GDNet: A Robust 2.5D Multimodal MRI Brain Tumor Segmentation Framework with EMA Stabilization and Tumor-Aware Sampling.

Overview

1. Department of Computer Science, School of Engineering, Central Asian University, Tashkent 111211, Uzbekistan
2. Faculty of Data Science and Information Technology, INTI International University, Nilai 71800, Malaysia

Abstract

Accurate, automated delineation of adult diffuse gliomas from multi-parametric magnetic resonance imaging (mpMRI) is central to quantitative neuro-oncology. 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. We introduce GDNet, a 2.5D multimodal MRI segmentation framework for adult glioma evaluated on the BraTS 2024 cohort. GDNet consumes a stack of three adjacent axial slices from the four standard BraTS modalities (T1, T1ce, T2, FLAIR) as a 12-channel input to a compact U-shaped encoder–decoder with Group Normalization and predicts whole tumor (WT), tumor core (TC), and enhancing tumor (ET) masks for the central slice. The training pipeline pairs the 2.5D backbone with: (i) Exponential Moving Average (EMA) of model weights with decay 0.999, (ii) mixed tumor-aware slice sampling (p_tumor = 0.50), (iii) a compound Cross-Entropy + Soft-Dice loss, and (iv) AdamW with warm-up plus cosine annealing under Automatic Mixed Precision. We performed a systematic, step-by-step ablation covering a 2D baseline, EMA + mixed sampling, tumor-centered crop fine-tuning, a GDNet-inspired architectural integration, a region-aware loss, 3-slice and 5-slice 2.5D inputs, and connected-component post-processing, and we report multi-seed results to quantify reproducibility. On the held-out BraTS 2024 test partition, the final 3-slice 2.5D GDNet achieved positive-only Dice scores of 0.791 ± 0.000 (WT), 0.736 ± 0.003 (TC), 0.654 ± 0.004 (ET), and a mean foreground positive-only Dice of 0.820 ± 0.000 across seeds; the all-slice mean foreground Dice exceeded 0.927 ± 0.000. Validation positive-only scores were 0.805 ± 0.002 (WT), 0.757 ± 0.004 (TC), 0.683 ± 0.009 (ET). The inter-seed standard deviation was small for every region (≤0.01 Dice points), indicating low inter-seed variance across the two seeds evaluated; with only two seeds, we regard this as preliminary evidence of training stability rather than a strong reproducibility claim. The ablation isolated EMA + mixed tumor sampling and the 2.5D context window as the dominant sources of improvement; notably, a GDNet-style architectural integration with a region-aware loss did not outperform the simpler 2.5D U-Net on positive-only WT/TC/ET, and light post-processing improved only all-slice Dice. A failure-mode audit found that the residual catastrophic predictions are concentrated on a small minority of diffuse, infiltrative tumors with mass effect. Conclusions: Carefully engineered training strategies, tumor-aware sampling, EMA 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.

Reproduced under the paper's license (CC BY), from the paper cited above.

Code

The paper links to its data, not to its authors' code: see the Data section.

Tracing map

A tracing map links a paper to the code its authors published: this paper has none, so it has no map.

Data

Datasets cited

• synapse.org/brats, at Synapse; found in “Data Availability Statement”

Data Availability Statement

The BraTS 2024 multi-parametric MRI dataset analyzed in this study is publicly available through the BraTS challenge organizing committee at https://www.synapse.org/brats (accessed on 23 November 2025), subject to the corresponding Data Use Agreement.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 3 authors, 8 keywords, 18 references.

Cite

This paper

Kiani Kalejahi, B., Khan, S., & Rajabi, M. J. (2026). GDNet: A Robust 2.5D Multimodal MRI Brain Tumor Segmentation Framework with EMA Stabilization and Tumor-Aware Sampling. Journal of imaging, 12(7), 288. https://doi.org/10.3390/jimaging12070288

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.3390/diagnostics16172806

Brain Tumor Segmentation and Grading on MRI Using Deep Learning: A Systematic Literature Review and Benchmark-Driven Comparative Analysis.

Journal: Diagnostics (Basel, Switzerland)

In common: methods / tools, structural MRI / diffusion, other condition, 6 references

[2] doi:10.3390/brainsci16050468

Advancing Brain Tumor Diagnosis Using Deep Learning: A Systematic and Critical Review on Methodological Approaches to Glioma Segmentation and Classification Through Multiparametric MRI.

Journal: Brain sciences

In common: structural MRI / diffusion, other condition, 5 references

[3] doi:10.1038/s41598-026-48496-1 [code]

A unified FLAIR hyperintensity segmentation model for various CNS tumor types and acquisition time points.

Journal: Scientific reports

In common: methods / tools, structural MRI / diffusion, other condition, 4 references

[4] doi:10.3390/s26165021

Automatic Segmentation of Ischaemic Stroke Lesions Using Transformers and Convolutional Neural Networks Applied to Multimodal Neuroimaging.

Journal: Sensors (Basel, Switzerland)

In common: methods / tools, 4 references

[5] doi:10.3389/fonc.2026.1928589

LiteFreqMamba:lightweight frequency-enhanced Mamba for efficient 3D brain tumor segmentation.

Journal: Frontiers in oncology

In common: methods / tools, other condition, 4 references

[6] doi:10.3389/fonc.2026.1834400

BrainFusionNet: an attention-augmented deep convolutional framework with hybrid loss optimisation and test-time augmentation for multi-class brain tumour detection in magnetic resonance images.

Journal: Frontiers in oncology

In common: methods / tools, structural MRI / diffusion, other condition, 3 references

[7] doi:10.3390/s26154963

NeuroSPFNet: Vision-Language Semantic Prior-Guided Frequency-Enhanced Network for Brain Tumor Segmentation in Multimodal MRI.

Journal: Sensors (Basel, Switzerland)

In common: methods / tools, structural MRI / diffusion, other condition, 3 references

[8] doi:10.1371/journal.pone.0351405 [code]

Unimodal vs. multimodal deep learning for non-invasive MGMT promoter methylation prediction in glioblastoma: A systematic evaluation on the BraTS 2021 dataset.

Journal: PloS one

In common: methods / tools, structural MRI / diffusion, other condition, 3 references

[9] doi:10.1038/s41598-026-60525-7

Superfast prediction of brain tumours through adaptive incremental pre-training and re-training through transfer learning using state-of-the-art ResNet18 architectural model.

Journal: Scientific reports

In common: structural MRI / diffusion, other condition, 3 references

[10] doi:10.21037/qims-2026-0792 [code]

An nnU-Net-based framework with adaptive feature representation for 3D brain tumor segmentation.

Journal: Quantitative imaging in medicine and surgery

In common: methods / tools, structural MRI / diffusion, other condition, 3 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.