an adjunct to EP risk assessment when considering therapy strategies or trial candidacy.

Machine Assist for Pediatric Posterior Fossa Tumor Diagnosis: A Multinational Study

BACKGROUND

Clinicians and machine classifiers reliably diagnose pilocytic astrocytoma (PA) on magnetic resonance imaging (MRI) but less accurately distinguish medulloblastoma (MB) from ependymoma (EP). One strategy is to first rule out the most identifiable diagnosis.

OBJECTIVE

To hypothesize a sequential machine-learning classifier could improve diagnostic performance by mimicking a clinician's strategy of excluding PA before distinguishing MB from EP.

METHODS

We extracted 1800 total Image Biomarker Standardization Initiative (IBSI)-based features from T2- and gadolinium-enhanced T1-weighted images in a multinational cohort of 274 MB, 156 PA, and 97 EP. We designed a 2-step sequential classifier - first ruling out PA, and next distinguishing MB from EP. For each step, we selected the best performing model from 6-candidate classifier using a reduced feature set, and measured performance on a holdout test set with the microaveraged F1 score.

RESULTS

Optimal diagnostic performance was achieved using 2 decision steps, each with its own distinct imaging features and classifier method. A 3-way logistic regression classifier first distinguished PA from non-PA, with T2 uniformity and T1 contrast as the most relevant IBSI features (F1 score 0.8809). A 2-way neural net classifier next distinguished MB from EP, with T2 sphericity and T1 flatness as most relevant (F1 score 0.9189). The combined, sequential classifier was with F1 score 0.9179.

CONCLUSION

An MRI-based sequential machine-learning classifiers offer high-performance prediction of pediatric posterior fossa tumors across a large, multinational cohort. Optimization of this model with demographic, clinical, imaging, and molecular predictors could provide significant advantages for family counseling and surgical 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.

MRI-based radiomics for prognosis of pediatric diffuse intrinsic pontine glioma: an international study

Background

Diffuse intrinsic pontine gliomas (DIPGs) are lethal pediatric brain tumors. Presently, MRI is the mainstay of disease diagnosis and surveillance. We identify clinically significant computational features from MRI and create a prognostic machine learning model.

Methods

We isolated tumor volumes of T1-post-contrast (T1) and T2-weighted (T2) MRIs from 177 treatment-naïve DIPG patients from an international cohort for model training and testing. The Quantitative Image Feature Pipeline and PyRadiomics was used for feature extraction. Ten-fold cross-validation of least absolute shrinkage and selection operator Cox regression selected optimal features to predict overall survival in the training dataset and tested in the independent testing dataset. We analyzed model performance using clinical variables (age at diagnosis and sex) only, radiomics only, and radiomics plus clinical variables.

Results

All selected features were intensity and texture-based on the wavelet-filtered images (3 T1 gray-level co-occurrence matrix (GLCM) texture features, T2 GLCM texture feature, and T2 first-order mean). This multivariable Cox model demonstrated a concordance of 0.68 (95% CI: 0.61–0.74) in the training dataset, significantly outperforming the clinical-only model (C = 0.57 [95% CI: 0.49–0.64]). Adding clinical features to radiomics slightly improved performance (C = 0.70 [95% CI: 0.64–0.77]). The combined radiomics and clinical model was validated in the independent testing dataset (C = 0.59 [95% CI: 0.51–0.67], Noether’s test P = .02).

Conclusions

In this international study, we demonstrate the use of radiomic signatures to create a machine learning model for DIPG prognostication. Standardized, quantitative approaches that objectively measure DIPG changes, including computational MRI evaluation, could offer new approaches to assessing tumor phenotype and serve a future role for optimizing clinical trial eligibility and tumor surveillance.

Artificial intelligence for automatic cerebral ventricle segmentation and volume calculation: a clinical tool for the evaluation of pediatric hydrocephalus

OBJECTIVE

Imaging evaluation of the cerebral ventricles is important for clinical decision-making in pediatric hydrocephalus. Although quantitative measurements of ventricular size, over time, can facilitate objective comparison, automated tools for calculating ventricular volume are not structured for clinical use. The authors aimed to develop a fully automated deep learning (DL) model for pediatric cerebral ventricle segmentation and volume calculation for widespread clinical implementation across multiple hospitals.

METHODS

The study cohort consisted of 200 children with obstructive hydrocephalus from four pediatric hospitals, along with 199 controls. Manual ventricle segmentation and volume calculation values served as "ground truth" data. An encoder-decoder convolutional neural network architecture, in which T2-weighted MR images were used as input, automatically delineated the ventricles and output volumetric measurements. On a held-out test set, segmentation accuracy was assessed using the Dice similarity coefficient (0 to 1) and volume calculation was assessed using linear regression. Model generalizability was evaluated on an external MRI data set from a fifth hospital. The DL model performance was compared against FreeSurfer research segmentation software.

RESULTS

Model segmentation performed with an overall Dice score of 0.901 (0.946 in hydrocephalus, 0.856 in controls). The model generalized to external MR images from a fifth pediatric hospital with a Dice score of 0.926. The model was more accurate than FreeSurfer, with faster operating times (1.48 seconds per scan).

CONCLUSIONS

The authors present a DL model for automatic ventricle segmentation and volume calculation that is more accurate and rapid than currently available methods. With near-immediate volumetric output and reliable performance across institutional scanner types, this model can be adapted to the real-time clinical evaluation of hydrocephalus and improve clinician workflow.

Developmentally anomalous cerebellar encephalocele arising within the cerebellopontine angle and extending into the adjacent skull base in a pediatric patient

Lesions of the cerebellopontine angle (CPA) in young children are rare, with the most common being arachnoid cysts and epidermoid inclusion cysts. The authors report a case of an encephalocele containing heterotopic cerebellar tissue arising from the right middle cerebellar peduncle and filling the right internal acoustic canal in a 2-year-old female patient. Her initial presentation included a focal left 6th nerve palsy. Magnetic resonance imaging was suggestive of a high-grade tumor of the right CPA. The lesion was removed via a retrosigmoid approach, and histopathologic analysis revealed heterotopic atrophic cerebellar tissue. This report is the first description of a heterotopic cerebellar encephalocele within the CPA and temporal skull base of a pediatric patient.

Semaphorin 3A mediated brain tumor stem cell proliferation and invasion in EGFRviii mutant gliomas

BACKGROUND

Glioblastoma multiforme (GBM) is the most common primary brain tumor in adults, with a median survival of approximately 15 months. Semaphorin 3A (Sema3A), known for its axon guidance and antiangiogenic properties, has been implicated in GBM growth. We hypothesized that Sema3A directly inhibits brain tumor stem cell (BTSC) proliferation and drives invasion via Neuropilin 1 (Nrp1) and Plexin A1 (PlxnA1) receptors.

METHODS

GBM BTSC cell lines were assayed by immunostaining and PCR for levels of Semaphorin 3A (Sema3A) and its receptors Nrp1 and PlxnA1. Quantitative BrdU, cell cycle and propidium iodide labeling assays were performed following exogenous Sema3A treatment. Quantitative functional 2-D and 3-D invasion assays along with shRNA lentiviral knockdown of Nrp1 and PlxnA1 are also shown. In vivo flank studies comparing tumor growth of knockdown versus control BTSCs were performed. Statistics were performed using GraphPad Prism v7.

RESULTS

Immunostaining and PCR analysis revealed that BTSCs highly express Sema3A and its receptors Nrp1 and PlxnA1, with expression of Nrp1 in the CD133 positive BTSCs, and absence in differentiated tumor cells. Treatment with exogenous Sema3A in quantitative BrdU, cell cycle, and propidium iodide labeling assays demonstrated that Sema3A significantly inhibited BTSC proliferation without inducing cell death. Quantitative functional 2-D and 3-D invasion assays showed that treatment with Sema3A resulted in increased invasion. Using shRNA lentiviruses, knockdown of either NRP1 or PlxnA1 receptors abrogated Sema3A antiproliferative and pro-invasive effects. Interestingly, loss of the receptors mimicked Sema3A effects, inhibiting BTSC proliferation and driving invasion. Furthermore, in vivo studies comparing tumor growth of knockdown and control infected BTSCs implanted into the flanks of nude mice confirmed the decrease in proliferation with receptor KD.

CONCLUSIONS

These findings demonstrate the importance of Sema3A signaling in GBM BTSC proliferation and invasion, and its potential as a therapeutic target.

Deep Learning for Automated Delineation of Pediatric Cerebral Arteries on Pre-operative Brain Magnetic Resonance Imaging

Introduction: Surgical resection of brain tumors is often limited by adjacent critical structures such as blood vessels. Current intraoperative navigations systems are limited; most are based on two-dimensional (2D) guidance systems that require manual segmentation of any regions of interest (ROI; eloquent structures to avoid or tumor to resect). They additionally require time- and labor-intensive processing for any reconstruction steps. We aimed to develop a deep learning model for real-time fully automated segmentation of the intracranial vessels on preoperative non-angiogram imaging sequences. Methods: We identified 48 pediatric patients (10-months to 22-years old) with high resolution (0.5-1 mm axial thickness) isovolumetric, pre-operative T2 magnetic resonance images (MRIs). Twenty-eight patients had anatomically normal brains, and 20 patients had tumors or other lesions near the skull base. Manually segmented intracranial vessels (internal carotid, middle cerebral, anterior cerebral, posterior cerebral, and basilar arteries) served as ground truth labels. Patients were divided into 80/5/15% training/validation/testing sets. A modified 2-D Unet convolutional neural network (CNN) architecture implemented with 5 layers was trained to maximize the Dice coefficient, a measure of the correct overlap between the predicted vessels and ground truth labels. Results: The model was able to delineate the intracranial vessels in a held-out test set of normal and tumor MRIs with an overall Dice coefficient of 0.75. While manual segmentation took 1-2 h per patient, model prediction took, on average, 8.3 s per patient. Conclusions: We present a deep learning model that can rapidly and automatically identify the intracranial vessels on pre-operative MRIs in patients with normal vascular anatomy and in patients with intracranial lesions. The methodology developed can be translated to other critical brain structures. This study will serve as a foundation for automated high-resolution ROI segmentation for three-dimensional (3D) modeling and integration into an augmented reality navigation platform.

Diffusion tensor magnetic resonance imaging of the optic nerves in pediatric hydrocephalus

OBJECTIVE

While conventional imaging can readily identify ventricular enlargement in hydrocephalus, structural changes that underlie microscopic tissue injury might be more difficult to capture. MRI-based diffusion tensor imaging (DTI) uses properties of water motion to uncover changes in the tissue microenvironment. The authors hypothesized that DTI can identify alterations in optic nerve microstructure in children with hydrocephalus.

METHODS

The authors retrospectively reviewed 21 children (< 18 years old) who underwent DTI before and after neurosurgical intervention for acute obstructive hydrocephalus from posterior fossa tumors. Their optic nerve quantitative DTI metrics of mean diffusivity (MD) and fractional anisotropy (FA) were compared to those of 21 age-matched healthy controls.

RESULTS

Patients with hydrocephalus had increased MD and decreased FA in bilateral optic nerves, compared to controls (p < 0.001). Normalization of bilateral optic nerve MD and FA on short-term follow-up (median 1 day) after neurosurgical intervention was observed, as was near-complete recovery of MD on long-term follow-up (median 1.8 years).

CONCLUSIONS

DTI was used to demonstrate reversible alterations of optic nerve microstructure in children presenting acutely with obstructive hydrocephalus. Alterations in optic nerve MD and FA returned to near-normal levels on short- and long-term follow-up, suggesting that surgical intervention can restore optic nerve tissue microstructure. This technique is a safe, noninvasive imaging tool that quantifies alterations of neural tissue, with a potential role for evaluation of pediatric hydrocephalus.

Brain Iron Assessment after Ferumoxytol-enhanced MRI in Children and Young Adults with Arteriovenous Malformations: A Case-Control Study

Background Iron oxide nanoparticles are an alternative contrast agent for MRI. Gadolinium deposition has raised safety concerns, but it is unknown whether ferumoxytol administration also deposits in the brain. Purpose To investigate whether there are signal intensity changes in the brain at multiecho gradient imaging following ferumoxytol exposure in children and young adults. Materials and Methods This retrospective case-control study included children and young adults, matched for age and sex, with brain arteriovenous malformations who received at least one dose of ferumoxytol from January 2014 to January 2018. In participants who underwent at least two brain MRI examinations (subgroup), the first and last available examinations were analyzed. Regions of interests were placed around deep gray structures on quantitative susceptibility mapping and R2* images. Mean susceptibility and R2* values of regions of interests were recorded. Measurements were assessed by linear regression analyses: a between-group comparison of ferumoxytol-exposed and unexposed participants and a within-group (subgroup) comparison before and after exposure. Results Seventeen participants (mean age ± standard deviation, 13 years ± 5; nine male) were in the ferumoxytol-exposed (case) group, 21 (mean age, 14 years ± 5; 11 male) were in the control group, and nine (mean age, 12 years ± 6; four male) were in the subgroup. The mean number of ferumoxytol administrations was 2 ± 1 (range, one to four). Mean susceptibility (in parts per million [ppm]) and R2* (in inverse seconds [sec^-1]) values of the dentate (case participants: 0.06 ppm ± 0.04 and 23.87 sec^-1 ± 4.13; control participants: 0.02 ppm ± 0.03 and 21.7 sec^-1 ± 5.26), substantia nigrae (case participants: 0.08 ppm ± 0.06 and 27.46 sec^-1 ± 5.58; control participants: 0.04 ppm ± 0.05 and 24.96 sec^-1 ± 5.3), globus pallidi (case participants: 0.14 ppm ± 0.05 and 30.75 sec^-1 ± 5.14; control participants: 0.08 ppm ± 0.07 and 28.82 sec^-1 ± 6.62), putamina (case participants: 0.03 ppm ± 0.02 and 20.63 sec^-1 ± 2.44; control participants: 0.02 ppm ± 0.02 and 19.65 sec^-1 ± 3.6), caudate (case participants: -0.1 ppm ± 0.04 and 18.21 sec^-1 ± 3.1; control participants: -0.06 ppm ± 0.05 and 18.83 sec^-1 ± 3.32), and thalami (case participants: 0 ppm ± 0.03 and 16.49 sec^-1 ± 3.6; control participants: 0.02 ppm ± 0.02 and 18.38 sec^-1 ± 2.09) did not differ between groups (susceptibility, P = .21; R2*, P = .24). For the subgroup, the mean interval between the first and last ferumoxytol administration was 14 months ± 8 (range, 1-25 months). Mean susceptibility and R2* values of the dentate (first MRI: 0.06 ppm ± 0.05 and 25.78 sec^-1 ± 5.9; last MRI: 0.06 ppm ± 0.02 and 25.55 sec^-1 ± 4.71), substantia nigrae (first MRI: 0.06 ppm ± 0.06 and 28.26 sec^-1 ± 9.56; last MRI: 0.07 ppm ± 0.06 and 25.65 sec^-1 ± 6.37), globus pallidi (first MRI: 0.13 ppm ± 0.07 and 27.53 sec^-1 ± 8.88; last MRI: 0.14 ppm ± 0.06 and 29.78 sec^-1 ± 6.54), putamina (first MRI: 0.03 ppm ± 0.03 and 19.78 sec^-1 ± 3.51; last MRI: 0.03 ppm ± 0.02 and 19.73 sec^-1 ± 3.01), caudate (first MRI: -0.09 ppm ± 0.05 and 21.38 sec^-1 ± 4.72; last MRI: -0.1 ppm ± 0.05 and 18.75 sec^-1 ± 2.68), and thalami (first MRI: 0.01 ppm ± 0.02 and 17.65 sec^-1 ± 5.16; last MRI: 0 ppm ± 0.02 and 15.32 sec^-1 ± 2.49) did not differ between the first and last MRI examinations (susceptibility, P = .95; R2*, P = .54). Conclusion No overall significant differences were found in susceptibility and R2* values of deep gray structures to suggest retained iron in the brain between ferumoxytol-exposed and unexposed children and young adults with arteriovenous malformations and in those exposed to ferumoxytol over time. © RSNA, 2020.

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.