DPEA-Net: Glioma Segmentation Results on BraTS 2019 and 2020 MRI

DPEA-Net, a lightweight 3D CNN, achieved mean Dice scores of 90.43%/89.96% for whole tumor, 85.56%/86.52% for tumor core, and 81.89%/80.31% for enhancing tumor on BraTS validation sets. The model, proposed by Caijian Hua, uses a computationally efficient architecture requiring only 17.48 GFLOPs. The results are documented in a PDF file last updated on 2026-05-25.

Use Cases

• Benchmarking new segmentation models based on the reported Dice scores for glioma subregions
• Developing lightweight clinical AI tools based on the described 3D CNN architecture for multiparametric MRI
• Refining tumor boundary segmentation based on the described attention mechanisms for handling ambiguous boundaries
• Planning radiotherapy treatments based on automated delineation of whole tumor, tumor core, and enhancing tumor

Strengths

• Performance metrics are explicitly quantified with Dice scores for three tumor subregions on two validation sets
• The model's computational efficiency is specified as 17.48 GFLOPs, enabling sub-2-second inference
• The document is openly licensed under CC-BY-4.0

Limitations

• The dataset is a 241.5 KB PDF containing results, not the underlying MRI image data or model code
• Description metadata is limited; actual data quality requires manual inspection after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

figshare

Collection Method

Results from evaluating the proposed DPEA-Net model on the BraTS 2019 and 2020 validation sets.

Freshness

Last updated 2026-05-25 08:39:18; freshness should be verified