The authors conducted a narrative review of clinical presentations, molecular changes, imaging features, radiology-molecular correlations, and current therapies regarding four high-grade types of pediatric gliomas: diffuse midline glioma (DMG-H3K27a), diffuse hemispheric glioma (DHG-H3G34m), diffuse pediatric-type high-grade glioma (H3/IDH-wt, phGG), and infant-type hemispheric glioma (IHG). The article reviewed relevant literature through January 2025 with a preference for peer-reviewed publications published after the 2016 WHO classification revision of these tumor types. Within this review, the article also discusses the use of artificial intelligence and deep learning in radiogenomics to detect imaging biomarkers that correlate with the molecular features that distinguish these types of gliomas. Of the four gliomas examined, DMG-H3K27a and DHG-H3G34m have demonstrated the most robust benefit from AI integration in diagnosis and treatment planning. Meta-analysis of radiomics studies using machine learning for DMG-H3K27a had a pooled sensitivity of 0.91, specificity of 0.81, and AUC of 0.84 for this tumor type. For DHG-H3G34m, a comparison of radiomics to the Visually AcceSAble Rembrandt Images (VASARI) MRI feature analysis (a standardized framework for visually evaluating MRIs to classify gliomas) showed an AUC of 0.925 that outperformed the AUC of the VASARI feature of 0.843. While the study is limited by small sample sizes and variable imaging protocols, it is clear that radiogenomics and AI have a promising role in tumor characterization and treatment planning.