Fig. 4: Prognostic Modeling Using Clinical Information and Tumor Volumes Derived from Manual and Automated Segmentations Under Complete and Incomplete MRI Inputs.
From: AI-powered segmentation and prognosis with missing MRI in pediatric brain tumors
Top two rows: Kaplan–Meier survival curves and corresponding forest plots from Cox models using age, sex, treatment group, and tumor volumes derived from manual segmentations, Baseline, and Dropout models with complete MRI input. Bottom two rows: Corresponding Kaplan-Meier curves and forest plots under missing FLAIR, with segmentation handled via Dropout, Synthesis, or Baseline models. Kaplan–Meier curves are annotated with the concordance index (C-index) and log-rank p-value. Forest plots show hazard ratios with 95% confidence intervals and p-values for each covariate.
