DEEP learning techniques significantly enhance the diagnosis and forecasting of COVID-19 using medical imaging, offering faster and more accurate results compared to traditional methods.
The COVID-19 pandemic, declared a global emergency by the WHO in March 2020, necessitated rapid and accurate diagnostic methods. While RT-PCR remained the gold standard, combining it with imaging modalities such as chest X-rays and CT scans enhanced diagnostic precision. However, manual interpretation of these images is labour-intensive and prone to error. This review focuses on how artificial intelligence, particularly deep learning, revolutionises COVID-19 diagnosis. Deep learning models autonomously extract, select, and classify features, addressing the limitations of conventional machine learning. By reviewing the application of deep learning in diagnosing COVID-19 from medical imaging, the study aims to highlight advancements and challenges in this field.
A systematic review was conducted using PRISMA guidelines to assess research from 2020 to 2023. The initial search yielded 400 studies on deep learning in COVID-19 imaging, with key focus areas including segmentation, classification, uncertainty quantification, and predictive modelling. Among these, methods like explainable AI and Internet of Medical Things (IoMT)-based rehabilitation systems demonstrated promise in real-time monitoring and treatment. Results consistently showed that deep learning models outperformed traditional machine learning in diagnostic accuracy and automation. For instance, studies using chest X-rays achieved over 90% accuracy in detecting COVID-19, underscoring the efficacy of computer-aided diagnostic systems (CADS).
This review concludes that deep learning offers transformative potential in medical imaging for COVID-19 diagnosis, paving the way for enhanced diagnostic tools in clinical practice. However, challenges such as data variability, model explainability, and integration into healthcare workflows remain critical. Future research should prioritise overcoming these obstacles, focusing on refining uncertainty quantification methods and improving model generalisability. The lessons learned from applying AI to COVID-19 can guide the development of robust systems for future public health crises, ensuring preparedness and resilience in clinical settings.
Reference
Shoeibi A et al. Automated detection and forecasting of COVID-19 using deep learning techniques: a review. Neurocomputing. 2024;127317.
