GDNet: A Robust 2.5D Multimodal MRI Brain Tumor Segmentation Framework with EMA Stabilization and Tumor-Aware Sampling.
A 2.5D deep-learning model segmented whole brain tumor on MRI with 0.791 Dice accuracy
GDNet: A Robust 2.5D Multimodal MRI Brain Tumor Segmentation Framework with EMA Stabilization and Tumor-Aware Sampling.
Accurate, automated delineation of adult diffuse gliomas from multi-parametric magnetic resonance imaging (mpMRI) is central to quantitative neuro-oncology.
We introduce GDNet, a 2.5D multimodal MRI segmentation framework for adult glioma evaluated on the BraTS 2024 cohort.
Automated whole-tumor outlines overlapped expert outlines fairly well (1.0 = perfect match)
Volumetric 3D networks dominate the BraTS leaderboard but require expensive GPUs, long training cycles, and provide diminishing returns relative to their compute budget.
Slice-wise 2D models, by contrast, discard inter-slice context that is informative for thin tumor rims and small enhancing foci.
Validation positive-only scores were 0.805 ± 0.002 (whole tumor), 0.757 ± 0.004 (tumor core), 0.683 ± 0.009 (enhancing tumor).
Carefully engineered training strategies, tumor-aware sampling, Exponential Moving Average stabilization, and a modest 2.5D context window recover a substantial fraction of the accuracy of much heavier 3D networks at a fraction of the compute, are reproducible across seeds, and outperform a heavier GDNet-inspired architectural variant on the same data.
GDNet is therefore a practical and, pending external validation, potentially clinically deployable framework for multimodal glioma segmentation on workstation-class GPU hardware.