Scholarlink

From past few years, artificial intelligence (AI) had significantly advanced in the field of computer vision due to the growing volume and unique characteristics of medical image data. The existing approaches for image segmentation for medical image analysis had faced significant challenges which include complex anatomical structures and generalizability issues. Therefore, this research proposes Multi-Scale Feature Fusion Semantic Segmentation Network (MSSNet) for image segmentation for medical image analysis. Initially, the data collected from International Symposium on Biomedical Imaging (ISBI) challenge dataset and preprocessed by using data augmentation, quantum annealing-based denoising (QAD) and histogram equalization. Then, this preprocessed data is used for extracting features like textures, edges and structures by using quantum convolutional neural network (QCNN). Finally, the segmentation is done by using proposed MSSNet which effectively segments the medical images. The proposed MSSNet achieved better results in terms of accuracy (0.9965), dice (0.8734), and intersection over union (IoU) (0.9748) when compared with existing Universal Segmentation Transformer with Pre-training (USegTranformer-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

Multi-scale Feature Fusion Semantic Segmentation Network-Based Enhanced Image Segmentation for Analysis of Medical Images

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Haayder M. Abbas,
  • K. Aruna Kumari

摘要

From past few years, artificial intelligence (AI) had significantly advanced in the field of computer vision due to the growing volume and unique characteristics of medical image data. The existing approaches for image segmentation for medical image analysis had faced significant challenges which include complex anatomical structures and generalizability issues. Therefore, this research proposes Multi-Scale Feature Fusion Semantic Segmentation Network (MSSNet) for image segmentation for medical image analysis. Initially, the data collected from International Symposium on Biomedical Imaging (ISBI) challenge dataset and preprocessed by using data augmentation, quantum annealing-based denoising (QAD) and histogram equalization. Then, this preprocessed data is used for extracting features like textures, edges and structures by using quantum convolutional neural network (QCNN). Finally, the segmentation is done by using proposed MSSNet which effectively segments the medical images. The proposed MSSNet achieved better results in terms of accuracy (0.9965), dice (0.8734), and intersection over union (IoU) (0.9748) when compared with existing Universal Segmentation Transformer with Pre-training (USegTranformer-P).