CU-Net: a U-Net architecture for efficient brain-tumor segmentation on BraTS 2019 dataset

TL;DR

The paper addresses the need for accurate brain-tumor segmentation in MRI by introducing CU-Net, a high‑resolution, symmetric U‑Net variant tailored for BraTS 2019 data. It details a contracting–bottleneck–expansive architecture with skip connections and upsampling to preserve boundary details, trained on multimodal MRI with binary masks. CU‑Net achieves a Dice score of 82.41% on BraTS 2019, outperforming Swin UNet and TransUNet in the reported comparisons, underscoring improved tumor delineation. The results have potential clinical impact for surgical planning and radiation therapy, and the discussion points toward incorporating self‑supervised learning to further enhance segmentation under limited labeled data scenarios.

Abstract

Accurately segmenting brain tumors from MRI scans is important for developing effective treatment plans and improving patient outcomes. This study introduces a new implementation of the Columbia-University-Net (CU-Net) architecture for brain tumor segmentation using the BraTS 2019 dataset. The CU-Net model has a symmetrical U-shaped structure and uses convolutional layers, max pooling, and upsampling operations to achieve high-resolution segmentation. Our CU-Net model achieved a Dice score of 82.41%, surpassing two other state-of-the-art models. This improvement in segmentation accuracy highlights the robustness and effectiveness of the model, which helps to accurately delineate tumor boundaries, which is crucial for surgical planning and radiation therapy, and ultimately has the potential to improve patient outcomes.

Figures (3)

• Figure 1: Sample BraTS 2019 images
• Figure 2: CU-Net Architecture
• Figure 3: Comparison of Mask Differences