Related Experiment Video
Updated: Sep 13, 2025
04:17
DNA Virus Detection System Based on RPA-CRISPR/Cas12a-SPM and Deep Learning
Published on: May 10, 2024
880
Deep Learning Network Selection and Optimized Information Fusion for Enhanced COVID-19 Detection: A Literature
Olga Adriana Caliman Sturdza1,2, Florin Filip1,2, Monica Terteliu Baitan1,2
1Faculty of Medicine and Biological Sciences, Stefan cel Mare University of Suceava, 720229 Suceava, Romania.
Diagnostics (Basel, Switzerland)
|July 29, 2025
Summary
Deep learning models, including vision transformers, show promise for diagnosing COVID-19 lung abnormalities using chest imaging. Challenges remain in data standardization and generalization for widespread clinical use.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Pulmonology
Background:
- The COVID-19 pandemic accelerated the need for rapid diagnostic tools.
- Deep learning (DL) applications utilizing chest imaging (CXR, CT) have been extensively researched for COVID-19 detection.
Purpose of the Study:
- To review the development and performance of DL architectures for identifying COVID-19 lung abnormalities.
- To explore multimodal diagnostic systems and information fusion techniques.
Main Methods:
- Examination of deep learning architectures, including Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs).
- Analysis of multimodal diagnostic approaches incorporating lung ultrasounds, clinical data, and cough sounds.
- Review of information fusion techniques at data, feature, and decision levels.
Main Results:
- CNNs, particularly ResNet architectures, show strong performance via transfer learning.
- Vision Transformers (ViTs) offer superior performance with enhanced interpretability and lower data demands.
- Multimodal systems and information fusion improve diagnostic accuracy.
Conclusions:
- Despite promising results, challenges like limited/non-uniform datasets, domain differences, overfitting, and poor generalization hinder progress.
- Recent advancements focus on large datasets, clinical AI, and distributed learning for data security.
- Broader validation, regulatory approval, and continuous adaptation are crucial for clinical deployment and future pandemic preparedness.
