Scholarlink

<p>Aortic aneurysm (AA) is still the predominant cause of global mortality and morbidity, making early and precise diagnosis necessary. In this work, a novel hybrid framework—Hybrid Attention-Augmented Deep Neural Network (HA-DNN) is proposed which is optimized using Ant Colony Optimization and Grey Wolf Optimizer (ACO-GWO) paradigm for precise diagnosis of AA. The proposed framework combines ACO to perform initial feature subset selection and GWO for fine-tuned optimisation to eliminate redundancy and enhance generalizability. The model has a two-branch structure: a deep feedforward network to learn structured clinical features and an attentional BiLSTM (Bi-directional long short term memory) network to learn time-series ECG (electrocardiograph) features. The two-branch integration makes it possible to learn robustly across heterogeneous data modalities. The proposed framework is tested on two open-access benchmark datasets: Cleveland Heart Disease Dataset and MIT-BIH Arrhythmia Dataset. The results show remarkable enhancements in terms of classification accuracy, F1-score, and generalizability when compared to existing methods.</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

Hybrid Attention-Augmented Deep Neural Network framework based on Ant Colony and Grey Wolf Optimization for Diagnosis of aortic aneurysm

  • 期刊论文

  • 发表日期：

    2025年10月25日

查看全文
  • Manisha Verma,
  • Jagendra Singh,
  • Sangeeta Kumari

摘要

Aortic aneurysm (AA) is still the predominant cause of global mortality and morbidity, making early and precise diagnosis necessary. In this work, a novel hybrid framework—Hybrid Attention-Augmented Deep Neural Network (HA-DNN) is proposed which is optimized using Ant Colony Optimization and Grey Wolf Optimizer (ACO-GWO) paradigm for precise diagnosis of AA. The proposed framework combines ACO to perform initial feature subset selection and GWO for fine-tuned optimisation to eliminate redundancy and enhance generalizability. The model has a two-branch structure: a deep feedforward network to learn structured clinical features and an attentional BiLSTM (Bi-directional long short term memory) network to learn time-series ECG (electrocardiograph) features. The two-branch integration makes it possible to learn robustly across heterogeneous data modalities. The proposed framework is tested on two open-access benchmark datasets: Cleveland Heart Disease Dataset and MIT-BIH Arrhythmia Dataset. The results show remarkable enhancements in terms of classification accuracy, F1-score, and generalizability when compared to existing methods.