Fig. 3: Performance comparison.

Comparative performance of HELMET models and EDEMA baseline models across all cohorts and prediction tasks. Each radial axis represents a distinct performance metric, with higher performance values further from center. Polyhedrons with greater total area on the plots show higher-performing models, and smaller polyhedrons show lower-performing models. Blue lines represent the overall performance for the HELMET models, green lines represent the filtered performance for the HELMET models, red lines represent the overall performance for the baseline EDEMA models, and orange lines represent the filtered performance for the baseline EDEMA models. Specifically, we summarize model performance across tasks and cohorts in the following panels: a derivation cohort for the 8-hr prediction task; b derivation cohort for the 24-hr prediction task; c external validation cohort for the 8-hr prediction task; d external validation cohort for the 24-hr task.