Scholarlink

<p>Recent progresses in deep learning greatly enhanced diagnostic performance, especially in the prediction of heart disease. In this paper, we present a new hybrid model based on CNNs and GANs to improve predictive performance with imbalanced and scarce medical datasets. CNN module is good at capturing useful features that are indispensible to the clinical data, and the proposed GAN can produce realistic synthetic samples, which can enhance the data set to fortify the model’s robustness. In comparison with conventional machine learning models such as SVM, Naive Bayes, and Decision Trees, also neural network models such as LSTM, ANN, CNN-GAN hybrid model provides better results i.e. 92% accuracy, 91.5% precision, 92.5% recall, and 92% F1-score. Furthermore, the architecture of the model is compatible with Internet-Of-Things (IoT) devices, which allows the real-time integration of IoT devices with the proposed model in order to monitor health continuously and thereby in the very early stage. These results demonstrate the potential of hybrid deep learning models in the development of cardiovascular diagnostics.</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

Enhancing Heart Disease Prediction Through a CNN-GAN Hybrid Deep Learning Model

  • 期刊论文

  • 发表日期：

    2025年12月26日

查看全文
  • Lam Rathnakumari,
  • Ganga Rama Koteswara Rao

摘要

Recent progresses in deep learning greatly enhanced diagnostic performance, especially in the prediction of heart disease. In this paper, we present a new hybrid model based on CNNs and GANs to improve predictive performance with imbalanced and scarce medical datasets. CNN module is good at capturing useful features that are indispensible to the clinical data, and the proposed GAN can produce realistic synthetic samples, which can enhance the data set to fortify the model’s robustness. In comparison with conventional machine learning models such as SVM, Naive Bayes, and Decision Trees, also neural network models such as LSTM, ANN, CNN-GAN hybrid model provides better results i.e. 92% accuracy, 91.5% precision, 92.5% recall, and 92% F1-score. Furthermore, the architecture of the model is compatible with Internet-Of-Things (IoT) devices, which allows the real-time integration of IoT devices with the proposed model in order to monitor health continuously and thereby in the very early stage. These results demonstrate the potential of hybrid deep learning models in the development of cardiovascular diagnostics.