Scholarlink

<p>The medical diagnostics and treatment have not helped to remove breast cancer as a significant health issue in the world. Risk stratification and informed clinical decision-making depend on accurate prediction of breast cancer outcomes. In this study, we propose a deep learning-based outcome prediction framework called as a Random Optimization based Fully Connected Neural Network (RObFCNN) for breast cancer outcome prediction. The proposed approach integrates a random optimization strategy for feature subset selection with a fully connected neural network, enabling effective modeling of complex and nonlinear relationships within high-dimensional clinical and genomic data.</p><p>The publicly available METABRIC dataset was used to conduct experiments. Proposed RObFCNN framework’s performance was compared to the widely used machine learning models such as standard neural networks, Random Forests, Decision Trees, and Extreme Gradient Boosting. In the framework of this experiment, the accuracy of the proposed framework reached 95.28%, sensitivity 94.56%, specificity 95.73%, precision 93.29%, and Matthews Correlation Coefficient (MCC) 0.9006, which is better than the predictive performance of the considered baseline methods. The findings indicate that the RObFCNN framework is an effective research-level framework that can be used to model breast cancer prognostics based on optimized feature selection and deep learning. This should be further validated on independent datasets and a survival specific modeling methodology should be included to determine its generalizability and potential clinical applicability.</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

Random optimization-based feature selection for breast cancer outcome prediction using fully connected neural networks on multi-omic data

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月2日

查看全文
  • Poi Tamrakar,
  • Ganesh R. Pathak

摘要

The medical diagnostics and treatment have not helped to remove breast cancer as a significant health issue in the world. Risk stratification and informed clinical decision-making depend on accurate prediction of breast cancer outcomes. In this study, we propose a deep learning-based outcome prediction framework called as a Random Optimization based Fully Connected Neural Network (RObFCNN) for breast cancer outcome prediction. The proposed approach integrates a random optimization strategy for feature subset selection with a fully connected neural network, enabling effective modeling of complex and nonlinear relationships within high-dimensional clinical and genomic data.

The publicly available METABRIC dataset was used to conduct experiments. Proposed RObFCNN framework’s performance was compared to the widely used machine learning models such as standard neural networks, Random Forests, Decision Trees, and Extreme Gradient Boosting. In the framework of this experiment, the accuracy of the proposed framework reached 95.28%, sensitivity 94.56%, specificity 95.73%, precision 93.29%, and Matthews Correlation Coefficient (MCC) 0.9006, which is better than the predictive performance of the considered baseline methods. The findings indicate that the RObFCNN framework is an effective research-level framework that can be used to model breast cancer prognostics based on optimized feature selection and deep learning. This should be further validated on independent datasets and a survival specific modeling methodology should be included to determine its generalizability and potential clinical applicability.