Scholarlink

Blood clots within deep veins cause deep vein thrombosis (DVT), a serious medical condition that typically occurs in the lower extremities. Preventing potentially fatal consequences like pulmonary embolism requires early and precise diagnosis of DVT. In this research, a novel Deep Thrombosis Net that has been proposed to identify DVT in CT images. The process begins with preprocessing the input CT images using a contrast stitching adaptive Gaussian star filter to eliminate noise artifacts. Next, the denoised images undergo feature extraction utilizing dilated convolutional neural network (Dilated CNN) to capture essential features. The pelican algorithm is employed to select pertinent features from the CT images. Subsequently, a deep belief network (DBN) is utilized to classify cases into normal and DVT categories. The accuracy of the proposed Deep Thrombosis Net is 99.03%. The proposed Deep Thrombosis Net improves the overall accuracy of 4.33%, 26.48% better than CTPAs, and ResNet-101, respectively.

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

Detection of Deep Vein Thrombosis via Deep Learning-Based Dilated Pelican Convolutional Neural Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Bastin Rogers Cross Joseph,
  • Immanuel Johnraja Jebadurai,
  • Getzi Jeba Leelipushpam Paulraj,
  • Mulli Mary Varuvel,
  • Jebaveerasingh Jebadurai

摘要

Blood clots within deep veins cause deep vein thrombosis (DVT), a serious medical condition that typically occurs in the lower extremities. Preventing potentially fatal consequences like pulmonary embolism requires early and precise diagnosis of DVT. In this research, a novel Deep Thrombosis Net that has been proposed to identify DVT in CT images. The process begins with preprocessing the input CT images using a contrast stitching adaptive Gaussian star filter to eliminate noise artifacts. Next, the denoised images undergo feature extraction utilizing dilated convolutional neural network (Dilated CNN) to capture essential features. The pelican algorithm is employed to select pertinent features from the CT images. Subsequently, a deep belief network (DBN) is utilized to classify cases into normal and DVT categories. The accuracy of the proposed Deep Thrombosis Net is 99.03%. The proposed Deep Thrombosis Net improves the overall accuracy of 4.33%, 26.48% better than CTPAs, and ResNet-101, respectively.