Scholarlink

<p>Coronary artery disease (CAD) has emerged as a leading cause of death globally in recent years. Accurate segmentation of coronary arteries is crucial for the clinical diagnosis and treatment of CAD, particularly in detecting stenosis and plaques. Deep learning techniques have proven valuable in assisting medical professionals with disease diagnosis through biomedical imaging. The existing methods leverage 3D UNet and its variants focusing on global processing of input image. These networks capture only global features without considering local details. This research investigates the use of the 3D VNet model to segment coronary arteries from 3D Coronary Computed Tomography Angiography (CCTA) images. The performance of this model is evaluated using a Patch Based Method where the input 3D image is divided into patches to capture local details. Experimental results demonstrate a validation&#xa0;Dice Similarity Coefficient (DSC) of 0.70 and test DSC of 0.68&#xa0;with the Patch Based Method on a subset of 200 images from the ImageCAS dataset.</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

Patch based coronary artery segmentation using 3D VNet architecture

  • 期刊论文

  • 发表日期：

    2025年8月18日

查看全文
  • Kavita Chachadi,
  • S. R. Nirmala,
  • Sheetal Lamani

摘要

Coronary artery disease (CAD) has emerged as a leading cause of death globally in recent years. Accurate segmentation of coronary arteries is crucial for the clinical diagnosis and treatment of CAD, particularly in detecting stenosis and plaques. Deep learning techniques have proven valuable in assisting medical professionals with disease diagnosis through biomedical imaging. The existing methods leverage 3D UNet and its variants focusing on global processing of input image. These networks capture only global features without considering local details. This research investigates the use of the 3D VNet model to segment coronary arteries from 3D Coronary Computed Tomography Angiography (CCTA) images. The performance of this model is evaluated using a Patch Based Method where the input 3D image is divided into patches to capture local details. Experimental results demonstrate a validation Dice Similarity Coefficient (DSC) of 0.70 and test DSC of 0.68 with the Patch Based Method on a subset of 200 images from the ImageCAS dataset.