•  
  •  
 

Corresponding Author

Hagar Hesham Eldawoudy

Subject Area

Electronics and Communication Engineering

Article Type

Original Study

Abstract

The new coronavirus 2019 (COVID-19) pandemic has been destructive to human life and has resulted in a significant number of deaths globally. Detecting COVID-19 as early as possible is a critical way to prohibit the virus from quickly passing among people. RT-PCR is a method used for COVID-19 detection, but it is very expensive and isn’t an accurate method due to its sensitivity of 60–70%. For these reasons, deep learning and imaging techniques are combined by doctors, scientific experts, and professionals for accurate and rapid COVID-19 detection. In this work, we propose a stacked ensemble model based on Linear Regression for automatic COVID-19 detection on the chest using X-ray scans based on the techniques of deep learning. The suggested model concept is based on fusing the features of the three pre-trained models, namely ResNet-152, DenseNet-201, and Vgg-19, and its performance is higher than using single models. We collected a balanced dataset from various repositories on the "Kaggle" website and split it into training and testing sets. Our proposed model was tested on 11100 COVID-19, normal, and pneumonia images. Also, we used different optimizers for evaluations, such as Adam, Adagrad, ASGD, and SGD optimizers, using 6 epochs for the training process and a learning rate of 0.01. Using the SGD optimizer with our proposed model achieved the highest prediction accuracy, recall, precision, F1-score, and AUC for 3-class classification among other optimizers.

Keywords

Chest X-ray (CXR), Transfer Learning (TL), Artificial Intelligence (AI), Deep Learning (DL), Ensemble Model (EM), Artificial Neural Networks(ANNs), Linear Regression(LR), Deep Neural Network(DNN).

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS