Pandemic has led dramatically loss of human life and a challenge in livelihoods. An international pandemic has been caused by the contagious viral illness COVID-19. A considerable number of COVID-19 deaths are brought on by pneumonia, which is a frequent consequence of the illness. In COVID-19 patients, early diagnosis of pneumonia can enhance clinical outcomes and save lives. In this study, we suggest a modified Convolutional Neural Network-CNN model to predict patients with pneumonia and COVID-19. Dropout regularization, additional convolutional layers, and transfer learning improve the suggested model's performance. In this study attempts to develop a model that has been trained and evaluated using a publicly available dataset of chest X-ray images from COVID-19 patients, pneumonia patients, and healthy individuals.The results demonstrate that the modified CNN model outperforms the original CNN model which gives clear insight for better results on image identification.
Top1. Introductions
The severe acute respiratory syndrome (SARS-CoV-2) coronavirus is also known as COVID-19. In December 2019, Wuhan, China, reported the illness for the first time. The pandemic has since spread throughout the world (Tiwari, Pant, Elarabawy et al, 2022).
The respiratory system is the system that is most commonly impacted by COVID-19, and pneumonia is a very serious consequence that can result in hospitalization, artificial ventilation, and even deaths (Tiwari, Pant, Elarabawy et al, 2022; Tiwari, Upadhyay, Pant et al, 2022).
In COVID-19 patients, a pneumonia diagnosis is crucial for efficient management and therapy. Patients with COVID-19 are typically diagnosed with pneumonia using chest X-ray imaging. The discussion presented (Chauhan et al., 2023; El-Shafai et al., 2022; Kantheti & Javvaji, 2022; Srivastava et al., 2020; Vasal et al., 2020) depicts the use of convolutional neural network based on image analysis and diagnosis with chest X-ray images
In this research, we recommend a customized CNN model comparison along with Attention Mechanisms used in CNN in both parallel and serial ways to forecast pneumonia in Figure 3 and COVID-19 patients in Figure 1 (Nahiduzzaman et al., 2023). A number of convolution layers, a number of pooling layers, and then a specific number of fully connected neural networks are included in the suggested design of the proposed model (Srinivasarao et al., 2022).
Figure 1. (Source: Primary) Figure 2. (Source: Primary) Figure 3. (Source: Primary) Top2. Literature Review
Numerous studies have looked into the use of CNNs for the diagnosis of pneumonia and COVID-19. A Postolopoulos et al (Bhatele et al., 2022) in 2020 also suggested a CNN-based method for COVID-19 diagnosis using CT scans in order to detect COVID-19 using chest X-ray images. An accuracy of 94.38% on the validation set was attained by the model, which used a modified version of the InceptionV3 architecture. In a comparison between the model's performance and that of three radiologists, the authors found that the model outperformed two of the radiologists.
The material that is currently available on COVID-19 and CNN-based pneumonia diagnosis does have significant drawbacks, though. One major disadvantage of CNNs is that they are often utilized as “black-box” models, meaning that it can be challenging to understand the decisions made by the models (see Table 1). Another drawback is the models' poor resistance to changes in data quality and imaging modalities.