Scholarlink

<p>Recent advancements in technology have significantly expanded the scope of image processing applications particularly in the field of medical imaging. Key processes in medical imaging such as segmentation, detection and classification have become pivotal in various medical applications. The dermoscopic images require precise segmentation for different medical purposes. Skin cancer especially melanoma poses a significant threat which kills an average of 60,000 individuals globally every year. Detecting skin cancer at an early stage is crucial for saving millions of lives worldwide. In this context a robust automatic skin segmentation model has been developed using deep learning techniques and enhanced with metaheuristic algorithms like ant colony optimization. The developed model is novel in its hybrid architecture combining UNet, ResNet and ant colony optimization algorithms. This integrated approach is designed to perform automatic skin lesion segmentation across multiple datasets offering accurate and reliable results that can assist medical experts in diagnosis and analysis. This innovative combination was applied to the openly available international skin imaging collaboration (ISIC) 2018 and 2017 datasets for segmentation purpose. The model's segmentation performance was thoroughly assessed using multiple segmentation metrics such as the accuracy, dice similarity coefficient and Jaccard index. The proposed ResUNet-52 with ant colony optimization model demonstrated remarkable results for the ISIC 2018 dataset achieving a dice score of 95.48%, an accuracy of 99.07% and a Jaccard Index of 91.56%. This breakthrough underscores the potential of advanced technologies in revolutionizing early skin cancer detection and ultimately saving lives on a global scale.</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

Deep learning based skin lesion detection in dermoscopic images enhanced by ant colony optimization

  • 期刊论文

  • 发表日期：

    2026年3月7日

查看全文
  • Yashaswini Gowda N,
  • Manjunath R V

摘要

Recent advancements in technology have significantly expanded the scope of image processing applications particularly in the field of medical imaging. Key processes in medical imaging such as segmentation, detection and classification have become pivotal in various medical applications. The dermoscopic images require precise segmentation for different medical purposes. Skin cancer especially melanoma poses a significant threat which kills an average of 60,000 individuals globally every year. Detecting skin cancer at an early stage is crucial for saving millions of lives worldwide. In this context a robust automatic skin segmentation model has been developed using deep learning techniques and enhanced with metaheuristic algorithms like ant colony optimization. The developed model is novel in its hybrid architecture combining UNet, ResNet and ant colony optimization algorithms. This integrated approach is designed to perform automatic skin lesion segmentation across multiple datasets offering accurate and reliable results that can assist medical experts in diagnosis and analysis. This innovative combination was applied to the openly available international skin imaging collaboration (ISIC) 2018 and 2017 datasets for segmentation purpose. The model's segmentation performance was thoroughly assessed using multiple segmentation metrics such as the accuracy, dice similarity coefficient and Jaccard index. The proposed ResUNet-52 with ant colony optimization model demonstrated remarkable results for the ISIC 2018 dataset achieving a dice score of 95.48%, an accuracy of 99.07% and a Jaccard Index of 91.56%. This breakthrough underscores the potential of advanced technologies in revolutionizing early skin cancer detection and ultimately saving lives on a global scale.