Artificial intelligence in the diagnosis and management of COVID-19: a narrative review

Abstract

Abstract: As per November 2020, there have been over 51.5 million cases of COVID-19 in the world with its mortality rate being close to 7%, causing a major burden on health care systems. Artificial intelligence (AI) is a promising tool, the use of which has been encouraged for the development of an automated diagnosis system for COVID-19 minimising the drawback of limited reverse transcription polymerase chain reaction (RT-PCR) tests. It is a time-saving, cost-effective approach, which is being promoted for reducing the physician burden during the pandemic crisis. For this narrative review, most recent data sources were collected from PubMed and Cochrane Library. Deep Learning is a promising technology for the automated diagnosis of COVID-19 through the use of advanced algorithms that identify hidden patterns on patient radiographs. Machine learning is useful in predicting patient prognosis and biomarker analysis is helpful for customised treatment planning. Infrared thermal scanners, chatbot applications, AI-based decision-making systems and image analysers are some generic contributions of AI assisting in the contactless diagnosis in suspected patients. Overall, deep neural network-based approaches have found to be superior to RT-PCR in diagnosing COVID-19 having a sensitivity of 85.35% and a specificity of 92.18% in the image-intensive diagnosis of pneumonia. In patients with comorbid conditions, telemedicine is a significant contribution of AI for monitoring and diagnosis positive cases through the use of applications such as My Day for Senior on Alexa Daily Check. Despite these advantages, the use of AI is only recommended under the guidance of the physician until sufficient clinical trials are not conducted supporting its independent use. Conclusively, the role of AI is prominent in the detection and diagnosis of COVID-19 through the use of technologies such as machine learning, deep learning and deep neural networks. However, its careful use is recommended until suitable clinical trials confirming safety are not conducted.

Ähnliche Arbeiten