Scholarlink

<p>The high mortality rate of lung cancer worldwide is largely attributed to its late-stage diagnosis. Assessing lung cancer severity promptly plays a vital role in guiding treatment decisions and enhancing patient survival rates. This study introduces an Adaptive Neuro-Fuzzy SqueezeNet (ANFSq-Net) framework for lung cancer detection and severity level classification using Computed Tomography (CT) images. The proposed method integrates the strengths of the Adaptive Neuro-Fuzzy Inference System (ANFIS) and SqueezeNet, supported by fractional calculus (FC) regression, to enhance feature adaptability, reduce computational complexity, and improve learning efficiency. To preprocess the CT images, noise is removed with a bilateral filter, followed by lung lobe segmentation through a Conditional Generative Adversarial Network (CGAN). Key features such as Local Ternary Pattern (LTP), homogeneity, and entropy are extracted for accurate classification of lung cancer into mild, moderate, and severe levels. The ANFSq-Net model achieves excellent performance metrics, including 94.876% accuracy, 96.877% sensitivity, 92.987% specificity, 93.988% precision, and a 95.411% F1-score. It is concluded that the combined application of ANFIS, SqueezeNet, and FC regression enables precise and efficient classification performance.</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

ANFSq-Net: Adaptive neuro fuzzy inference system for lung cancer detection and severity level classification using CT image

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • Sudhakar Raju,
  • Peddireddy Veera Venkateswara Rao

摘要

The high mortality rate of lung cancer worldwide is largely attributed to its late-stage diagnosis. Assessing lung cancer severity promptly plays a vital role in guiding treatment decisions and enhancing patient survival rates. This study introduces an Adaptive Neuro-Fuzzy SqueezeNet (ANFSq-Net) framework for lung cancer detection and severity level classification using Computed Tomography (CT) images. The proposed method integrates the strengths of the Adaptive Neuro-Fuzzy Inference System (ANFIS) and SqueezeNet, supported by fractional calculus (FC) regression, to enhance feature adaptability, reduce computational complexity, and improve learning efficiency. To preprocess the CT images, noise is removed with a bilateral filter, followed by lung lobe segmentation through a Conditional Generative Adversarial Network (CGAN). Key features such as Local Ternary Pattern (LTP), homogeneity, and entropy are extracted for accurate classification of lung cancer into mild, moderate, and severe levels. The ANFSq-Net model achieves excellent performance metrics, including 94.876% accuracy, 96.877% sensitivity, 92.987% specificity, 93.988% precision, and a 95.411% F1-score. It is concluded that the combined application of ANFIS, SqueezeNet, and FC regression enables precise and efficient classification performance.