Scholarlink

<p>Accurate brain tumour segmentation from Magnetic Resonance Imaging (MRI) remains a challenging task due to noise, intensity inhomogeneity, and complex tumour morphology. Although deep learning models such as U-Net have shown promising results, existing approaches often lack optimization-driven preprocessing and robustness validation, while many hybrid methods remain pipeline-based without a unified mathematical framework. This paper proposes an optimization-driven integrated framework combining an Adaptive Cuckoo Optimization Algorithm (ACOA), Weighted Singular Value Decomposition (WSVD), and an enhanced U-Net architecture. The method formulates segmentation as a multi-objective optimization problem that maximizes accuracy while minimizing boundary error and noise sensitivity. ACOA improves convergence through adaptive search, while WSVD enhances feature representation. A composite loss function further improves segmentation under class imbalance. Experimental results on the BRATS dataset demonstrate superior performance, achieving a Dice score of 0.94, IoU of 0.89, and accuracy above 98%, with statistically significant improvements (p &lt; 0.01). Robustness and cross-dataset evaluations confirm strong generalization. The proposed framework provides an accurate and reliable solution for automated brain tumour segmentation.</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

Integrated Brain Tumour Detection: Cuckoo Optimization and SVD- Guided Segmentation with U-Net

  • 期刊论文

  • 发表日期：

    2026年5月30日

查看全文
  • Ajay Singh Yadav,
  • Navin Ahlawat,
  • Bhavani Viswanathan,
  • Garima Pandey,
  • Anupam Swami,
  • Anjali Malik,
  • Priyanka Agarwal,
  • M. Vinothkumar

摘要

Accurate brain tumour segmentation from Magnetic Resonance Imaging (MRI) remains a challenging task due to noise, intensity inhomogeneity, and complex tumour morphology. Although deep learning models such as U-Net have shown promising results, existing approaches often lack optimization-driven preprocessing and robustness validation, while many hybrid methods remain pipeline-based without a unified mathematical framework. This paper proposes an optimization-driven integrated framework combining an Adaptive Cuckoo Optimization Algorithm (ACOA), Weighted Singular Value Decomposition (WSVD), and an enhanced U-Net architecture. The method formulates segmentation as a multi-objective optimization problem that maximizes accuracy while minimizing boundary error and noise sensitivity. ACOA improves convergence through adaptive search, while WSVD enhances feature representation. A composite loss function further improves segmentation under class imbalance. Experimental results on the BRATS dataset demonstrate superior performance, achieving a Dice score of 0.94, IoU of 0.89, and accuracy above 98%, with statistically significant improvements (p < 0.01). Robustness and cross-dataset evaluations confirm strong generalization. The proposed framework provides an accurate and reliable solution for automated brain tumour segmentation.