Early prognostication of overall survival for pediatric diffuse midline gliomas using MRI radiomics and machine learning

BACKGROUND: Diffuse midline gliomas (DMG) are aggressive pediatric brain tumors. MRI is the standard non-invasive tool for DMG diagnosis and monitoring. We developed an automatic pipeline to segment subregions of DMG and select radiomic features to predict patient overall survival (OS). METHODS: We acquired diagnostic and post-radiation therapy (RT) multisequence MRI (T1, T1ce, T2, and T2 FLAIR) and manual segmentations of 53 (internal cohort) and 16 (external cohort) DMG patients. We pretrained a deep learning model on an adult brain tumor dataset, and finetuned the model on our internal cohort to segment tumor core (TC) and whole tumor (WT). PyRadiomics and sequential feature selection were used for feature extraction and selection based on the segmented volumes. Two machine learning models were trained on our internal cohort to predict patient 1-year survival from diagnosis. One model used only diagnostic features (baseline study) and the other used both diagnostic and post-RT features (post-RT study). RESULTS: For segmentation, Dice score (mean [median]±SD) was 0.91 (0.94)±0.12/0.74 (0.83)±0.32 for TC and 0.88 (0.91)±0.07/0.86 (0.89)±0.06 for WT of internal/external cohorts. For OS prediction, accuracy was 77%/81% for the baseline study and 85%/78% for the post-RT study of internal/external cohorts. Our results suggest post-RT features are more discriminative and reliable compared with diagnostic features. Smaller post-RT TC/WT volume ratio indicates longer OS. Our model predicts with high accuracy which patients have short OS. CONCLUSIONS: We demonstrated how a fully automatic approach to compute imaging biomarkers of DMG from multisequence MRI can accurately and non-invasively predict overall survival for impacted pediatric patients.