Abstract

Respiratory infections are a confusing and time-consuming task of constantly looking at clinical pictures of patients. Therefore, there is a need to develop and improve the respiratory case prediction model as soon as possible to control the spread of disease. Deep learning makes it possible to discover a virus such as COVID-19 can be effectively detected using classification tools as CNN (Convolutional Neural Network). MFCC (Mel Frequency Cepstral Coefficients) is a common and effective classification tool. MFCC-CNN’s the proposed learning model is used to speed up the prediction process that assists medical professionals. MFCC is used to extract image features that are related to presence of COVID-19 or not. Prediction is based on convolutional neural network. This makes time-consuming process easier, faster with more accurate results reducing the spread of the virus and saves lives. Experimental results show that using a CT image converted to Mel-frequency cepstral spectrogram as an input to CNN can perform better results; with the validation data that include 99.08% accuracy for appropriate COVID categories and images with the non-COVID labels. Thus, it can probably be used to detect in CT images the presence of COVID-19. The work here provides evidence of the idea that high accuracy can be achieved with a trusted dataset, which can have a significant impact on this area.

Keywords: Biomedical imaging; COVID-19; Computed tomography; Feature extraction; MFCC; Image classification; Convolutional neural network