Prediction of COVID-19 Epidemiology with Chest Computerized Tomography Images Using Machine Learning Techniques

The COVID-19 pandemic placed immense pressure healthcare systems, emphasizing the need for efficient prediction methods. This research proposes a deep learning-based solution using Convolutional Neural Networks (CNNs) to classify chest CT scan images as either COVID-19 positive or negative. A dataset of 400 normal and 399 COVID-19 CT scan images was used to evaluate CNN models, including VGG19, DenseNet201, Xception, InceptionV3, and VGG16, over 25 and 50 epochs. At 25 epochs, DenseNet201 achieved the highest accuracy of 97%, followed by VGG16 at 96%, InceptionV3 at 93%, and VGG19 and Xception at 91%. At 50 epochs, accuracy improved, with VGG16 and DenseNet201 both reaching 98%, and VGG19 at 97%. Xception and Inception V3also saw improvements, reaching 96%. VGG16 consistently achieved 100% for both Precision and Recall across all epochs, while DenseNet201 maintained 100% Precision at 25 epochs but dropped to 95% at 50 epochs. Xception showed significant improvement in Recall, rising from 85% to 97%. Enhanced preprocessing, including noise reduction and contrast adjustment, further boosted model performance. A COVID-19 Prediction Web Application (CPWA) was developed using the fine-tuned VGG19 model to offer a practical tool for healthcare professionals. This study demonstrates the potential of AI-driven diagnostics to streamline COVID-19 screening and improve early diagnosis, providing scalable solutions for clinical settings.