Scholarlink

<p>Brain tumor segmentation is the process of precisely finding and outlining tumor areas within the brain and is commonly carried out using Magnetic Resonance Imaging (MRI) images. The differences in tumor appearance, noise, and irregularities between MRI scans make precise segmentation challenging. Enhancing the diagnosis accuracy requires the development of strong algorithms that can manage these variances and produce reliable outcomes. A new technique termed Serval Ablation Optimization_Fuzzy Attention-SegNet (SeAO_FA-SegNet) is devised in this work for this process employing MRI. The input MRI image is firstly obtained and is denoised employing Attention Guided Denoising CNN (ADNet) to remove inherent noise. Once denoised, segmentation is done with Fuzzy Attention-SegNet (FA-SegNet), which is trained by exploiting the devised Serval Ablation Optimization (SeAO). Further, FA-SegNet is employed with Log-Cosh Softmax Loss, which is the incorporation of Lovász-Softmax Loss and Log-Cosh Dice Loss. Also, SeAO is developed by merging the Serval Optimization Algorithm (SOA) and Snow Ablation Optimization (SAO). Here, the segmented tumor regions are finally categorized as whole tumor, tumor core, or enhancing tumor. The evaluation of the devised SeAO_FA-SegNet is performed using the BraTS 2019 and BraTS 2020 datasets. The SeAO_FA-SegNet model demonstrates high segmentation performance, with segmentation accuracy, dice coefficient, Intersection over Union (IoU), and Hausdorff Distance values of 92.192%, 92.846%, 92.265%, and 0.752&#xa0;mm, indicating reliable delineation of tumor regions across diverse MRI scans.</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

Fuzzy Attention SegNet Model with Serval Ablation Optimization and Log-Cosh Softmax Loss for Effective Brain Tumor Segmentation Using MRI Imaging

  • 期刊论文

  • 发表日期：

    2026年3月11日

查看全文
  • C. K. Jyothi,
  • Anupama S. Awati,
  • Dattaprasad A. Torse

摘要

Brain tumor segmentation is the process of precisely finding and outlining tumor areas within the brain and is commonly carried out using Magnetic Resonance Imaging (MRI) images. The differences in tumor appearance, noise, and irregularities between MRI scans make precise segmentation challenging. Enhancing the diagnosis accuracy requires the development of strong algorithms that can manage these variances and produce reliable outcomes. A new technique termed Serval Ablation Optimization_Fuzzy Attention-SegNet (SeAO_FA-SegNet) is devised in this work for this process employing MRI. The input MRI image is firstly obtained and is denoised employing Attention Guided Denoising CNN (ADNet) to remove inherent noise. Once denoised, segmentation is done with Fuzzy Attention-SegNet (FA-SegNet), which is trained by exploiting the devised Serval Ablation Optimization (SeAO). Further, FA-SegNet is employed with Log-Cosh Softmax Loss, which is the incorporation of Lovász-Softmax Loss and Log-Cosh Dice Loss. Also, SeAO is developed by merging the Serval Optimization Algorithm (SOA) and Snow Ablation Optimization (SAO). Here, the segmented tumor regions are finally categorized as whole tumor, tumor core, or enhancing tumor. The evaluation of the devised SeAO_FA-SegNet is performed using the BraTS 2019 and BraTS 2020 datasets. The SeAO_FA-SegNet model demonstrates high segmentation performance, with segmentation accuracy, dice coefficient, Intersection over Union (IoU), and Hausdorff Distance values of 92.192%, 92.846%, 92.265%, and 0.752 mm, indicating reliable delineation of tumor regions across diverse MRI scans.