COVID-19 Pneumonia Detection Using Optimized Deep Learning Techniques

Abstract

It became apparent that mankind has to learn to live with and adapt to COVID-19, especially because the developed vaccines thus far do not prevent the infection but rather just reduce the severity of the symptoms. The manual classification and diagnosis of COVID-19 pneumonia requires specialized personnel and is time consuming and very costly. On the other hand, automatic diagnosis would allow for real-time diagnosis without human intervention resulting in reduced costs. Therefore, the objective of this research is to propose a novel optimized Deep Learning (DL) approach for the automatic classification and diagnosis of COVID-19 pneumonia using X-ray images. For this purpose, a publicly available dataset of chest X-rays on Kaggle was used in this study. The dataset was developed over three stages in a quest to have a unified COVID-19 entities dataset available for researchers. The dataset consists of 21,165 anterior-to-posterior and posterior-to-anterior chest X-ray images classified as: Normal (48%), COVID-19 (17%), Lung Opacity (28%) and Viral Pneumonia (6%). Data Augmentation was also applied to increase the dataset size to enhance the reliability of results by preventing overfitting. An optimized DL approach is implemented in which chest X-ray images go through a three-stage process. Image Enhancement is performed in the first stage, followed by Data Augmentation stage and in the final stage the results are fed to the Transfer Learning algorithms (AlexNet, GoogleNet, VGG16, VGG19, and DenseNet) where the images are classified and diagnosed. Extensive experiments were performed under various scenarios, which led to achieving the highest classification accuracy of 95.63% through the application of VGG16 transfer learning algorithm on the augmented enhanced dataset with freeze weights. This accuracy was found to be better as compared to the results reported by other methods in the recent literature. Thus, the proposed approach proved superior in performance as compared with that of other similar approaches in the extant literature, and it made a valuable contribution to the body of knowledge. Although the results achieved so far are promising, further work is planned to correlate the results of the proposed approach with clinical observations to further enhance the efficiency and accuracy of COVID-19 diagnosis.

CovidConvLSTM: A fuzzy ensemble model for COVID-19 detection from chest X-rays

“…From the table, it can be observed that the highest accuracy, average precision, recall, and F1-score is achieved by the proposed CovidConvLSTM model with values 86.75%, 88.71%, 86.75%, and 86.58% respectively. The second highest accuracy of 85.50% is achieved by the Optimized CNN model Bashar et al (2021) , which is 1.25% less than that of the proposed method. Also, in terms of accuracy, the CovidConvLSTM model outperforms the Optimized CNN model in terms of average precision, recall, and F1-score by 1.21%, 1.25%, and 1.58% respectively.…”

“…In this section, the performance of the present Sugeno fuzzy integral aided ensemble method is compared with several state-of-the-art methods proposed by Makris, Kontopoulos, and Tserpes (2020) , Horry, et al (2020) , Hemdan et al (2020) , Ardakani et al (2020) , Aslan et al (2021) , Bashar et al (2021) , Chowdhury, et al (2020) , Das, Roy, et al (2021) , Goel et al (2021) , Islam et al (2020) , Ismael and Şengür (2021) , Jain et al (2021) , Kedia et al (2021) , Khan et al (2020) , Mukherjee et al (2021a) , Naeem and Bin-Salem (2021) , Panetta et al (2021) , Paul et al (2022) , Roy et al (2021) , Sedik, Hammad, Abd El-Samie, Gupta, and Abd El-Latif (2021) , Senan et al (2021) , and Yang, et al (2021) on all three datasets used here whenever applicable i.e., the results obtained by the methods on the respective datasets are cited or in few cases, their performances are evaluated using the proposed setup. The comparative results are recorded in Table 2 , Table 3 , Table 4 for dataset 1, 2 and 3 (mentioned in Table 1 ) respectively.…”

“… Work ref. Technique Performance (in %) in terms of Precision Recall F1-score Accuracy Jain et al (2021) a Xception 79.50 74.00 73.00 74.25 Ardakani et al (2020) a ResNet101 80.50 76.50 76.00 76.75 Chowdhury, et al (2020) a CheXNet 82.50 75.00 73.50 74.75 Bashar et al (2021) a Optimized CNN 87.50 85.50 85.00 85.50 Senan et al (2021) a ResNet50 80.50 77.00 76.50 77.25 Islam et al (2020) a CNN-LSTM 70.47 70.25 70.16 70.25 Mukherjee et al (2021a) a End-to-End CNN 71.00 64.00 73.50 64.00 …”

“…However, there are some researchers who used data augmentation methods explicitly to include variations in the training data, or concentrated on image enhancement to improve the quality of chest X-ray images, or both. For example, Bashar, Latif, Ben Brahim, Mohammad, and Alghazo (2021) proposed an optimized approach in three stages - initially, image enhancement was performed, followed by data augmentation, and finally, the images were fed to transfer learning algorithms using some standard CNN architectures through which the images were classified. Chowdhury, et al (2020) used image augmentation but did not employ image enhancement.…”

A Review Study of the Deep Learning Techniques used for the Classification of Chest Radiological Images for COVID-19 Diagnosis

“…Bashar et al ( Bashar et al, 2021 ) managed to train their model on bigger public datasets. Their research team optimized the Deep Learning approach for automatic classification and diagnosis, using what appears to be the largest open-source dataset on Kaggle in their range of knowledge.…”