3D Capsule Networks for Brain Image Segmentation.

BACKGROUND AND PURPOSE: Current autosegmentation models such as UNets and nnUNets have limitations, including the inability to segment images that are not represented during training and lack of computational efficiency. 3D capsule networks have the potential to address these limitations. MATERIALS AND METHODS: We used 3430 brain MRIs, acquired in a multi-institutional study, to train and validate our models. We compared our capsule network with standard alternatives, UNets and nnUNets, on the basis of segmentation efficacy (Dice scores), segmentation performance when the image is not well-represented in the training data, performance when the training data are limited, and computational efficiency including required memory and computational speed. RESULTS: The capsule network segmented the third ventricle, thalamus, and hippocampus with Dice scores of 95%, 94%, and 92%, respectively, which were within 1% of the Dice scores of UNets and nnUNets. The capsule network significantly outperformed UNets in segmenting images that were not well-represented in the training data, with Dice scores 30% higher. The computational memory required for the capsule network is less than one-tenth of the memory required for UNets or nnUNets. The capsule network is also >25% faster to train compared with UNet and nnUNet. CONCLUSIONS: We developed and validated a capsule network that is effective in segmenting brain images, can segment images that are not well-represented in the training data, and is computationally efficient compared with alternatives.

PET/MRI in Pediatric Neuroimaging: Primer for Clinical Practice.

Modern pediatric imaging seeks to provide not only exceptional anatomic detail but also physiologic and metabolic information of the pathology in question with as little radiation penalty as possible. Hybrid PET/MR imaging combines exquisite soft-tissue information obtained by MR imaging with functional information provided by PET, including metabolic markers, receptor binding, perfusion, and neurotransmitter release data. In pediatric neuro-oncology, PET/MR imaging is, in many ways, ideal for follow-up compared with PET/CT, given the superiority of MR imaging in neuroimaging compared with CT and the lower radiation dose, which is relevant in serial imaging and long-term follow-up of pediatric patients. In addition, although MR imaging is the main imaging technique for the evaluation of spinal pathology, PET/MR imaging may provide useful information in several clinical scenarios, including tumor staging and follow-up, treatment response assessment of spinal malignancies, and vertebral osteomyelitis. This review article covers neuropediatric applications of PET/MR imaging in addition to considerations regarding radiopharmaceuticals, imaging protocols, and current challenges to clinical implementation.

Machine Learning in Differentiating Gliomas from Primary CNS Lymphomas: A Systematic Review, Reporting Quality, and Risk of Bias Assessment.

BACKGROUND: Differentiating gliomas and primary CNS lymphoma represents a diagnostic challenge with important therapeutic ramifications. Biopsy is the preferred method of diagnosis, while MR imaging in conjunction with machine learning has shown promising results in differentiating these tumors. PURPOSE: Our aim was to evaluate the quality of reporting and risk of bias, assess data bases with which the machine learning classification algorithms were developed, the algorithms themselves, and their performance. DATA SOURCES: Ovid EMBASE, Ovid MEDLINE, Cochrane Central Register of Controlled Trials, and the Web of Science Core Collection were searched according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. STUDY SELECTION: From 11,727 studies, 23 peer-reviewed studies used machine learning to differentiate primary CNS lymphoma from gliomas in 2276 patients. DATA ANALYSIS: Characteristics of data sets and machine learning algorithms were extracted. A meta-analysis on a subset of studies was performed. Reporting quality and risk of bias were assessed using the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) and Prediction Model Study Risk Of Bias Assessment Tool. DATA SYNTHESIS: The highest area under the receiver operating characteristic curve (0.961) and accuracy (91.2%) in external validation were achieved by logistic regression and support vector machines models using conventional radiomic features. Meta-analysis of machine learning classifiers using these features yielded a mean area under the receiver operating characteristic curve of 0.944 (95% CI, 0.898-0.99). The median TRIPOD score was 51.7%. The risk of bias was high for 16 studies. LIMITATIONS: Exclusion of abstracts decreased the sensitivity in evaluating all published studies. Meta-analysis had high heterogeneity. CONCLUSIONS: Machine learning-based methods of differentiating primary CNS lymphoma from gliomas have shown great potential, but most studies lack large, balanced data sets and external validation. Assessment of the studies identified multiple deficiencies in reporting quality and risk of bias. These factors reduce the generalizability and reproducibility of the findings.

Diffusion Characteristics of Pediatric Diffuse Midline Gliomas with Histone H3-K27M Mutation Using Apparent Diffusion Coefficient Histogram Analysis.

BACKGROUND AND PURPOSE: Diffuse midline gliomas with histone H3 K27M mutation are biologically aggressive tumors with poor prognosis defined as a new diagnostic entity in the 2016 World Health Organization Classification of Tumors of the Central Nervous System. There are no qualitative imaging differences (enhancement, border, or central necrosis) between histone H3 wildtype and H3 K27M-mutant diffuse midline gliomas. Herein, we evaluated the utility of diffusion-weighted imaging to distinguish H3 K27M-mutant from histone H3 wildtype diffuse midline gliomas. MATERIALS AND METHODS: We identified 31 pediatric patients (younger than 21 years of age) with diffuse gliomas centered in midline structures that had undergone assessment for histone H3 K27M mutation. We measured ADC within these tumors using a voxel-based 3D whole-tumor measurement method. RESULTS: Our cohort included 18 infratentorial and 13 supratentorial diffuse gliomas centered in midline structures. Twenty-three (74%) tumors carried H3-K27M mutations. There was no difference in ADC histogram parameters (mean, median, minimum, maximum, percentiles) between mutant and wild-type tumors. Subgroup analysis based on tumor location also did not identify a difference in histogram descriptive statistics. Patients who survived <1 year after diagnosis had lower median ADC (1.10 × 10-3mm2/s; 95% CI, 0.90-1.30) compared with patients who survived >1 year (1.46 × 10-3mm2/s; 95% CI, 1.19-1.67; P < .06). Average ADC values for diffuse midline gliomas were 1.28 × 10-3mm2/s (95% CI, 1.21-1.34) and 0.86 × 10-3mm2/s (95% CI, 0.69-1.01) for hemispheric glioblastomas with P < .05. CONCLUSIONS: Although no statistically significant difference in diffusion characteristics was found between H3-K27M mutant and H3 wildtype diffuse midline gliomas, lower diffusivity corresponds to a lower survival rate at 1 year after diagnosis. These findings can have an impact on the anticipated clinical course for this patient population and offer providers and families guidance on clinical outcomes.

Imaging Characteristics of Pediatric Diffuse Midline Gliomas with Histone H3 K27M Mutation.

BACKGROUND AND PURPOSE: The 2016 World Health Organization Classification of Tumors of the Central Nervous System includes "diffuse midline glioma with histone H3 K27M mutation" as a new diagnostic entity. We describe the MR imaging characteristics of this new tumor entity in pediatric patients. MATERIALS AND METHODS: We retrospectively reviewed imaging features of pediatric patients with midline gliomas with or without the histone H3 K27 mutation. We evaluated the imaging features of these tumors on the basis of location, enhancement pattern, and necrosis. RESULTS: Among 33 patients with diffuse midline gliomas, histone H3 K27M mutation was present in 24 patients (72.7%) and absent in 9 (27.3%). Of the tumors, 27.3% (n = 9) were located in the thalamus; 42.4% (n = 14), in the pons; 15% (n = 5), within the vermis/fourth ventricle; and 6% (n = 2), in the spinal cord. The radiographic features of diffuse midline gliomas with histone H3 K27M mutation were highly variable, ranging from expansile masses without enhancement or necrosis with large areas of surrounding infiltrative growth to peripherally enhancing masses with central necrosis with significant mass effect but little surrounding T2/FLAIR hyperintensity. When we compared diffuse midline gliomas on the basis of the presence or absence of histone H3 K27M mutation, there was no significant correlation between enhancement or border characteristics, infiltrative appearance, or presence of edema. CONCLUSIONS: We describe, for the first time, the MR imaging features of pediatric diffuse midline gliomas with histone H3 K27M mutation. Similar to the heterogeneous histologic features among these tumors, they also have a diverse imaging appearance without distinguishing features from histone H3 wildtype diffuse gliomas.