Scholarlink

Purpose <p>Ovarian cancer (OC) remains one of the most fatal malignancies among women worldwide, with epithelial ovarian cancer (EOC) being particularly difficult to diagnose due to complex tumor morphology, low tissue contrast in histopathology images, and noise corruption. Conventional diagnostic approaches suffer from poor feature representation, inaccurate segmentation, and weak generalization, leading to reduced classification accuracy.</p> Methods <p>We propose an adaptive contextual propagation deep graph neural network optimized with the migrating walrus algorithm (ACPDGNNet-MWA). The pipeline integrates multiple stages: (i) image enhancement through an observability-constrained resampling-free cubature kalman filter (O-CRCKF) for noise reduction, (ii) accurate segmentation via a prompt-tuned multi-task taxonomic transformer (PTMT-TT), (iii) robust feature extraction using enhanced VGG19 (E-VGG19), and (iv) efficient classification through ACPDGNNet, which combines a contextual attention network (CAN) with an adaptive propagation deep graph neural network (AP-DGNN). Model parameters are further optimized with MWA to enhance convergence and generalization.</p> Results <p>Extensive experiments on the EOC histopathology dataset demonstrated the superiority of the proposed method, achieving 99.9% accuracy, 99.8% precision, 99.7% sensitivity, and a dice similarity coefficient (DSC) of 99.6%, surpassing state-of-the-art benchmarks.</p> Conclusion <p>The ACPDGNNet-MWA framework provides a highly accurate and computationally efficient approach for EOC detection, offering substantial potential for clinical decision support in ovarian cancer diagnosis.</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

Ovarian Cancer Detection Using Adaptive Contextual Propagation Deep Graph Neural Network with Migrating Walrus Algorithm

  • 期刊论文

  • 发表日期：

    2025年11月2日

查看全文
  • B. R. Tapas Bapu,
  • R. Anitha,
  • K. M. Dhanalakshmi,
  • S. Sridhar

摘要

Purpose

Ovarian cancer (OC) remains one of the most fatal malignancies among women worldwide, with epithelial ovarian cancer (EOC) being particularly difficult to diagnose due to complex tumor morphology, low tissue contrast in histopathology images, and noise corruption. Conventional diagnostic approaches suffer from poor feature representation, inaccurate segmentation, and weak generalization, leading to reduced classification accuracy.

Methods

We propose an adaptive contextual propagation deep graph neural network optimized with the migrating walrus algorithm (ACPDGNNet-MWA). The pipeline integrates multiple stages: (i) image enhancement through an observability-constrained resampling-free cubature kalman filter (O-CRCKF) for noise reduction, (ii) accurate segmentation via a prompt-tuned multi-task taxonomic transformer (PTMT-TT), (iii) robust feature extraction using enhanced VGG19 (E-VGG19), and (iv) efficient classification through ACPDGNNet, which combines a contextual attention network (CAN) with an adaptive propagation deep graph neural network (AP-DGNN). Model parameters are further optimized with MWA to enhance convergence and generalization.

Results

Extensive experiments on the EOC histopathology dataset demonstrated the superiority of the proposed method, achieving 99.9% accuracy, 99.8% precision, 99.7% sensitivity, and a dice similarity coefficient (DSC) of 99.6%, surpassing state-of-the-art benchmarks.

Conclusion

The ACPDGNNet-MWA framework provides a highly accurate and computationally efficient approach for EOC detection, offering substantial potential for clinical decision support in ovarian cancer diagnosis.