Scholarlink

<p>Early detection of lung cancer significantly enhances patient survival rates, combating one of the foremost causes of death globally. Nonetheless, early detection of the disease is still difficult as lung cancer is asymptomatic in its early stages and conventional diagnostic techniques have limitations, including a low sensitivity in detecting small tumors, and the inability to distinguish between benign and malignant lesions efficiently. This research devises a new scheme for detecting lung cancer by exploiting Computed Tomography (CT) images. Firstly, the input images are gathered and then preprocessed employing a median filter. Further, the lung lobe is segmented exploiting the Modified U-Net framework. Once the segmentation process is accomplished, feature extraction is conducted using Maximally Stable Extremal Regions (MSER), Discrete Wavelet Transform (DWT), and geometric features. Lastly, lung cancer detection is done by using Dense residual forward Taylor network (DenResFT-Net). This technique integrates DenseNet, Deep Residual Network (DRN), and Taylor Series. Further, the experimental results validate that DenResFT-Net attains high performance with an accuracy, True Positive Rate (TPR) of 92.755%, precision, F1-score of 91.877%, 91.334%, 92.039%, and measured a minimal False Positive Rate (FPR) of 7.433%.</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

DenResFT-Net: Dense Residual Forward Taylor Network-Based Lung Cancer Detection Using CT Image

  • 期刊论文

  • 发表日期：

    2026年2月23日

查看全文
  • Karthika K,
  • Antonidoss A

摘要

Early detection of lung cancer significantly enhances patient survival rates, combating one of the foremost causes of death globally. Nonetheless, early detection of the disease is still difficult as lung cancer is asymptomatic in its early stages and conventional diagnostic techniques have limitations, including a low sensitivity in detecting small tumors, and the inability to distinguish between benign and malignant lesions efficiently. This research devises a new scheme for detecting lung cancer by exploiting Computed Tomography (CT) images. Firstly, the input images are gathered and then preprocessed employing a median filter. Further, the lung lobe is segmented exploiting the Modified U-Net framework. Once the segmentation process is accomplished, feature extraction is conducted using Maximally Stable Extremal Regions (MSER), Discrete Wavelet Transform (DWT), and geometric features. Lastly, lung cancer detection is done by using Dense residual forward Taylor network (DenResFT-Net). This technique integrates DenseNet, Deep Residual Network (DRN), and Taylor Series. Further, the experimental results validate that DenResFT-Net attains high performance with an accuracy, True Positive Rate (TPR) of 92.755%, precision, F1-score of 91.877%, 91.334%, 92.039%, and measured a minimal False Positive Rate (FPR) of 7.433%.