MULTI-DEEP: A novel CAD system for coronavirus (COVID-19) diagnosis from CT images using multiple convolution neural networks

摘要: Coronavirus (COVID-19) was first observed in Wuhan, China, and quickly propagated worldwide. It is considered the supreme crisis of present era one most crucial hazards threatening worldwide health. Therefore, early detection COVID-19 essential. The common way to detect reverse transcription-polymerase chain reaction (RT-PCR) test, although it has several drawbacks. Computed tomography (CT) scans can enable suspected patients, however, overlap between patterns other types pneumonia makes difficult for radiologists diagnose accurately. On hand, deep learning (DL) techniques especially convolutional neural network (CNN) classify non-COVID-19 cases. In addition, DL that use CT images deliver an accurate diagnosis faster than RT-PCR which consequently saves time disease control provides efficient computer-aided (CAD) system. shortage publicly available datasets images, CAD system's design a challenging task. systems literature are based on either individual CNN or two-fused CNNs; used segmentation classification diagnosis. this article, novel system proposed diagnosing fusion multiple CNNs. First, end-to-end performed. Afterward, features extracted from each individually classified using support vector machine (SVM) classifier. Next, principal component analysis applied feature set, network. Such sets then train SVM classifier individually. selected number components set fused compared with CNN. results show effective capable detecting distinguishing cases accuracy 94.7%, AUC 0.98 (98%), sensitivity 95.6%, specificity 93.7%. Moreover, efficient, as fusing reduced computational cost final model by almost 32%.