An Explainable Artificial Intelligence based Prospective Framework for COVID-19 Risk Prediction

Abstract

Given the spread of COVID-19 to vast geographical regions and populations, it is not feasible to undergo or recommend the RT-PCR based tests to all individuals with flu-like symptoms. The reach of RT-PCR based testing is still limited due to the high cost of the test and huge population in few countries. Thus, alternative methods for COVID-19 infection risk prediction can be useful. We built an explainable artificial intelligence (AI) based integrated web-based prospective framework for COVID-19 risk prediction. We employed a two-step procedure for the non-clinical prediction of COVID19 infection risk. In the first step we assess the initial risk of COVID19 infection based on carefully selected parameters associated with COVID-19 positive symptoms from recent research. Generally, X-ray scans are cheaper and easily available in most government and private health centres. Therefore, based on the outcome of the computed initial risk in first step, we further provide an optional prediction using the chest X-ray scans in the second step of our proposed AI based prospective framework. Since there is a bottleneck to undergo an expensive RT-PCR based confirmatory test in economically backward nations, this is a crucial part of our explainable AI based prospective framework. The initial risk assessment outcome is analysed in combination with the advanced deep learning-based analysis of chest X-ray scans to provide an accurate prediction of COVID-19 infection risk. This prospective web-based AI framework can be employed in limited resource settings after clinical validation in future. The cost and time associated with the adoption of this prospective AI based prospective framework will be minimal and hence it will be beneficial to majority of the population living in low-income settings such as small towns and rural areas that have limited access to advanced healthcare facilities.