An overview of deep learning techniques for COVID-19 detection: methods, challenges, and future works

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• The COVID-19 pandemic, declared by the WHO in 2020, caused significant global mortality and continues to affect millions.
• Accurate COVID-19 identification relies on RT-PCR tests or medical data analysis, which can be costly and time-consuming.
• Automated computer-aided detection methods are increasingly crucial for efficient and scalable COVID-19 diagnosis.

Purpose of the Study:

• To review the applications of deep learning (DL) and machine learning (ML) in detecting COVID-19 using diverse medical data.
• To analyze data preprocessing techniques, feature extraction, and current detection methodologies.
• To present and compare publicly available datasets and identify future research directions for enhanced COVID-19 classification.

Main Methods:

• Systematic review of studies applying DL and ML for COVID-19 detection.
• Analysis of medical data modalities including CT scans, X-rays, cough sounds, MRIs, ultrasound, and clinical markers.
• Examination of data preprocessing, feature engineering, and algorithmic approaches used in COVID-19 classification.

Main Results:

• DL and ML models demonstrate significant potential in detecting COVID-19 from various medical imaging and non-imaging data.
• A comprehensive overview of datasets, their characteristics, and comparison materials is provided.
• Key similarities, differences, gaps, and limitations in current research are highlighted.

Conclusions:

• AI-driven methods, particularly DL and ML, show promise for rapid and accurate COVID-19 detection.
• Further research is needed to address existing gaps and limitations for improved diagnostic tools.
• Future directions include refining models and exploring novel data sources for robust COVID-19 classification.