Scholarlink

This paper proposes an ensemble learning method for 2D brain tumor segmentation utilizing multi-modal MRI data from the BraTS 2021 dataset. The proposed framework integrates three distinct architectures: the standard U-Net, a conventional convolutional neural network (CNN), and an Attention U-Net. Each model is trained independently on four MRI modalities—T1, T2,T1ce and FLAIR following extensive preprocessing and slice-wise segmentation. Experimental results indicate that the Attention U-Net achieves the highest individual performance, with a Dice similarity coefficient of 0.95 and an Intersection over Union (IoU) of 0.90. The ensemble model further enhances segmentation accuracy by leveraging the complementary strengths of individual networks. This demonstrates the effectiveness of architectural diversity in improving the delineation of tumor boundaries in 2D MRI scans.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Ensemble Learning for 2D Brain Tumor Segmentation Using Multi-modal MRI Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Pratik Raj Gurung,
  • Bijoyeta Roy

摘要

This paper proposes an ensemble learning method for 2D brain tumor segmentation utilizing multi-modal MRI data from the BraTS 2021 dataset. The proposed framework integrates three distinct architectures: the standard U-Net, a conventional convolutional neural network (CNN), and an Attention U-Net. Each model is trained independently on four MRI modalities—T1, T2,T1ce and FLAIR following extensive preprocessing and slice-wise segmentation. Experimental results indicate that the Attention U-Net achieves the highest individual performance, with a Dice similarity coefficient of 0.95 and an Intersection over Union (IoU) of 0.90. The ensemble model further enhances segmentation accuracy by leveraging the complementary strengths of individual networks. This demonstrates the effectiveness of architectural diversity in improving the delineation of tumor boundaries in 2D MRI scans.