Scholarlink

<p>Colorectal cancer is one of the most prevalent cancers in the world. It illustrates the effectiveness of early detection and treatment of precursor polyps to prevent progression to malignancy. Despite the pivotal role of colonoscopy in detecting colorectal cancer and polyps, its efficacy has been marred by high miss rates attributed to heterogeneity in polyps and observer variability. Recent advancements in deep learning have significantly improved the automation of polyp detection and segmentation systems. In this study, we introduce the Pyramid Vision Transformer Adapter Residual Network (PVTAdpNet) to enhance existing models. As such, PVTAdpNet also presents an encoder-decoder architecture with additional upsampling layers by fusing some principles from the Pyramid Vision Transformer model with a novel residual block and adapter base skip connection. The lightweight design of PVTAdapNet enables real-time inference, making it suitable for clinical integration. This research contributes to developing advanced computer-aided diagnosis systems for improved polyp detection and early cancer diagnosis. PVTAdpNet obtains a high Dice coefficient of 0.8851 and a mean Intersection over Union of 0.8167 on out-of-distribution polyp datasets. Evaluation of the PolypGen dataset demonstrates PVTAdpNet's capability for real-time, accurate performance within familiar distributions. The source code of our network is available at <a href="https://github.com/ayousefinejad/PVTAdpNet.git">https://github.com/ayousefinejad/PVTAdpNet.git</a>.</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

PVTAdpNet: polyp segmentation using pyramid vision transformer with a novel adapter block

  • 期刊论文

  • 发表日期：

    2025年2月20日

查看全文
  • Arshia Yousefi Nezhad,
  • Helia Aghaei,
  • Hedieh Sajedi

摘要

Colorectal cancer is one of the most prevalent cancers in the world. It illustrates the effectiveness of early detection and treatment of precursor polyps to prevent progression to malignancy. Despite the pivotal role of colonoscopy in detecting colorectal cancer and polyps, its efficacy has been marred by high miss rates attributed to heterogeneity in polyps and observer variability. Recent advancements in deep learning have significantly improved the automation of polyp detection and segmentation systems. In this study, we introduce the Pyramid Vision Transformer Adapter Residual Network (PVTAdpNet) to enhance existing models. As such, PVTAdpNet also presents an encoder-decoder architecture with additional upsampling layers by fusing some principles from the Pyramid Vision Transformer model with a novel residual block and adapter base skip connection. The lightweight design of PVTAdapNet enables real-time inference, making it suitable for clinical integration. This research contributes to developing advanced computer-aided diagnosis systems for improved polyp detection and early cancer diagnosis. PVTAdpNet obtains a high Dice coefficient of 0.8851 and a mean Intersection over Union of 0.8167 on out-of-distribution polyp datasets. Evaluation of the PolypGen dataset demonstrates PVTAdpNet's capability for real-time, accurate performance within familiar distributions. The source code of our network is available at https://github.com/ayousefinejad/PVTAdpNet.git.