Effect of multimodal imaging on Covid-19 and lung cancer classification via deep learning

Abstract

The recent Coronavirus pandemic that affected every part of our society has been spreading rapidly due to its high infectious rate. Covid-19 is an illness that affects the respiratory system and its early symptoms include tiredness, cough and fever. It is generally diagnosed using reverse transcription-polymerase chain reaction (RT-PCR), and in some cases using computed tomography (CT) scans or radiography. However, the similarities in medical image structure of Covid-19 and lung cancer can lead to wrong treatment approaches. In this paper, the aim is to investigate if a deep learning model, specifically AlexNet, can accurately distinguish between lung cancer and Covid-19 from their CT and X-ray images. During this analysis, we carried out 3 different analyses, which included the classification of Covid-19 and lung cancer CT images, Covid-19 and lung cancer X-rays, and Covid-19 and lung cancer CT and X-rays. The results clearly demonstrated that deep learning was able to distinguish Covid-19 and lung cancer with very high accuracy from the CT images in comparison to X-ray and multimodal imaging. However, there was really no significant improvement as a result of multimodal imaging.