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. 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. 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). 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.
Bilateral basal ganglia abnormalities on MRI are observed in a wide variety of childhood disorders. MRI pattern recognition can enable rationalization of investigations and also complement clinical and molecular findings, particularly confirming genomic findings and also enabling new gene discovery. A pattern recognition approach in children with bilateral basal ganglia abnormalities on brain MRI was undertaken in this international multicentre cohort study. Three hundred and five MRI scans belonging to 201 children with 34 different disorders were rated using a standard radiological scoring proforma. In addition, literature review on MRI patterns was undertaken in these 34 disorders and 59 additional disorders reported with bilateral basal ganglia MRI abnormalities. Cluster analysis on first MRI findings from the study cohort grouped them into four clusters: Cluster 1—T2-weighted hyperintensities in the putamen; Cluster 2—T2-weighted hyperintensities or increased MRI susceptibility in the globus pallidus; Cluster 3—T2-weighted hyperintensities in the globus pallidus, brainstem and cerebellum with diffusion restriction; Cluster 4—T1-weighted hyperintensities in the basal ganglia. The 34 diagnostic categories included in this study showed dominant clustering in one of the above four clusters. Inflammatory disorders grouped together in Cluster 1. Mitochondrial and other neurometabolic disorders were distributed across clusters 1, 2 and 3, according to lesions dominantly affecting the striatum (Cluster 1: glutaric aciduria type 1, propionic acidaemia, 3-methylglutaconic aciduria with deafness, encephalopathy and Leigh-like syndrome and thiamine responsive basal ganglia disease associated with SLC19A3), pallidum (Cluster 2: methylmalonic acidaemia, Kearns Sayre syndrome, pyruvate dehydrogenase complex deficiency and succinic semialdehyde dehydrogenase deficiency) or pallidum, brainstem and cerebellum (Cluster 3: vigabatrin toxicity, Krabbe disease). The Cluster 4 pattern was exemplified by distinct T1-weighted hyperintensities in the basal ganglia and other brain regions in genetically determined hypermanganesemia due to SLC39A14 and SLC30A10. Within the clusters, distinctive basal ganglia MRI patterns were noted in acquired disorders such as cerebral palsy due to hypoxic ischaemic encephalopathy in full-term babies, kernicterus and vigabatrin toxicity and in rare genetic disorders such as 3-methylglutaconic aciduria with deafness, encephalopathy and Leigh-like syndrome, thiamine responsive basal ganglia disease, pantothenate kinase-associated neurodegeneration, TUBB4A and hypermanganesemia. Integrated findings from the study cohort and literature review were used to propose a diagnostic algorithm to approach bilateral basal ganglia abnormalities on MRI. After integrating clinical summaries and MRI findings from the literature review, we developed a prototypic decision-making electronic tool to be tested using further cohorts and clinical practice.
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.
27 November 2024 in Journal of Medical Imaging and Radiation Oncology
SUMMARY: X-linked deafness-2 (DFNX2) is an X-linked recessive disorder characterized by profound sensorineural hearing loss and a pathognomonic temporal bone deformity. Because hypothalamic malformations associated with DFNX2 have been rarely described, we aimed to further describe these lesions and compare them with features of a nonaffected population. All patients diagnosed with DFNX2 between 2006 and 2019 were included and compared with age-matched patients with normal MR imaging findings and without hypothalamic dysfunction. MR imaging features differing between groups were selected to help identify DFNX2. Sensitivity and specificity were calculated for these features. Agreement among 3 radiologists was quantified using the index κ. Information on the presence or absence of gelastic seizures, precocious puberty, or delayed puberty was also gathered. We selected distinctive MR imaging features of hypothalamic malformations in DFNX2. The feature selected on axial T2 images was the folded appearance of the ventromedial hypothalamus (sensitivity, 100%; specificity, 95.8%) characterized by an abnormal internal/external cleft (sensitivity, 100%; specificity, 95.7%). On coronal T2, the first distinctive feature was a concave morphology of the medial eminence (sensitivity, 100%; specificity, 97.1%), the second feature was at least 1 hypothalamic-septum angle ≥90° (sensitivity, 90%; specificity, 72.5%), and the third feature was a forebrain-hypothalamic craniocaudal length of ≥6 mm (sensitivity, 70%; specificity, 79.7%). Clinical features were also distinctive because 9 patients with DFNX2 did not present with gelastic seizures or precocious puberty. One patient had delayed puberty. The κ index and intraclass correlation coefficient ranged between 0.78 and 0.95. Imaging and clinical features of the hypothalamus suggest that there is a hypothalamic malformation associated with DFNX2. Early assessment for pubertal delay is proposed.
9 June 2022 in European Journal of Paediatric Neurology