C-LVQ: A Convolutional Neural Network with Learning Vector Quantization for the Diagnosis of Covid-19

Globally, the spread of Covid-19 started in December 2019 and led to world chaos. For the detection of Covid-19, the standard Reverse Transcription Polymerase Chain Reaction (RT-PCR) test is famous for the initial diagnosis. This standard RT-PCR test has limitations, such as being time consuming, having low sensitivity, etc. Chest X-Rays (CXR) and CT scans may also detect lung infections, which can aid doctors in detecting Covid-19. The detection of Covid-19 using CXR via artificial intelligence based diagnosis is more efficient and accurate than traditional medical practice. For the automated diagnosis of Covid-19, a hybrid of Convolutional Neural Network and Learning Vector Quantization (C-LVQ) is proposed. First, five pre-trained Convolutional Neural Network (CNN) models are selected for feature extraction, followed by Learning Vector Quantization (LVQ) for classification between Covid-19, Pneumonia, and Healthy subjects. The results show that of all the hybrid networks studied, the MobileNetV2-LVQ, a hybrid of the MobileNetV2 architecture of CNN and LVQ, has the highest accuracy of 91.61%.