Scholarlink

<p>Brain tumor segmentation is critical in medical image analysis, aiding in early diagnosis and treatment planning. This study proposes an optimized deep learning approach for automated brain tumor segmentation using the BRATS 2016 and 2017 datasets. We utilized FLAIR, T1w, T1gd, and T2w MRI modalities, which underwent preprocessing and conversion from .nii to PNG format for model input. Our segmentation framework is based on MiniSegNet, FCN, and U-Net, with both original and compressed versions. The results show that MiniSegNet achieved 97.8% accuracy (compressed: 97.67%), FCN achieved 97.37% (compressed: 96.47%), and U-Net achieved 97.37% (compressed: 97.26%). The compressed models significantly reduced computational costs while maintaining comparable accuracy. In this paper, Particle Swarm Optimization (PSO) is utilized to optimize the hyperparameters of MiniSegNet, U-Net, and FCN models, enhancing their performance for brain tumor segmentation. The proposed approach enhances real-time clinical applicability by enabling integration into small IoT-based medical devices.</p>

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

Optimized brain tumor segmentation using MiniSegNet, U-Net, and FCN with compressed learning techniques

  • 期刊论文

  • 发表日期：

    2025年11月11日

查看全文
  • Arun Kumar,
  • Mohit Agarwal,
  • Mohd Aquib

摘要

Brain tumor segmentation is critical in medical image analysis, aiding in early diagnosis and treatment planning. This study proposes an optimized deep learning approach for automated brain tumor segmentation using the BRATS 2016 and 2017 datasets. We utilized FLAIR, T1w, T1gd, and T2w MRI modalities, which underwent preprocessing and conversion from .nii to PNG format for model input. Our segmentation framework is based on MiniSegNet, FCN, and U-Net, with both original and compressed versions. The results show that MiniSegNet achieved 97.8% accuracy (compressed: 97.67%), FCN achieved 97.37% (compressed: 96.47%), and U-Net achieved 97.37% (compressed: 97.26%). The compressed models significantly reduced computational costs while maintaining comparable accuracy. In this paper, Particle Swarm Optimization (PSO) is utilized to optimize the hyperparameters of MiniSegNet, U-Net, and FCN models, enhancing their performance for brain tumor segmentation. The proposed approach enhances real-time clinical applicability by enabling integration into small IoT-based medical devices.