<![CDATA[The COVID-19 pandemic has underscored the need for rapid and accurate diagnostic methods. Although Reverse Transcription Polymerase Chain Reaction (RT-PCR) is the gold standard for detecting COVID-19, it presents limitations such as high costs, lengthy processing times, and the requirement for specialized personnel. Medical imaging, particularly lung X-rays, offers a viable alternative for COVID-19 detection. This study evaluates five Convolutional Neural Network (CNN) models: a handcrafted CNN, VGG-16, VGG-19, ResNet50, and InceptionV3, with the aim of enhancing classification accuracy between COVID-19 and normal lung images. The dataset, obtained from Kaggle, comprises 13,808 X-ray images, which were balanced using random oversampling to address class imbalance. Data augmentation techniques were applied to improve model generalization and mitigate overfitting. After training the models for 100 epochs, the results revealed that both VGG-19 and InceptionV3 achieved the highest accuracy, each attaining 100%, outperforming the other models. VGG-16 and CNN Handcraft also demonstrated strong performance with an accuracy of 99% and 97%, whereas ResNet50 exhibited the lowest accuracy at 78%. These findings suggest that more complex CNN architectures, such as VGG- 19 and InceptionV3, are highly effective in detecting COVID-19 from X-ray images. Future research should explore additional CNN models and employ further model tuning to optimize performance.]]>
Copyrights © 2025
