ABSTRACT
It is well known that the symptoms of Coronavirus disease (COVID) and common pneumonia (CP) disease are very similar though the first one often leads to severe complications and may even be fatal. Hence, it is of vital importance to be able to correctly distinguish between the two. This paper attempts to achieve this task using whole 3-D CT scans of lungs. A number of models have been experimented with, using convolutional and radiomic features as well as their concatenations, and different classifiers (MLP and Random Forest) with two different sizes of input CT images (50 × 128 × 128 and 25 × 256 × 256 ) and their performances have been compared. The most significant contribution of this work is the postulation of a 3-D dual-scale framework using CT scans, employing both intra-scale and inter-scale information, thereby achieving performance scores which are much higher than the state of the art methods to distinguish between COVID-19 and CP using lung CT scans. Specifically, Accuracy of 98.67% and Receiver Operating Characteristics-Area Under The Curve (AUC) of 99% are worth mentioning. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.