Scholarlink

<p>The accurate and timely detection of lung cancer significantly enhances patient survival outcomes given its status as one of the most deadly cancers. This study proposes a Clustering-based Capsule Network with Stacked Heterogeneous Ensemble Learning as a new diagnostic framework that improves lung tumor detection capabilities by addressing existing model limitations. K-means clustering works within the framework to choose features effectively along with Capsule Networks which analyze spatial image relationships. The model uses L2 regularization as a method to decrease overfitting issues and improve predictive accuracy. A meta-learner completes final prediction refinement in Stacked Heterogeneous Ensemble Learning by using a Support Vector Machine (SVM), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost) as component classifiers to optimize classification precision. Data alteration strategies added with optimized parameter adjustment that helps in strengthening the proposed system design. Experimental results indicate CCN-SHEL outperforms existing approaches, achieving 98.52% accuracy, and 98.19% precision together with 97.81% specificity and 98.14% F1-score and 98.1% recall. These outputs underscore the model’s effectiveness in improving diagnostic accuracy and reliability for lung cancer thus advancing AI-based medical imaging technology systems.</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

Capsule network-driven feature extraction and ensemble learning for robust lung tumor classification

  • 期刊论文

  • 发表日期：

    2025年8月9日

查看全文
  • V. Nisha Jenipher,
  • S. Radhika

摘要

The accurate and timely detection of lung cancer significantly enhances patient survival outcomes given its status as one of the most deadly cancers. This study proposes a Clustering-based Capsule Network with Stacked Heterogeneous Ensemble Learning as a new diagnostic framework that improves lung tumor detection capabilities by addressing existing model limitations. K-means clustering works within the framework to choose features effectively along with Capsule Networks which analyze spatial image relationships. The model uses L2 regularization as a method to decrease overfitting issues and improve predictive accuracy. A meta-learner completes final prediction refinement in Stacked Heterogeneous Ensemble Learning by using a Support Vector Machine (SVM), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost) as component classifiers to optimize classification precision. Data alteration strategies added with optimized parameter adjustment that helps in strengthening the proposed system design. Experimental results indicate CCN-SHEL outperforms existing approaches, achieving 98.52% accuracy, and 98.19% precision together with 97.81% specificity and 98.14% F1-score and 98.1% recall. These outputs underscore the model’s effectiveness in improving diagnostic accuracy and reliability for lung cancer thus advancing AI-based medical imaging technology systems.