In Studies in health technology and informatics ; h5-index 23.0
COVID-19 when left undetected can lead to a hazardous infection spread, leading to an unfortunate loss of life. It's of utmost importance to diagnose COVID-19 in Infected patients at the earliest, to avoid further complications. RT-PCR, the gold standard method is routinely used for the diagnosis of COVID-19 infection. Yet, this method comes along with few limitations such as its time-consuming nature, a scarcity of trained manpower, sophisticated laboratory equipment and the possibility of false positive and negative results. Physicians and global health care centers use CT scan as an alternate for the diagnosis of COVID-19. But this process of detection too, might demand more manual work, effort and time. Thus, automating the detection of COVID-19 using an intelligent system has been a recent research topic, in the view of pandemic. This will also help in saving the physician's time for carrying out further treatment. In this paper, a hybrid learning model has been proposed to identify the COVID-19 infection using CT scan images. The Convolutional Neural Network (CNN) was used for feature extraction and Multilayer Perceptron was used for classification. This hybrid learning model's results were also compared with traditional CNN and MLP models in terms of Accuracy, F1-Score, Precision and Recall. This Hybrid CNN-MLP model showed an Accuracy of 94.89% when compared with CNN and MLP giving 86.95% and 80.77% respectively.
Rajasekar Sakthi Jaya Sundar, Narayanan Vasumathi, Perumal Varalakshmi
CNN, COVID-19, Classification, Deep Learning, Multilayer Perceptron