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

Researchers

Journal

Modalities

Models

Abstract

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).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).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.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.This is the first fully automatic deep learning/machine learning MRI study to predict DMG survival.Post-radiation therapy features are more discriminative and reliable than diagnostic features. Smaller post-radiation therapy tumor core/whole tumor volume ratio indicates better prognosis.Previous studies on pediatric DMG prognostication relied on manual tumor segmentation, which is time-consuming and has high inter-operator variability. There is a great need for non-invasive prognostic imaging tools that can be universally used. Such tools should be automatic, objective, and easy to use in multi-institutional clinical trials. We developed a fully automatic imaging tool to segment subregions of DMG and select radiomic features to predict patient overall survival (OS). Our acquired 4 sequences of MRI for each patient, at both diagnostic and post-radiation therapy from 2 institutions, were more comprehensive than previous studies. The proposed method achieved high accuracy in DMG segmentation and survival prediction, especially for patients having short OS. The proposed method will be the foundation of increasing the utility of MRI as a tool for predicting clinical outcome, stratifying patients into risk-groups for improved therapeutic management and monitoring therapeutic response with greater sensitivity and an opportunity to adapt treatment.