Comparison of Convolutional Neural Networks in SARS-CoV-2 Identification

Severe Acute Respiratory Syndrome SARS-CoV-2 is a global pandemic that has resulted in numerous fatalities and affected millions of people worldwide. The global community has been experiencing conditions resembling lockdowns due to the COVID-19 pandemic. This public health crisis has posed a significant challenge for scientists, researchers, and healthcare professionals worldwide, extending from the virus's detection to its treatment. Healthcare professionals would greatly benefit from a technological tool that enables swift and precise screening for COVID-19 infections. Prompt recognition of this specific virus can contribute to easing the burden on healthcare systems. X-rays have demonstrated their significance in pinpointing ailments like Pneumonia. The notable advancements achieved in the realm of Machine Learning (ML) have paved the way for the development of artificial intelligent systems proficient in differentiating between COVID-19 cases and those considered normal. The latter is contributed by Deep Learning (DL) advancements. This research utilizes advanced deep learning methods, particularly the training of CNN models using the Python programming language. Its primary aim is to differentiate between chest X-ray images of COVID-19 patients and those considered normal. The CNN models which were used are VGG19, Xception and VGG16. The dataset incorporated was 400 normal X-ray chest images and 399 COVID-19 X-ray chest images. The primary performance metric employed is classification accuracy. Remarkably, the VGG19 model outperformed the others with the highest accuracy of 99%. VGG16 model achieved the accuracy 97%, while the Xception model demonstrated the lowest accuracy at 96%. The above results prove that deep learning holds a promising detection of COVID-19 from the chest X-ray images.