tsta

فارسی

Home / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

Application of Attention Mechanisms in Deep Learning Models for COVID-19 Detection and Classification from Medical Images: A Systematic Review

Authors :

Jafar Abdollahi¹ Babak Nouri-Moghaddam² Abbas Mirzaei³

1- Department of Computer Engineering, CT.T, Faculty of Engineering, Central Tehran Branch, Islamic Azad University, Tehran, Iran 2- Department of Computer Engineering, Ardabil Branch, Islamic Azad University, Ardabil, Iran 3- Department of Computer Engineering, Ard.C., Islamic Azad University, Ardabil, Iran

Keywords :

COVID-19،Deep Learning،Attention Mechanisms،Medical Image Analysis،Chest X-ray،CT scan،Disease Classification

Abstract :

The COVID-19 pandemic, declared by the World Health Organization in March 2020, strained healthcare systems globally, highlighting the urgent need for rapid and accurate diagnosis to control its spread and optimize treatment. This systematic review investigates recent advancements in attention-based deep learning models for detecting and classifying COVID-19 using medical images, primarily chest X-rays and CT scans. We explore key attention mechanisms—spatial, channel, and self-attention—assessing their roles, strengths, and limitations in improving diagnostic precision. The study reviews their integration with architectures like convolutional neural networks (CNNs), Vision Transformers, and our proposed hybrid model, RMT-Net, through a comparative analysis of performance on public datasets such as COVIDx. Leading models achieve accuracies above 95%, demonstrating their efficacy. Clinically, these tools enhance triage and decision-making, yet challenges like data scarcity and interpretability persist. Future research should focus on improving data quality, developing interpretable models, and integrating multi-modal data. This review underscores the potential of attention mechanisms to revolutionize computer-aided diagnosis for COVID-19, offering a foundation for future innovations in medical imaging.