Prediction of COVID-19 with Computed Tomography Images using Hybrid Learning Techniques

Abstract

Reverse Transcription Polymerase Chain Reaction (RT-PCR) used for diagnosing COVID-19 has been found to give low detection rate during early stages of infection. Radiological analysis of CT images has given higher prediction rate when compared to RT-PCR technique. In this paper, hybrid learning models are used to classify COVID-19 CT images, Community-Acquired Pneumonia (CAP) CT images, and normal CT images with high specificity and sensitivity. The proposed system in this paper has been compared with various machine learning classifiers and other deep learning classifiers for better data analysis. The outcome of this study is also compared with other studies which were carried out recently on COVID-19 classification for further analysis. The proposed model has been found to outperform with an accuracy of 96.69%, sensitivity of 96%, and specificity of 98%.

Figure 1

CT scan images of a COVID-19 patient as time goes by. (a) 7th Day (as soon as symptoms show up): CT scan presents opacities formed in left lower lobe and right upper lobe. (b, c) 9th Day: CT scan depicts ground-glass opacities which are bilateral and multifocal. (d) 15th Day: CT scan presents that virus has evolved into mixture of consolidations and opacities. (e) 19th Day: CT scan shows partial disappearance of ground-glass opacities and consolidations using antiviral treatments. (f) 31st Day: CT scan shows absence of pleural effusions, pulmonary cavitation, and lymphadenopathy [–4].

Figure 2

System architecture that was proposed for screening COVID-19 chest CT scans using feature extraction and classification.

Figure 3

(a) Original COVID-19 CT scan image (b) Histogram equalized COVID-19 CT scan image (c) Weiner filtered COVID-19 CT scan image.

Figure 4

Histogram of original and equalized images.

Figure 5

Convolutional and max-pooling layer intermediate image.

Figure 6

Colormap for COVID-19-affected chest CT scan images correctly classified by AlexNet+SVM and AlexNet+Random Forest.

Figure 7

Colormap for CAP-affected chest CT scan images correctly classified by AlexNet+SVM and AlexNet+Random Forest.

Figure 8

Colormap for normal chest CT scan images correctly classified by AlexNet+SVM and AlexNet+Random Forest.

Figure 9

Images that were tested as negative by RT-PCR were actually positive cases and were correctly predicted as positive by the proposed work.