COVID-19 Diagnosis and Classification Using Radiological Imaging and Deep Learning Techniques: A Comparative Study.

Area of Science:

• Medical Imaging and Artificial Intelligence
• Radiology and Diagnostic Imaging
• Infectious Disease Diagnostics

Background:

• The COVID-19 pandemic highlighted the need for rapid and accessible diagnostic tools.
• Limitations of traditional laboratory tests, such as PCR, include shortages and delays.
• Automated systems using radiological images offer a promising complementary diagnostic approach.

Purpose of the Study:

• To review and analyze artificial intelligence (AI) based automated diagnosis systems for COVID-19 detection.
• To evaluate the effectiveness of deep learning models trained on chest X-ray (CXR) and CT images.
• To compare the performance of AI models against established COVID-19 diagnostic methods.

Main Methods:

• Review of studies employing AI, specifically Convolutional Neural Networks (CNNs), for analyzing radiological images.
• Application of transfer learning and various classification techniques (binary and multi-class).
• Training and validation of models on diverse datasets, with discussion of data attributes and performance metrics (accuracy, F1 score, AUC).

Main Results:

• AI models demonstrate high accuracy in predicting COVID-19 severity, comparable to or exceeding traditional methods.
• Challenges include limited availability of COVID-19 image data.
• Automated CXR analysis systems show reliability for initial screening and diagnosis.

Conclusions:

• AI-driven automated diagnosis systems using chest imaging are valuable tools for radiologists.
• These systems provide a cost-effective, readily available, and rapid method for COVID-19 screening.
• Further development is needed to address data limitations and enhance model generalizability.