Acetazolamide-Challenged Arterial Spin Labeling Detects Augmented Cerebrovascular Reserve After Surgery for Moyamoya.

BACKGROUND: Cerebrovascular reserve (CVR) inversely correlates with stroke risk in children with Moyamoya disease and may be improved by revascularization surgery. We hypothesized that acetazolamide-challenged arterial spin labeling MR perfusion quantifies augmentation of CVR achieved by revascularization and correlates with currently accepted angiographic scoring criteria. METHODS: We retrospectively identified pediatric patients with Moyamoya disease or syndrome who received cerebral revascularization at ≤18 years of age between 2012 and 2019 at our institution. Using acetazolamide-challenged arterial spin labeling, we compared postoperative CVR to corresponding preoperative values and to postoperative perfusion outcomes classified by Matsushima grading. RESULTS: In this cohort, 32 patients (17 males) with Moyamoya underwent 29 direct and 16 indirect extracranial-intracranial bypasses at a median 9.7 years of age (interquartile range, 7.6-15.7). Following revascularization, median CVR increased within the ipsilateral middle cerebral artery territory (6.9 mL/100 g per minute preoperatively versus 16.5 mL/100 g per minute postoperatively, P<0.01). No differences were observed in the ipsilateral anterior cerebral artery (P=0.13) and posterior cerebral artery (P=0.48) territories. Postoperative CVR was higher in the ipsilateral middle cerebral artery territories of patients who achieved Matsushima grade A perfusion, in comparison to those with grades B or C (25.8 versus 17.5 mL, P=0.02). The method of bypass (direct or indirect) did not alter relative increases in CVR (8 versus 3.8 mL/100 g per minute, P=0.7). CONCLUSIONS: Acetazolamide-challenged arterial spin labeling noninvasively quantifies augmentation of CVR following surgery for Moyamoya disease and syndrome.

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.

Response assessment in paediatric low-grade glioma: recommendations from the Response Assessment in Pediatric Neuro-Oncology (RAPNO) working group.

Paediatric low-grade gliomas (also known as pLGG) are the most common type of CNS tumours in children. In general, paediatric low-grade gliomas show clinical and biological features that are distinct from adult low-grade gliomas, and the developing paediatric brain is more susceptible to toxic late effects of the tumour and its treatment. Therefore, response assessment in children requires additional considerations compared with the adult Response Assessment in Neuro-Oncology criteria. There are no standardised response criteria in paediatric clinical trials, which makes it more difficult to compare responses across studies. The Response Assessment in Pediatric Neuro-Oncology working group, consisting of an international panel of paediatric and adult neuro-oncologists, clinicians, radiologists, radiation oncologists, and neurosurgeons, was established to address issues and unique challenges in assessing response in children with CNS tumours. We established a subcommittee to develop consensus recommendations for response assessment in paediatric low-grade gliomas. Final recommendations were based on literature review, current practice, and expert opinion of working group members. Consensus recommendations include imaging response assessments, with additional guidelines for visual functional outcomes in patients with optic pathway tumours. As with previous consensus recommendations, these recommendations will need to be validated in prospective clinical trials.

Brain Tumors of Glial Origin.

Gliomas are a heterogeneous group of tumors with evolving classification based on genotype. Isocitrate dehydrogenase (IDH) mutation is an early event in the formation of some diffuse gliomas, and is the best understood mechanism of their epigenetic dysregulation. Glioblastoma may evolve from lower-grade lesions with IDH mutations, or arise independently from copy number changes in platelet-derived growth factor receptor alpha (PDGFRA) and phosphatase and tensin homolog (PTEN). Several molecular subtypes of glioblastoma arise from a common proneural precursor with a tendency toward transition to a mesenchymal subtype. Following oncogenic transformation, gliomas escape growth arrest through a distinct step of aberrant telomere reverse transcriptase (TERT) expression, or mutations in either alpha thalassemia/mental retardation syndrome (ATRX) or death-domain associated protein (DAXX) genes. Metabolic reprogramming allows gliomas to thrive in harsh microenvironments such as hypoxia, acidity, and nutrient depletion, which contribute to tumor initiation, maintenance, and treatment resistance.

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 Perfusion and Diffusion Abnormalities in Children Treated for Posterior Fossa Brain Tumors.

OBJECTIVE: To compare cerebral perfusion and diffusion in survivors of childhood posterior fossa brain tumor with neurologically normal controls and correlate differences with cognitive dysfunction. STUDY DESIGN: We analyzed retrospectively arterial spin-labeled cerebral blood flow (CBF) and apparent diffusion coefficient (ADC) in 21 patients with medulloblastoma (MB), 18 patients with pilocytic astrocytoma (PA), and 64 neurologically normal children. We generated ANCOVA models to evaluate treatment effects on the cerebral cortex, thalamus, caudate, putamen, globus pallidus, hippocampus, amygdala, nucleus accumbens, and cerebral white matter at time points an average of 5.7 years after original diagnosis. A retrospective review of patient charts identified 12 patients with neurocognitive data and in whom the relationship between IQ and magnetic resonance imaging variables was assessed for each brain structure. RESULTS: Patients with MB (all treated with surgery, chemotherapy, and radiation) had significantly lower global CBF relative to controls (10%-23% lower, varying by anatomic region, all adjusted P?

Citrate concentrations increase with hypoperfusion in pediatric diffuse intrinsic pontine glioma.

Citrate, a tricarboxylic acid cycle intermediate, is present in high concentrations in pediatric diffuse intrinsic pontine gliomas (DIPG). Since citrate increases during hypoxia in animal studies, we hypothesized that it accumulates in DIPG when hypoperfused. Relative tumor blood volumes (rTBV) were determined, using dynamic susceptibility contrast-enhanced magnetic resonance imaging, in twelve children [median age 8.2 (range 3.2-14.5) years] with DIPG and compared to citrate concentrations measured with in vivo proton magnetic resonance spectroscopy ((1)H MRS). Tissue perfusion and metabolite concentration were assessed at initial presentation and over the clinical course, yielding 36 and 46 perfusion and MR spectroscopy datasets, respectively. At presentation, DIPG blood volume was 60 ± 27 % of that measured for normal cerebellum. Citrate, which is not detectable in normal brain tissue, was present in DIPG at concentrations of 3.81 ± 1.44 mmol/kg tissue. Over the course of the disease and treatment, rTBV increased and citrate decreased (both p < 0.05) with an inverse correlation (p = 0.028). Citrate accumulation is associated with tissue hypoperfusion in DIPG.

Decreased tumor apparent diffusion coefficient correlates with objective response of pediatric low-grade glioma to bevacizumab.

Recent small, retrospective series suggest bevacizumab may be a therapeutic option for recurrent pediatric low-grade glioma (LGG). Assessment of therapeutic responses is complicated by the unpredictable natural history of these tumors. Because diffusion-weighted imaging quantifies microscopic water motion affected by cellular density and histologic features, we hypothesized that it may be helpful in monitoring therapeutic response of LGG to bevacizumab. We retrospectively reviewed eight consecutive patients, median age 4.8 (range 2.3-12.3) years at initiation of bevacizumab therapy for recurrent or refractory LGG. Patients received 10 mg/kg/dose every 2 weeks (median 16 doses/therapy course). Mean apparent diffusion coefficient (ADC) was measured and analyzed in respect to tumor volume. Following the first treatment course, seven of eight patients had reduced tumor volume (≥25 %) and ADC. The median decrease in tumor volume was 47% (range -6 to 78 %) and the median decrease in ADC was 14 % (range -5 to 30 %). The ADC was significantly decreased during therapy, whereas the decrease in volume was seen only after therapy completion. There was a positive correlation between percent change in tumor volume and ADC (p < 0.05). We report a decrease in tumor ADC during initial bevacizumab therapy that is accompanied by a decrease in volume following therapy. Imaging changes in microscopic water motion associated with histology may be useful in monitoring the therapeutic response of LGG to bevacizumab.

Diffusion-weighted MRI derived apparent diffusion coefficient identifies prognostically distinct subgroups of pediatric diffuse intrinsic pontine glioma.

While pediatric diffuse intrinsic pontine gliomas (DIPG) remain fatal, recent data have shown subgroups with distinct molecular biology and clinical behavior. We hypothesized that diffusion-weighted MRI can be used as a prognostic marker to stratify DIPG subsets with distinct clinical behavior. Apparent diffusion coefficient (ADC) values derived from diffusion-weighted MRI were computed in 20 consecutive children with treatment-naïve DIPG tumors. The median ADC for the cohort was used to stratify the tumors into low and high ADC groups. Survival, gender, therapy, and potential steroid effects were compared between the ADC groups. Median age at diagnosis was 6.6 (range 2.3-13.2) years, with median follow-up seven (range 1-36) months. There were 14 boys and six girls. Seventeen patients received radiotherapy, five received chemotherapy, and six underwent cerebrospinal fluid diversion. The median ADC of 1,295 × 10(-6) mm(2)/s for the cohort partitioned tumors into low or high diffusion groups, which had distinct median survivals of 3 and 13 months, respectively (log-rank p < 0.001). Low ADC tumors were found only in boys, whereas high ADC tumors were found in both boys and girls. Available tissue specimens in three low ADC tumors demonstrated high-grade histology, whereas one high ADC tumor demonstrated low-grade histology with a histone H3.1 K27M mutation and high-grade metastatic lesion at autopsy. ADC derived from diffusion-weighted MRI may identify prognostically distinct subgroups of pediatric DIPG.

Surveillance imaging in children with malignant CNS tumors: low yield of spine MRI.

Magnetic resonance imaging (MRI) is routinely obtained in patients with central nervous system (CNS) tumors, but few studies have been conducted to evaluate this practice. We assessed the benefits of surveillance MRI and more specifically spine MRI in a contemporary cohort. We evaluated MRI results of children diagnosed with CNS tumors from January 2000 to December 2011. Children with at least one surveillance MRI following the diagnosis of medulloblastoma (MB), atypical teratoid rhabdoid tumor (ATRT), pineoblastoma (PB), supratentorial primitive neuroectodermal tumor, supratentorial high-grade glioma (World Health Organization grade III-IV), CNS germ cell tumors or ependymoma were included. A total of 2,707 brain and 1,280 spine MRI scans were obtained in 258 patients. 97% of all relapses occurred in the brain and 3% were isolated to the spine. Relapse was identified in 226 (8%) brain and 48 (4%) spine MRI scans. The overall rate of detecting isolated spinal relapse was 9/1,000 and 7/1,000 for MB patients. MRI performed for PB showed the highest rate for detecting isolated spinal recurrence with 49/1,000. No initial isolated spinal relapse was identified in patients with glioma, supratentorial primitive neuroectodermal tumor and ATRT. Isolated spinal recurrences are infrequent in children with malignant CNS tumors and the yield of spine MRI is very low. Tailoring surveillance spine MRI to patients with higher spinal relapse risk such as PB, MB with metastatic disease and within 3 years of diagnosis could improve allocation of resources without compromising patient care.

The incidence and significance of multiple lesions in glioblastoma.

The location and distribution of glioblastoma (GBM) within the brain parenchyma plays an important role in surgical and radiation planning. Prior studies have reported incidences of multiple lesions at the time of diagnosis ranging from 0.5 to 20 %. Multiple lesions can be further categorized as multifocal (multiple areas involved, but with a clear path of spread from one lesion to another) or multicentric (multiple lesions, no clear path of spread). In this retrospective study, we reviewed our experience with GBM and found the incidence of multiple lesions at time of diagnosis was 35 %, much higher than previously suggested in the literature. Patients with single lesions had an improved overall survival when compared to patients with multiple lesions (18 vs. 10 months). Patients with multicentric lesions fared the worst, with average survival of 3 months. However, the difference between single and multiple lesions (multifocal or multicentric) was no longer significant when taking into consideration age, Karnofsky performance score (KPS) and extent of resection by multivariate analysis. Age, KPS, gross total resection, and MGMT status were independent predictors of outcome. Multiple lesions did not independently confer a worse outcome, but were associated with lower KPS scores and inability to perform gross total resection. These findings suggest that single, multiple and multicentric imaging exams represent a spectrum of presentations of a single disease. The rate of multiple lesions reported here may be the result of improved imaging technology, suggesting that incidence of multiple lesions will continue to increase as imaging technology advances.

Relapse patterns in pediatric embryonal central nervous system tumors.

Embryonal tumors of the central nervous system (CNS) share histological features and were therefore initially grouped as primitive neuroectodermal tumors (PNET) and treated similarly. We sought to determine the relapse patterns of specific embryonal CNS tumors. We conducted a historical cohort study of children diagnosed with CNS embryonal tumors from January 2000 to December 2011 in two pediatric neuro-oncology centers. Patients of 21 years of age or younger at time of presentation with a diagnosis of medulloblastoma, supratentorial PNET, pineoblastoma or atypical teratoid/rhabdoid tumor (ATRT) and at least one surveillance MRI were included. A total of 133 patients met inclusion criteria and 49 (37 %) patients relapsed during the observation period. The majority (79 %) of sPNET relapses were local, whereas all (100 %) PB relapses were associated with diffuse leptomeningeal disease. Relapse patterns for MB were more diverse with local recurrence in 27 %, distant recurrence in 35 % and diffuse leptomeningeal disease in 38 %. The frequency of relapses involving the spine differed (p < 0.001) between tumor types (MB 28/55 [51 %], sPNET 3/33 [9 %], ATRT 3/7 [43 %] and PB 12/12 [100 %]). No sPNET patients had isolated spinal relapse (0/14). Embryonal tumors were found to have divergent patterns of recurrence. While medulloblastoma has variable relapse presentations, sPNET relapses locally and pineoblastoma recurs with diffuse leptomeningeal disease involving the spine. These results point toward possibly new upfront treatment stratification among embryonal tumors in accordance with relapse pattern.

Distinctive MRI features of pediatric medulloblastoma subtypes.

OBJECTIVE: We hypothesized that the apparent diffusion coefficient (ADC) and other MRI features can be used to predict medulloblastoma histologic subtypes, as defined by the World Health Organization (WHO) in WHO Classification of Tumours of the Central Nervous System. MATERIALS AND METHODS: A retrospective review of pediatric patients with medulloblastoma between 1989 and 2011 identified 38 patients with both pretreatment MRI and original pathology slides. The mean and minimum tumor ADC values and conventional MRI features were compared among medulloblastoma histologic subtypes. RESULTS: The cohort of 38 patients included the following histologic subtypes: 24 classic medulloblastomas, nine large cell (LC) or anaplastic medulloblastomas, four desmoplastic medulloblastomas, and one medulloblastoma with extensive nodularity. The median age at diagnosis was 8 years (range, 1-21 years) and the median follow-up time was 33 months (range, 0-150 months). The mean ADC (× 10(-3) mm(2)/s) was lower in classic medulloblastoma (0.733 ± 0.046 [SD]) than in LC or anaplastic medulloblastoma (0.935 ± 0.127) (Mann-Whitney test, p = 0.004). Similarly, the minimum ADC was lower in classic medulloblastoma (average ± SD, 0.464 ± 0.056) than in LC or anaplastic medulloblastoma (0.630 ± 0.053) (p = 0.004). The MRI finding of focal cysts correlated with the classic and desmoplastic subtypes (Fisher exact test, p = 0.026). Leptomeningeal enhancement positively correlated with the LC or anaplastic medulloblastoma subtype and inversely correlated with the classic medulloblastoma and desmoplastic medulloblastoma subtypes (p = 0.04). Ring enhancement correlated with tumor necrosis (p = 0.022) and with the LC or anaplastic medulloblastoma histologic subtype (p < 0.001). CONCLUSION: The LC or anaplastic medulloblastoma subtype was associated with increased ADC and with ring enhancement, the latter of which correlated with tumor necrosis. These features could be considered in the evaluation of high-risk medulloblastoma subtypes.

Comparison of readout-segmented echo-planar imaging (EPI) and single-shot EPI in clinical application of diffusion-weighted imaging of the pediatric brain.

OBJECTIVE: Readout-segmented echo-planar imaging (EPI) has been suggested as an alternative to single-shot EPI for diffusion-weighted imaging (DWI) with reduced distortion. However, clinical comparisons of readout-segmented EPI and EPI DWI are limited by unmatched imaging parameters and reconstruction procedures. Our goal was to compare the clinical utility of generalized autocalibrating partial parallel acquisition (GRAPPA)-accelerated readout-segmented EPI DWI with GRAPPA-accelerated EPI DWI for visualization of the pediatric brain in regions prone to distortion, such as the orbit, skull base, and posterior fossa. SUBJECTS AND METHODS: Thirty consecutive patients (mean age, 7.8 years) presenting with orbital, skull base, and posterior fossa neuropathologic abnormalities were scanned at 3 T. Images were obtained using GRAPPA-accelerated readout-segmented EPI and GRAPPA-accelerated EPI with an identical scanning time, acceleration factor, target resolution, and image postprocessing procedure. The two datasets were independently reviewed by two blinded neuroradiologists. Imaging studies were evaluated for resolution, signal-to-noise ratio (SNR), contrast, distortion, lesion conspicuity, and diagnostic confidence and graded using a 7-point Likert scale (1, nondiagnostic; 7, outstanding). RESULTS: There was good reader agreement in the scores (κ = 0.66; 95% CI, 0.54-0.78). The mean scores for EPI and readout-segmented EPI, respectively, were as follows: resolution, 5.0 and 6.0; SNR, 5.5 and 3.0; contrast, 3.7 and 3.2; distortion, 4.8 and 6.0; lesion conspicuity, 4.6 and 5.1; and diagnostic confidence, 4.7 and 5.4. Readout-segmented EPI was superior in resolution, distortion reduction, lesion conspicuity, and diagnostic confidence, whereas EPI scored better in SNR and contrast. Readout-segmented EPI was considered the better sequence overall in 85% of the cases. CONCLUSION: This study shows the benefits of improved resolution and reduced distortion of readout-segmented EPI in evaluating the orbit, skull base, and posterior fossa, sites of common neuropathologic abnormalities in children.

A multi-institutional pediatric dataset of clinical radiology MRIs by the Children's Brain Tumor Network.

Pediatric brain and spinal cancers remain the leading cause of cancer-related death in children. Advancements in clinical decision-support in pediatric neuro-oncology utilizing the wealth of radiology imaging data collected through standard care, however, has significantly lagged other domains. Such data is ripe for use with predictive analytics such as artificial intelligence (AI) methods, which require large datasets. To address this unmet need, we provide a multi-institutional, large-scale pediatric dataset of 23,101 multi-parametric MRI exams acquired through routine care for 1,526 brain tumor patients, as part of the Children's Brain Tumor Network. This includes longitudinal MRIs across various cancer diagnoses, with associated patient-level clinical information, digital pathology slides, as well as tissue genotype and omics data. To facilitate downstream analysis, treatment-naïve images for 370 subjects were processed and released through the NCI Childhood Cancer Data Initiative via the Cancer Data Service. Through ongoing efforts to continuously build these imaging repositories, our aim is to accelerate discovery and translational AI models with real-world data, to ultimately empower precision medicine for children.

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 < .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 = .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.

Gigli saw facilitates safe minimal access sagittal suturectomy in infants.

The authors describe the use of the Gigli saw for craniectomy in minimal access surgery to address sagittal craniosynostosis. This modification allows for supine positioning and avoidance of potential brain compression with endoscopic instruments, and provides visually clear, safe, and facile removal of the fused suture and surrounding calvaria. The video can be found here: https://vimeo.com/511568750.

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.

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.